# UpStation

UpStation automates GPU and IT purchasing by delivering curated content to your LLM. Explore some solutions then book a meeting once you want to engage with the OEMs.

> UpStation sells pre-priced NVIDIA GPU servers, network switches, optics and cables, and colocation, at published prices, so buyers and their AI assistants can design and price a deployment without a sales call. Prices in USD as of 2026-09-09; UpStation confirms the final price on order. Delivery: Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call).

Every price on this site is on this page. Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

## [GPU servers (8 GPUs each)](https://ai.upstation.io/gpu-servers)

| GPU | Tier | Price | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [NVIDIA B300 (HGX NVL8)](https://ai.upstation.io/p/UPS-GPU-B300-8) | Tier 1 OEM build | $572,364 | 2× 64C Intel | 2TB | 1× 960GB M.2 | 8× 800G | 13.6 → 16.96 | SYS-822GS-NB3RT-01-G2 | Supermicro | Ships 1-3 wk |
| [NVIDIA B200 (HGX)](https://ai.upstation.io/p/UPS-GPU-B200-8) | Tier 1 OEM build | out of stock industry-wide, see B300 |  |  |  |  |  |  |  |  |
| [NVIDIA H200 (HGX)](https://ai.upstation.io/p/UPS-GPU-H200-8) | Tier 1 OEM build | $319,421.64 | 2× 48C Intel | 1TB | 1× 240GB SATA | 2× 25G | 9.1 → 11.34 | SYS-821GE-TNHR | Supermicro | 6-10 wk |
| [NVIDIA RTX PRO 6000 Server Edition](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8) | Tier 1 OEM build | $191,888.10 | 2× 72C Intel | 1TB | 2× 480GB + 1× 3.84TB | 2× 10G | 6.8 → 8.47 | SYS-422GA-NRT | Supermicro | 3-5 wk |
| [NVIDIA RTX PRO 6000 Server Edition](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8-T2) | Tier 2 system builder | $163,909.60 | 2× 96C AMD | 768GB | 1× 3.84TB + 2× 1TB | 2× 25G | 6.7 → 8.37 | 4U-E10G-GENOAX2 | Octoserver | 3-5 wk |
| [NVIDIA RTX 5090 (passive server edition)](https://ai.upstation.io/p/UPS-GPU-5090-8) | Tier 2 system builder | $76,679.60 | 2× 96C AMD | 512GB | 1× 3.84TB + 1× 1TB | 2× 25G | 6.4 → 7.99 | 4U-E10G-GENOAX2 | Octoserver | 5-6 wk |

## [GPU hosts without GPUs (build your own; 8 slots)](https://ai.upstation.io/gpu-hosts)

| Host | Price | Slots | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8) | $47,884.46 | no GPUs, 8 slots | 2× 72C Intel | 512GB | 2× 960GB | 2× 25G | 1.1 → 1.38 | SYS-422GA-NRT | Supermicro | 3-5 wk |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-X13) | $37,937.10 | no GPUs, 8 slots | 2× 28C Intel | 512GB | 2× 960GB | 2× 25G | 1.7 → 2.19 | SYS-521GE-TNRT | Supermicro | 3-5 wk |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-AMD) | $39,179.76 | no GPUs, 8 slots | 2× 32C AMD | 512GB | 1× 960GB | 2× 25G | 1.0 → 1.25 | AS-5126GS-TNRT | Supermicro | 3-5 wk |
| [8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN5) | $32,919.60 | no GPUs, 8 slots | 2× 32C AMD | 512GB | 2× 960GB | 2× 25G | 1.1 → 1.31 | 4U-E10G-GENOAX2 | Octoserver | 2-3 wk |
| [8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN4) | $15,146.20 | no GPUs, 8 slots, PCIe 4.0 | 2× 32C AMD | 512GB DDR4 | 2× 960GB | 2× 25G | 1.1 → 1.34 | 4U-E10G-ROMEX2 | Octoserver | 2-3 wk |

## [Switches, cables and optics (FS.com, PicOS included, generic optics)](https://ai.upstation.io/switches)

| Ports | Price | $/port | W nameplate | Provision kW | Part | Lead time | Spec |
|---|---|---|---|---|---|---|---|
| [64× 800G OSFP](https://ai.upstation.io/p/UPS-NET-800G-64) | $44,600 | $697 | 1,800 | 2.25 | N8650-64OD, PicOS | ~2 wk | 64x 800G OSFP (or 128x 400G breakout); Broadcom Tomahawk 5, 51.2T; PicOS: RoCEv2, EVPN-VXLAN, MLAG; 5-yr warranty; options: Same spec on Dell Z9864F-ON / Enterprise SONiC quote; Same spec on Arista 7060X6-64PE / EOS quote; NVIDIA Spectrum-4 SN5600 / Cumulus quote |
| [32× 800G OSFP](https://ai.upstation.io/p/UPS-NET-800G-32) | $37,499 | $1,172 | 1,000 | 1.25 | N8650-32OD, PicOS | ~2 wk | 32x 800G OSFP (or 64x 400G breakout); Broadcom Tomahawk 5, 25.6T; PicOS; options: Dell / Arista / NVIDIA equivalent quote |
| [64× 400G QSFP-DD](https://ai.upstation.io/p/UPS-NET-400G-64) | $23,999 | $375 | 1,000 | 1.25 | N8610-64DB, PicOS | ~2 wk | 64x 400G QSFP-DD; Broadcom Tomahawk 4, 25.6T; PicOS; options: Dell Z9664F-ON / Enterprise SONiC quote; Arista 7060DX5-64S / EOS quote |
| [32× 400G QSFP-DD](https://ai.upstation.io/p/UPS-NET-400G-32) | $19,999 | $625 | 600 | 0.75 | N8610-32D, PicOS | ~2 wk | 32x 400G QSFP-DD; Broadcom Tomahawk 4, 12.8T; PicOS; options: FS N5680-32D (Tomahawk 3) budget variant −$700; Dell Z9432F-ON / Enterprise SONiC quote; Arista 7060DX5-32 / EOS quote |
| [24× 200G + 8× 400G](https://ai.upstation.io/p/UPS-NET-200G-32) | $12,400 | $517 | 600 | 0.75 | N8510-24CD8D, PicOS | Backorder to 12/2026 | 24x 200G QSFP56 + 8x 400G QSFP-DD; Broadcom Trident 4-X9; PicOS; options: No 32/64-port pure-200G box at FS; use 400G-32 with breakout quote |
| [64× 100G QSFP28](https://ai.upstation.io/p/UPS-NET-100G-64) | $11,900 | $186 | 550 | 0.69 | N5630-64C, PicOS | ~2 wk | 64x 100G QSFP28; Broadcom Tomahawk 2, 6.4T; PicOS; options: Dell Z9264F-ON / Enterprise SONiC quote; Arista 7060CX2-32S class quote |
| [32× 100G QSFP28](https://ai.upstation.io/p/UPS-NET-100G-32) | $7,919 | $247 | 250 | 0.31 | N8560-32C, PicOS | ~2 wk | 32x 100G QSFP28; Broadcom Trident 3-X7, 3.2T; PicOS; options: Dell S5232F-ON / Enterprise SONiC quote; Arista 7050CX3-32S / EOS quote |
| [48× 25G + 8× 100G](https://ai.upstation.io/p/UPS-NET-25G-48) | $6,049 | $126 | 150 | 0.19 | S5470-48Y, PicOS | ~2 wk | 48x 25G SFP28 + 8x 100G QSFP28; PicOS, MLAG; options: FS S5580-48Y (Trident 3, EVPN-VXLAN) +$1,700; Dell S5248F-ON / Enterprise SONiC quote; Arista 7050SX3-48YC8 / EOS quote |
| [48× 10G + 8× 100G](https://ai.upstation.io/p/UPS-NET-10G-48) | $4,569 | $95 | 120 | 0.15 | S5860-48SC | ~2 wk | 48x 10G SFP+ + 8x 100G QSFP28; FS OS rather than PicOS (PicOS 10G options: S5860-20SQ, S5440-12S) |
| [48× 1G RJ45 + 4× 10G](https://ai.upstation.io/p/UPS-NET-1G-48) | $1,089 | $23 | 60 | 0.07 | S3410-48TS, PicOS | ~2 wk | 48x 1GbE RJ45 + 4x 10G SFP+ uplinks; PicOS L2+ managed (VLAN, LACP, MLAG); out-of-band / BMC / IPMI layer; Fanless-class 1U, redundant PSU not standard; 5-yr warranty; options: Budget: S3540L-48T, 48x 1G + 4x 10G, PicOS L2+ (369 in stock) −$480; L3: S5810-48TS, 48x 1G + 4x 10G, PicOS L3 (US warehouse) +$490; 24-port: S3410-24TS, 24x 1G + 4x 10G, PicOS L2+ −$290; PoE+: S3410-48TS-P, 48x 1G PoE+ + 2x 10G + 2x 1G +$530; 48x 1G SFP fiber: S5810-48FS, PicOS L3 +$1,440; Non-PicOS (FS OS) same hardware: S3910-48TS L2+ $1,049 / S5810-48TS L3 $1,509 quote |

### Cables and optics (FS, generic-coded; reaches are rated maxima, OM4 for multimode)

| Speed | 2 m passive DAC | Passive reach | Active DAC / AEC | AOC 10 m (max length) | Short-reach optic, each (reach) | Long-reach optic, each (reach) | Breakout cable |
|---|---|---|---|---|---|---|---|
| 800G | $239 | 3 m | AEC 5 m $1,424 (max 5 m) | $3,011 (50 m) | SR8 $879 (50 m) | DR8 $1,319 (500 m) · 2×FR4 $1,869 (2 km) | 800G→2×400G 2 m $299 |
| 400G | $174 | 3 m | Active DAC 5 m $611 (max 5 m) | $1,055 (30 m) | SR8 $219 (100 m) | DR4 $549 (500 m) · FR4 $719 (2 km) · LR4 $999 (10 km) | 400G→4×100G 2 m $199 |
| 200G | $120 | 3 m | Active DAC 7 m $325 (max 7 m) | $1,034 (50 m) | SR4 $299 (100 m) | FR4 $1,249 (2 km) · LR4 $2,624 (10 km) | 200G→2×100G 2 m $119 |
| 100G | $54 | 5 m | Active DAC 9 m $300 (max 9 m) | $221 (30 m) | SR4 $99 (100 m) | CWDM4 $209 (2 km) · LR4 $399 (10 km) | 100G→4×25G 3 m $136 |
| 25G | $39 | 5 m | Active DAC 9 m $73 (max 9 m) | $73 (30 m) | SR $49 (100 m) | LR $74 (10 km) | — |
| 10G | $18 | 7 m | Active DAC 10 m $105 (max 10 m) | $39 (30 m) | SR $25 (300 m) | LR $34 (10 km) | — |
| 1G | RJ45, Cat6 patch | — | — | — | 10G SFP+ uplinks: see 10G row |  | — |

### Fibre patch (generic FS)

| Cable | 1 m | 3 m | 10 m | 30 m | 50 m | Used with |
|---|---|---|---|---|---|---|
| LC-LC duplex OM4 | $8.40 | $10 | $20 | $47 | — | SR (10G, 25G) |
| MTP-12 OM4 | $59 | $89 | $150 | $301 | $452 | SR4, SR8, DR4, DR8 |
| LC-LC duplex OS2 | $8.40 | — | — | — | — | FR4, CWDM4, LR, LR4 |

### Link by distance

| Distance | Use | Cost per link |
|---|---|---|
| Inside passive reach | 2 m passive DAC | 1 cable |
| To active max | Active DAC or AEC | 1 cable |
| To AOC max, or to short-reach optic reach | AOC, or 2× short-reach optic + multimode fibre | 1 cable, or 2 optics + fibre |
| Beyond | 2× long-reach optic + single-mode fibre | 2 optics + fibre |
| Any | Connected cost per port = switch $/port + link |  |

Choosing a link by distance: 2 m passive DAC inside passive reach; active DAC or AEC up to its maximum; beyond that an AOC (10 m priced; longer lengths cost more) or two short-reach optics plus multimode fibre out to the optic's reach; past that, two long-reach optics over single-mode. At 100G and up the optics, not the fibre, set the cost. Connected cost per port = switch $/port + one cable, or + two optics plus fibre. GPU-fabric links are estimated all-optical: GPU node density (1–3 HGX nodes per 40 kW rack) puts most nodes past passive-copper reach of their leaf, since a 3 m DAC spans at most the adjacent rack. Mixing copper in is realistic — expect 25% or more of fabric cables to land as passive DAC once installation is engineered (peer links, co-racked leaf↔spine, nodes in or beside the leaf's rack) — but which links qualify depends on rack elevations, so estimates stay optical and copper is taken as a savings at engineering time.

| PicOS | 100G | 400G | 800G |
|---|---|---|---|
| Licence | included | included | included |
| 1-yr support | $369 | $439 | $659 |
| Dell/SONiC or Arista, same ports | quote | quote | quote |

## [Compute nodes](https://ai.upstation.io/compute-nodes)

| Product | Price | Form | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [1U compute node](https://ai.upstation.io/p/UPS-CN-INT-1U) | $33,298.33 | 1U | 2× 32C Intel | 512GB | diskless, 12 NVMe bays | 2× 25G | 1.0 → 1.2 | SYS-122C-TN | Supermicro | 3-5 wk |

## [Storage nodes](https://ai.upstation.io/storage-nodes)

| Product | Price | Form | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [2U all-NVMe storage node](https://ai.upstation.io/p/UPS-SN-INT-NVME) | $182,994.59 | 2U | 2× 32C Intel | 512GB | 24× 7.68TB NVMe, 184TB | 2× 100G | 1.3 → 1.65 | SYS-222C-TN | Supermicro | 3-5 wk |
| [2U HDD storage node](https://ai.upstation.io/p/UPS-SN-AMD-HDD) | $51,743.45 | 2U | 1× 16C AMD | 128GB | 24× 24TB HDD, 576TB | 2× 100G | 0.6 → 0.78 | ASG-2015S-E1CR24H | Supermicro | 3-5 wk |

## [Parts rate card](https://ai.upstation.io/rate-card) (marginal; apply to a base build)

| Item | Rate | Typical |
|---|---|---|
| DDR5-6400 RDIMM 32GB (Supermicro) | $46.06/GB | $1,474 per DIMM (MEM-DR532L-SL02-ER64), identical on every chassis probed; DDR5-5600 32GB $1,314 ($41.06/GB) on X13 |
| DDR5-6400 RDIMM 64GB (Supermicro) | $39.53/GB | $2,530 per DIMM; cheapest per GB |
| DDR5-6400 RDIMM 96GB (Supermicro) | $46.08/GB | $4,424 per DIMM |
| DDR5-6400 RDIMM 128GB (Supermicro) | $47.23/GB | $6,045 per DIMM; 256GB $10,311 ($40.28/GB, 09-03) |
| DDR5-4800 RDIMM 32GB (Octoserver) | $39.53/GB | $1,265 per DIMM |
| DDR4-3200 RDIMM 32GB (Octoserver Rome host) | $13.44/GB | $430 per DIMM |
| AMD EPYC 9005/9004, mainstream 16-96C (Supermicro) | $31–95/core | $66/core |
| Intel Xeon 6700P, mainstream 32-64C (Supermicro) | $101–166/core | $124/core |
| Intel Xeon 6900P, 72-128C (Supermicro) | $58–82/core | $65/core |
| AMD EPYC 9004/9005 at Octoserver | $29–51/core | $42/core |
| AMD EPYC 7003 at Octoserver (PCIe 4.0 host) | $13–29/core | $20/core |
| Nearline HDD 16-30TB SAS/SATA (Supermicro) | $49–63/TB | $58/TB |
| NVMe Gen4 U.2 TLC 1.9-15TB (Supermicro) | $690–1390/TB | $1006/TB |
| NVMe Gen5 U.2/E1.S/E3.S (Supermicro) | $715–1029/TB | $821/TB |
| SATA/SAS SSD (Supermicro) | $678–1050/TB | $800/TB |
| Exotic parts: Xeon 6768P/6788P/6748P, 3DS DIMMs, dual-port drives | quote |  |

Example: 1TB → 2TB with 64GB DIMMs = 1,024 GB × $39.53 = $40,478.72 (the H200 line lists this option at +$40,480). Marginal rates from non-exotic SKUs, taken from the eStore configurator (the price its cart charges). Within a family the price rises almost linearly with cores and capacity, so alternative configurations are not priced separately: apply these rates to the base build. Supermicro's drive pricing is high: buyers who source NVMe or HDD elsewhere should price drives from the market and the chassis from here. Exotic parts (Xeon 6768P/6788P/6748P at $250-320/core, 3DS 256GB DIMMs, dual-port or FIPS drives) sit outside these bands.

## [Power and colo](https://ai.upstation.io/power)

| Rule | Value |
|---|---|
| Provisioned kW | nameplate W ÷ 0.8 (no circuit above 80%) |
| Feeds | A + B, each sized for the whole box |
| Colo, 5-year term | $175/kW/month on provisioned kW, US or Canada |
| Colo, 3-year term | $225/kW/month on provisioned kW, US or Canada |
| Colo, 1-year term | $250/kW/month on provisioned kW, US or Canada |

| Circuit | Rated kW |
|---|---|
| 208V 1-ph 20A | 4.2 |
| 208V 1-ph 30A | 6.2 |
| 208V 3-ph 30A | 10.8 |
| 208V 3-ph 50A | 18.0 |
| 208V 3-ph 60A | 21.6 |
| 415V 3-ph 30A | 21.6 |
| 415V 3-ph 60A | 43.1 |
| 415V 3-ph 100A | 71.9 |

Nameplate max for PCIe GPU servers = (GPU TDP x count + CPU TDP x 2 + 8 W per DIMM + 15 W per drive + 400 W fans/board/NICs) / PSU efficiency (Titanium 0.94, Platinum 0.92); HGX servers use the vendor configurator total, which matches NVIDIA system guidance. Vendor configurator totals for PCIe boxes are kept in each product's basis as a cross-check only (Supermicro pads chassis overhead by 1.5-2.4 kW; Octoserver omits fans and PSU loss). Minimum circuit kW = nameplate / 0.8 so the circuit never runs above 80% (NEC continuous). Pick the smallest standard circuit whose rated kW (V x A x sqrt3 for 3-phase) >= that minimum; provision it twice (A+B) so either feed carries the box alone. Colo kW to buy = nameplate / 0.8; monthly colo = that kW x the term rate ($175 / $225 / $250 per kW-month for 5 / 3 / 1-year terms). Quoted to UpStation 2026-09-08, available in Canada and the US: $175/kW/mo on a 5-year term, $225 on 3-year, $250 on 1-year. Billed on the provisioned kW (nameplate / 0.8).

### Per box (computed)

| Item | Nameplate W | Provision kW | Circuit (A+B) | Colo/mo 5-yr | Colo/mo 3-yr | Colo/mo 1-yr |
|---|---|---|---|---|---|---|
| [8x NVIDIA B300 (HGX B300 NVL8), Intel](https://ai.upstation.io/p/UPS-GPU-B300-8) | 13,566 | 16.96 | 208V 3-ph 50A ×2 | $2,968 | $3,816 | $4,240 |
| [8x NVIDIA H200 (HGX H200), Intel](https://ai.upstation.io/p/UPS-GPU-H200-8) | 9,069 | 11.34 | 208V 3-ph 50A ×2 | $1,984 | $2,552 | $2,835 |
| [8x NVIDIA RTX PRO 6000 Blackwell Server Edition, Intel](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8) | 6,780 | 8.47 | 208V 3-ph 30A ×2 | $1,482 | $1,906 | $2,118 |
| [8x NVIDIA RTX PRO 6000 Blackwell Server Edition, AMD](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8-T2) | 6,692 | 8.37 | 208V 3-ph 30A ×2 | $1,465 | $1,883 | $2,092 |
| [8x NVIDIA RTX 5090 (passive server edition), AMD](https://ai.upstation.io/p/UPS-GPU-5090-8) | 6,389 | 7.99 | 208V 3-ph 30A ×2 | $1,398 | $1,798 | $1,998 |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8) | 1,100 | 1.38 | 208V 1-ph 20A ×2 | $241 | $310 | $345 |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-X13) | 1,748 | 2.19 | 208V 1-ph 20A ×2 | $383 | $493 | $548 |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-AMD) | 1,000 | 1.25 | 208V 1-ph 20A ×2 | $219 | $281 | $312 |
| [8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN5) | 1,050 | 1.31 | 208V 1-ph 20A ×2 | $229 | $295 | $328 |
| [8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN4) | 1,074 | 1.34 | 208V 1-ph 20A ×2 | $234 | $302 | $335 |
| [64× 800G OSFP switch](https://ai.upstation.io/p/UPS-NET-800G-64) | 1,800 | 2.25 | 208V 1-ph 20A ×2 | $394 | $506 | $562 |
| [32× 800G OSFP switch](https://ai.upstation.io/p/UPS-NET-800G-32) | 1,000 | 1.25 | 208V 1-ph 20A ×2 | $219 | $281 | $312 |
| [64× 400G QSFP-DD switch](https://ai.upstation.io/p/UPS-NET-400G-64) | 1,000 | 1.25 | 208V 1-ph 20A ×2 | $219 | $281 | $312 |
| [32× 400G QSFP-DD switch](https://ai.upstation.io/p/UPS-NET-400G-32) | 600 | 0.75 | 208V 1-ph 20A ×2 | $131 | $169 | $188 |
| [24× 200G + 8× 400G switch](https://ai.upstation.io/p/UPS-NET-200G-32) | 600 | 0.75 | 208V 1-ph 20A ×2 | $131 | $169 | $188 |
| [64× 100G QSFP28 switch](https://ai.upstation.io/p/UPS-NET-100G-64) | 550 | 0.69 | 208V 1-ph 20A ×2 | $121 | $155 | $172 |
| [32× 100G QSFP28 switch](https://ai.upstation.io/p/UPS-NET-100G-32) | 250 | 0.31 | 208V 1-ph 20A ×2 | $54 | $70 | $78 |
| [48× 25G + 8× 100G switch](https://ai.upstation.io/p/UPS-NET-25G-48) | 150 | 0.19 | 208V 1-ph 20A ×2 | $33 | $43 | $48 |
| [48× 10G + 8× 100G switch](https://ai.upstation.io/p/UPS-NET-10G-48) | 120 | 0.15 | 208V 1-ph 20A ×2 | $26 | $34 | $38 |
| [48× 1G RJ45 + 4× 10G switch](https://ai.upstation.io/p/UPS-NET-1G-48) | 60 | 0.07 | 208V 1-ph 20A ×2 | $12 | $16 | $18 |
| [1U compute node, dual socket, 64 cores, 512GB, diskless NVMe (virtualisation host)](https://ai.upstation.io/p/UPS-CN-INT-1U) | 960 | 1.2 | 208V 1-ph 20A ×2 | $210 | $270 | $300 |
| [2U all-NVMe storage node, dual socket, 24x 7.68TB NVMe (184TB raw)](https://ai.upstation.io/p/UPS-SN-INT-NVME) | 1,320 | 1.65 | 208V 1-ph 20A ×2 | $289 | $371 | $412 |
| [2U HDD storage node, single socket, 24x 24TB SAS (576TB raw)](https://ai.upstation.io/p/UPS-SN-AMD-HDD) | 620 | 0.78 | 208V 1-ph 20A ×2 | $136 | $176 | $195 |

## [How solutions are sized](https://ai.upstation.io/sizing) (digest)

- Size a GPU cluster on accelerators (GPU count and aggregate GPU memory), never on host cores. Every listed GPU server has 8 GPUs: nodes = ceil(GPUs / 8). Nodes ship fully populated; partial population is not offered. Model memory: required GPU memory (GB) = parameters (billions) × bytes per parameter × 1.2 (weights plus overhead) + KV cache (grows with context length × batch).
- HGX or PCIe: Does the model at its precision fit one GPU's memory, and can each GPU work alone? Yes: PCIe. No: HGX. Distributed training is never served by PCIe servers. An HGX cluster of 2 or more nodes carries a dedicated non-blocking GPU fabric (1 port per GPU), a storage tier sized to feed the GPUs, and up to 2× the power per node of a PCIe server; a PCIe inference deployment carries neither the fabric nor the parallel-file tier.
- Networks: GPU fabric 1 port per GPU (H200 400G, B300 800G) at 1:1, never carrying storage: one switch while the ports fit it, else on R-port switches leaves = ceil(ports ÷ (R/2)), spines = max(2, ceil(leaves × R/2 ÷ R)) (32 nodes on 64-port switches: 8 leaves + 4 spines). Front-end 2 ports per node at 3:1 on an MLAG pair; OOB 1× 1G per node and per switch. 1 cable per link; fibre links need 2 transceivers, DAC/AOC none. Estimate every GPU-fabric link as optical (2× short-reach optics + fibre, or AOC): node density puts most nodes in a different rack from their leaf, past passive-DAC reach.
- Storage: read GB/s = GPUs × 0.15 (text) to 0.8 (vision), aggregate; write = max(½ read, checkpoint GB ÷ pause s); plannable TB = raw × protection efficiency (RAID 6 10+2 0.833, EC 4+2 0.667, replica 3 0.333; Ceph also × (N−1)/N) × 0.85; nodes = max(by GB/s, by TB, protection floor). Per listed node: NVMe 184 TB raw, 20 GB/s planning; HDD 576 TB raw, 5 GB/s. Storage software is buyer-supplied.
- Racks: 40 U and 40 kW per rack unless the site says otherwise; nodes per rack = min(by U, by kW); A+B feeds. Management: 3× the 1U compute node from 4 GPU nodes up, one OOB switch per rack.

| GPU server | RU | Provisioned kW | Per 40 kW rack | Rack kW | Binds on |
|---|---|---|---|---|---|
| 8x NVIDIA B300 (HGX B300 NVL8), Intel | 8 | 16.96 | 2 | 33.92 | power |
| 8x NVIDIA H200 (HGX H200), Intel | 8 | 11.34 | 3 | 34.02 | power |
| 8x NVIDIA RTX PRO 6000 Blackwell Server Edition, Intel | 4 | 8.47 | 4 | 33.88 | power |
| 8x NVIDIA RTX PRO 6000 Blackwell Server Edition, AMD | 4 | 8.37 | 4 | 33.48 | power |
| 8x NVIDIA RTX 5090 (passive server edition), AMD | 4 | 7.99 | 5 | 39.95 | power |
| 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX) | 4 | 1.38 | 10 | 13.8 | space |
| 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX) | 5 | 2.19 | 8 | 17.52 | space |
| 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD | 5 | 1.25 | 8 | 10.0 | space |
| 8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD | 4 | 1.31 | 10 | 13.1 | space |
| 8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD | 4 | 1.34 | 10 | 13.4 | space |

Full rules, feed-rate tables, protection math and the questions to settle first: [How solutions are sized](https://ai.upstation.io/sizing). Worked designs computed with them (a 4× B300 cluster with storage; 2 PB usable on HDD nodes): [Reference builds and worked designs](https://ai.upstation.io/designs).

## [Reference builds and worked designs](https://ai.upstation.io/designs) (computed from this catalog)

| Build | Hardware | Provisioned kW | Colo term | Colo/month | Hardware + colo over term |
|---|---|---|---|---|---|
| [One 8× H200 node, 3-year colo](https://ai.upstation.io/designs) | $319,421.64 | 11.34 | 3-year | $2,552 | $411,293.64 |
| [Four 8× B300 nodes, 5-year colo](https://ai.upstation.io/designs) | $2,289,456 | 67.84 | 5-year | $11,872 | $3,001,776 |
| [Two 8× RTX PRO 6000 nodes, one 48× 25G switch, 1-year colo](https://ai.upstation.io/designs) | $390,763.20 | 17.13 | 1-year | $4,282 | $442,147.20 |
| [Worked design: 4× 8× B300 training cluster (32 GPUs), 800G fabric, 200G front-end, 5× NVMe storage nodes, 5-year colo](https://ai.upstation.io/designs) | $3,353,598.95 | 78.91 | 5-year | $13,809 | $4,182,138.95 |
| [Worked design: 2 PB usable archive on 5× HDD storage nodes (RAID 6), 100G front-end pair, 5-year colo](https://ai.upstation.io/designs) | $276,292.25 | 4.59 | 5-year | $803 | $324,472.25 |

## Not in this catalog (ask for a quote)

- NVIDIA B200 (HGX) servers: out of stock industry-wide; B300 is the listed alternative
- GB200 / GB300 NVL72 rack-scale systems and liquid-cooled racks
- AMD Instinct (MI300/MI325/MI355) servers
- 2-GPU and 4-GPU servers (every listed GPU server has 8 GPUs; hosts without GPUs have 8 slots)
- Dell, HPE and Lenovo priced builds (Tier 1 pricing shown is Supermicro; other OEMs on request)
- InfiniBand switches, and Arista or Dell/SONiC Ethernet switches (quote)
- Vendor-coded optics (listed optics are generic-coded for the FS switches)
- Racks, PDUs, structured cabling and installation labour
- Software licences other than the PicOS switch OS; managed services; GPU cloud or rental

## Questions and answers

- **What is UpStation?** UpStation sells pre-priced NVIDIA GPU servers, network switches, optics and cables, and colocation, at published prices, so buyers and their AI assistants can design and price a deployment without a sales call. Delivery: Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call).
- **What does an 8× NVIDIA B300 server cost?** $572,364 for the Tier 1 OEM build (Supermicro SYS-822GS-NB3RT-01-G2: 8× B300 · 2× 64C Intel · 2TB · 1× 960GB M.2 · 8× 800G), ships 1-3 wk. It draws 13,566 W nameplate, so provision 16.96 kW; colocation is $2,968/mo 5-yr · $3,816/mo 3-yr · $4,240/mo 1-yr.
- **What does an 8× NVIDIA H200 server cost?** $319,421.64 for the Tier 1 OEM build (Supermicro SYS-821GE-TNHR: 8× H200 · 2× 48C Intel · 1TB · 1× 240GB SATA · 2× 25G), 6-10 wk. Provision 11.34 kW; colocation $1,984/mo 5-yr · $2,552/mo 3-yr · $2,835/mo 1-yr.
- **What is the cheapest 8-GPU server listed?** $76,679.60 for 8× RTX 5090 (passive server edition) from Octoserver; the cheapest 8× RTX PRO 6000 build is $163,909.60 (Tier 2). A host with 8 GPU slots and no GPUs starts at $15,146.20.
- **What circuit does an 8× RTX PRO 6000 server need?** Any 8x RTX PRO 6000 server at 600 W caps: plan 6.8 kW nameplate (5.5 kW with cards capped at 450 W), provision 8.5 kW, one 208V 3-phase 30A circuit pair (A+B) per server; where three-phase is unavailable, 208V single-phase 50A. A 30A single-phase (L6-30) cannot carry it even at 450 W caps. Four servers per 40 kW rack. 200-240 V input only; the box needs four 3000-3200 W supplies in 3+1 for redundancy (Dell XE7745 matrix, ASUS/Supermicro/Exxact/GIGABYTE all ship that). Computed per build on the [power page](https://ai.upstation.io/power); the Tier 1 build is 6,780 W nameplate, 8.47 kW provisioned.
- **What does colocation cost?** Colocation is billed on provisioned kW: $175/kW/month on a 5-year term, $225/kW/month on a 3-year term, $250/kW/month on a 1-year term, US or Canada. Provisioned kW = nameplate W ÷ 0.8 so no circuit runs above 80%, with A and B feeds each sized for the whole box. Example: one B300 node at 16.96 kW is $2,968/mo 5-yr · $3,816/mo 3-yr · $4,240/mo 1-yr.
- **How are prices set, and how do I get a firm quote?** Every price is the vendor's published price (Supermicro, Octoserver, FS.com) in USD as of 2026-09-09. UpStation confirms the final price on order. [Book a meeting](https://calendly.com/alecglover/intro-call) with the bill of materials ready.
- **Can I get a different configuration?** Each class has one priced base build. Intel and AMD equivalents are listed with a host delta. Memory, CPU and drive changes are priced with the rate card (marginal $/GB, $/core, $/TB) applied to the base build; options such as NICs, drives and warranty have signed price deltas. Anything marked quote, or outside the rate-card bands, is quoted by UpStation.
- **How many storage nodes does 2 PB usable need?** With RAID 6 in 10+2 groups (0.833 efficiency) and an 85% fill ceiling, one 2U HDD storage node, single socket, 24x 24TB SAS (576TB raw) plans to 407.8 TB, so 2,000 TB needs 5 nodes (2,039.0 TB plannable, $258,717.25 for the nodes). Distributed with Ceph erasure coding 4+2 it takes 8 nodes (2,285.9 TB plannable) because Ceph also keeps (N−1)/N in reserve for self-healing and needs at least 6 hosts. Storage software is buyer-supplied; see the worked design.
- **How many switches does a B300 cluster need?** One 800G port per GPU at 1:1. 4 nodes (32 ports) fit one 32× 800G OSFP switch ($37,499); 8 nodes fit one 64-port switch; 32 nodes (256 ports) need 8 leaves and 4 spines on 64-port switches, 256 host links and 256 uplinks. Add a front-end switch pair and one 48× 1G OOB switch per rack; the worked design prices the 4-node case with storage.
- **How many GPU servers fit in a rack?** At 40 kW and 40 U per rack: 2× B300 (33.92 kW), 3× H200 (34.02 kW), 4× RTX PRO 6000 Tier 1 (33.88 kW), 4× RTX PRO 6000 Tier 2 (33.48 kW), 5× RTX 5090 (39.95 kW). Power binds before space in every case; a site limit replaces the 40 kW.

## Notes

- Prices as of 2026-09-09; final price confirmed on order. Continental US delivery; Canada case by case.
- Intel and AMD builds for every node; memory and drives set the price. Exotic parts: quote.
- 3 years parts and labour included; on-site and extended warranty are options.
- Draft, estimate and price-pending lines are marked; everything else is confirmed.
- 2026-09-08: First public catalog: B300, H200, RTX PRO 6000 (Tier 1 and Tier 2), RTX 5090, three GPU hosts without GPUs, FS switches with cables and optics, compute and storage nodes, rate card, power and colo model.
- 2026-09-09: Site rebuilt for people and AI assistants: visible tables, Markdown twin (catalog.md), short brief (brief.txt), computed power and colo fields and reference builds in catalog.json, JSON-LD, robots.txt and sitemap. Prices unchanged.
- 2026-09-09: Solution sizing added: how many nodes, switches, cables, storage nodes and racks a design needs (GPU count → nodes, fat-tree switch counts, storage protection math, racks and power), with two computed worked examples. Rack units and GPU memory per server added. Prices unchanged.
- 2026-09-09: Intel Xeon 6 host (SYS-422GA-NRT, 2x 6944P, TDX-capable) replaces the AMD AS-5126GS-TNRT as the Tier 1 host without GPUs; the AMD host stays as an equivalent. 48x 10G switch priced (S5860-48SC). Draft markers removed from every vendor-priced line. The AMD host is also kept as its own line.
- 2026-09-09: Supermicro lines re-priced from the eStore configurator (the price its cart charges) after an audit found that the AMD 5U host, the 1U compute node, both storage nodes, the H200 option deltas and the memory and drive rate-card rows had been priced from a stale secondary price list: hosts and the compute node moved a few percent, the storage nodes rose substantially because Supermicro's drive pricing is two to three times market (rate card updated). New Intel 5th Gen Xeon host (SYS-521GE-TNRT, 2x Gold 5520+, TDX) $37,937.10; Intel Xeon 6 host $47,884.46.
- 2026-09-09: Fabric cabling rule made explicit: GPU node density (1–3 HGX nodes per 40 kW rack) puts most nodes in a different rack from their leaf switches, so estimates price every GPU-fabric link as optical (2× short-reach optics + fibre, or AOC); passive DAC is a final-engineering substitution for proven in-rack or adjacent-rack links (25%+ of fabric cables landing as copper is a reasonable expectation, taken as a savings then, not in the estimate). Worked design updated to match. Prices unchanged.
- Generated 2026-09-22; prices as of 2026-09-09; re-checked monthly.

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# GPU servers (8 GPUs each)

> UpStation catalog · Home › GPU servers (8 GPUs each) · https://ai.upstation.io/gpu-servers

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

GPU servers: the table below is the same as on the [homepage](https://ai.upstation.io/); each product is then listed in full, and each has its own page (linked from its name). Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

| GPU | Tier | Price | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [NVIDIA B300 (HGX NVL8)](https://ai.upstation.io/p/UPS-GPU-B300-8) | Tier 1 OEM build | $572,364 | 2× 64C Intel | 2TB | 1× 960GB M.2 | 8× 800G | 13.6 → 16.96 | SYS-822GS-NB3RT-01-G2 | Supermicro | Ships 1-3 wk |
| [NVIDIA B200 (HGX)](https://ai.upstation.io/p/UPS-GPU-B200-8) | Tier 1 OEM build | out of stock industry-wide, see B300 |  |  |  |  |  |  |  |  |
| [NVIDIA H200 (HGX)](https://ai.upstation.io/p/UPS-GPU-H200-8) | Tier 1 OEM build | $319,421.64 | 2× 48C Intel | 1TB | 1× 240GB SATA | 2× 25G | 9.1 → 11.34 | SYS-821GE-TNHR | Supermicro | 6-10 wk |
| [NVIDIA RTX PRO 6000 Server Edition](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8) | Tier 1 OEM build | $191,888.10 | 2× 72C Intel | 1TB | 2× 480GB + 1× 3.84TB | 2× 10G | 6.8 → 8.47 | SYS-422GA-NRT | Supermicro | 3-5 wk |
| [NVIDIA RTX PRO 6000 Server Edition](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8-T2) | Tier 2 system builder | $163,909.60 | 2× 96C AMD | 768GB | 1× 3.84TB + 2× 1TB | 2× 25G | 6.7 → 8.37 | 4U-E10G-GENOAX2 | Octoserver | 3-5 wk |
| [NVIDIA RTX 5090 (passive server edition)](https://ai.upstation.io/p/UPS-GPU-5090-8) | Tier 2 system builder | $76,679.60 | 2× 96C AMD | 512GB | 1× 3.84TB + 1× 1TB | 2× 25G | 6.4 → 7.99 | 4U-E10G-GENOAX2 | Octoserver | 5-6 wk |

## [8x NVIDIA B300 (HGX B300 NVL8), Intel](https://ai.upstation.io/p/UPS-GPU-B300-8): $572,364

- $572,364 — 8× B300 · 2× 64C Intel · 2TB · 1× 960GB M.2 · 8× 800G · 13.6 kW nameplate — SYS-822GS-NB3RT-01-G2, Supermicro — Ships 1-3 wk
    spec: 8U, NVIDIA HGX B300 NVL8 8-GPU (5th-gen NVLink, 1.8TB/s); 2x Xeon 6768P 64C 2.4GHz (128 cores); 2TB DDR5-6400 (16x128GB); 1x 960GB M.2 NVMe boot; hot-swap bays free; 8-port 800G IB/XDR800 or 16-port 400GbE on-board; Redundant 6600W Titanium (6x, 3+3); 3-yr parts and labor + 1-yr cross-ship
    options (price delta): +2× 3.84TB E1.S NVMe +$6,914; +2× 200G CX7 (4× 200G ports) north-south +$3,848; +8× 7.68TB E1.S NVMe (61TB) +$48,384; +2× 400G CX7 north-south +$3,061; Warranty: 3 yr + 2 yr cross-ship +$42,927; Warranty: 3 yr + 3 yr cross-ship +$85,855; Warranty: 4 yr + 4 yr cross-ship +$128,782; Warranty: 5 yr + 5 yr cross-ship +$171,709; On-site Silver NBD, 3 yr +$57,985
    power: 13,566 W nameplate; provision 16.96 kW (nameplate ÷ 0.8); circuit 208V 3-ph 50A (18.0 kW rated) ×2 A+B; colo $2,968/mo 5-yr · $3,816/mo 3-yr · $4,240/mo 1-yr
    per GPU: $71,545.50 (8 GPUs)
    page: https://ai.upstation.io/p/UPS-GPU-B300-8

## [8x NVIDIA B200 (HGX B200), Intel or AMD](https://ai.upstation.io/p/UPS-GPU-B200-8)

Out of stock industry-wide; see [8x NVIDIA B300 (HGX B300 NVL8), Intel](https://ai.upstation.io/p/UPS-GPU-B300-8). Not stocked. Mostly out of stock everywhere; B300 recommended.

## [8x NVIDIA H200 (HGX H200), Intel](https://ai.upstation.io/p/UPS-GPU-H200-8): $319,421.64

- $319,421.64 — 8× H200 · 2× 48C Intel · 1TB · 1× 240GB SATA · 2× 25G · 9.1 kW nameplate — SYS-821GE-TNHR, Supermicro — 6-10 wk
    spec: 8U, NVIDIA HGX H200 8-GPU (8x 141GB HBM3e, NVLink/NVSwitch); 2x Xeon Platinum 8558 48C 2.1GHz; 1TB DDR5 (16x64GB, 16 of 32 slots free); 1x 240GB SATA boot; 15x 2.5in bays (12 NVMe + 3 SATA) free; 1x 25G 2x SFP28 CX-4; 8x PCIe 5.0 GPU + 9x PCIe 5.0 slots; 6x 3000W Titanium (4+2); 3-yr parts and labor + 1-yr cross-ship; rail kit
    options (price delta): Memory 1TB -> 2TB (+16x64GB) +$40,480; Memory 1TB -> 3TB (32x96GB) +$101,088; 8x 400G CX7 OSFP compute fabric NICs +$12,248; +8x 3.84TB U.2 NVMe (31TB) +$27,656; Boot 240GB -> 2x 1.92TB SATA PM893 +$2,119; CPU 2x 8558 48C -> 2x 8592V 64C quote; TPM 2.0 +$66; On-site Silver NBD, 3 yr (approx 10-16% of system) quote
    power: 9,069 W nameplate; provision 11.34 kW (nameplate ÷ 0.8); circuit 208V 3-ph 50A (18.0 kW rated) ×2 A+B; colo $1,984/mo 5-yr · $2,552/mo 3-yr · $2,835/mo 1-yr
    per GPU: $39,927.71 (8 GPUs)
    page: https://ai.upstation.io/p/UPS-GPU-H200-8

## [8x NVIDIA RTX PRO 6000 Blackwell Server Edition, Intel](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8): $191,888.10

- $191,888.10 — 8× RTX PRO 6000 · 2× 72C Intel · 1TB · 2× 480GB + 1× 3.84TB · 2× 10G · 6.8 kW nameplate — SYS-422GA-NRT, Supermicro — 3-5 wk
    spec: 4U X14 PCIe 5.0 dual-root, 8x double-width 600W slots; 8x RTX PRO 6000 SE 96GB (768GB GDDR7 total), factory-installed; 2x Xeon 6960P 72C 2.7GHz; 1TB DDR5-6400 (16x64GB, 8 slots free); 2x 480GB M.2 boot + 1x 3.84TB E3.S NVMe (8 bays); 2x 10GbE RJ45 on-board; PCIe slots free for 25/100/400G; TPM 2.0, rail kit; 4x 3200W Titanium (3+1); 3-yr parts and labor + 1-yr cross-ship
    options (price delta): Memory 1TB -> 1.5TB (24x64GB) +$19,336; Memory 1TB -> 2TB (16x128GB) +$55,248; Swap QLC E3.S for 3.84TB TLC U.2 (PM9A3) +$673; +2x 3.84TB U.2 NVMe +$6,914; +1x 25G 2x SFP28 CX6 NIC (AOC-631102AN-ADAT) +$391; +1x 100G CX6-DX 2x QSFP56 +$1,228; +1x 400G CX7 OSFP +$1,531; +2x 200G CX7 (2x QSFP112, crypto) +$3,848; On-site Silver NBD, 3 yr (approx 16% of system) +$30,700; Alternative chassis: Supermicro SYS-422GL-NR (6944P, 768GB, 2x 3.84TB E1.S, 25G NIC, DCMS) −$2,090.82
    power: 6,780 W nameplate; provision 8.47 kW (nameplate ÷ 0.8); circuit 208V 3-ph 30A (10.8 kW rated) ×2 A+B; colo $1,482/mo 5-yr · $1,906/mo 3-yr · $2,118/mo 1-yr
    per GPU: $23,986.01 (8 GPUs)
    page: https://ai.upstation.io/p/UPS-GPU-RTXPRO-8

## [8x NVIDIA RTX PRO 6000 Blackwell Server Edition, AMD](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8-T2): $163,909.60

- $163,909.60 — 8× RTX PRO 6000 · 2× 96C AMD · 768GB · 1× 3.84TB + 2× 1TB · 2× 25G · 6.7 kW nameplate — 4U-E10G-GENOAX2, Octoserver — 3-5 wk
    spec: 4U, dual SP5, PCIe 5.0, 8x dual-slot GPU slots; 8x RTX PRO 6000 SE 96GB passive (768GB total); 2x EPYC 9654 96C 2.4GHz (192 cores); 768GB DDR5-4800 (24x32GB); 1x 3.84TB U.2 NVMe + 2x 1TB M.2; 2x 25GbE SFP28 + 2x 1GbE on-board, IPMI; 4x 1600W Platinum (6400W total; 3+1 redundancy only below 4.8kW, so 4+0 at full GPU load), ~6.2kW DC configured; 3-yr parts warranty, rails and cables included
    options (price delta): Memory 768GB -> 1TB (16x64GB) +$13,640; Memory 768GB -> 1.5TB (24x64GB) +$35,640; +1x 3.84TB U.2 (2 total) +$2,050; +1x 100G 2x QSFP28 NIC +$578.20; Ubuntu Server 24.04 LTS preinstalled +$20; 5-year warranty upgrade +$9,834.58
    power: 6,692 W nameplate; provision 8.37 kW (nameplate ÷ 0.8); circuit 208V 3-ph 30A (10.8 kW rated) ×2 A+B; colo $1,465/mo 5-yr · $1,883/mo 3-yr · $2,092/mo 1-yr
    per GPU: $20,488.70 (8 GPUs)
    page: https://ai.upstation.io/p/UPS-GPU-RTXPRO-8-T2

## [8x NVIDIA RTX 5090 (passive server edition), AMD](https://ai.upstation.io/p/UPS-GPU-5090-8): $76,679.60

- $76,679.60 — 8× RTX 5090 · 2× 96C AMD · 512GB · 1× 3.84TB + 1× 1TB · 2× 25G · 6.4 kW nameplate — 4U-E10G-GENOAX2, Octoserver — 5-6 wk
    spec: 4U, dual SP5, PCIe 5.0, 8x dual-slot GPU slots; 8x Octoserver RTX 5090 Ai 32GB GDDR7, passive dual-slot (256GB total), 3-yr warranty; 2x EPYC 9654 96C 2.4GHz (192 cores); 512GB DDR5-4800 (16x32GB); 1x 3.84TB U.2 NVMe + 1x 1TB M.2; 2x 25GbE SFP28 + 2x 1GbE on-board, IPMI; 4x 1600W Platinum (6400W total; 4+0 at full GPU load), ~5.9kW DC configured; 3-yr parts warranty, rails and cables included
    options (price delta): CPU 2x 9654 -> 2x 9554 64C 3.1GHz (cheaper) −$360; CPU 2x 9654 -> 2x 9754 128C +$2,360; Memory 512GB -> 768GB (24x32GB) +$10,120; Memory 512GB -> 1TB (16x64GB) +$23,760; +1x 3.84TB U.2 (2 total) +$2,050; +1x 100G 2x QSFP28 NIC +$578.20; Ubuntu Server 24.04 LTS preinstalled +$20; 5-year warranty upgrade +$4,600.78
    power: 6,389 W nameplate; provision 7.99 kW (nameplate ÷ 0.8); circuit 208V 3-ph 30A (10.8 kW rated) ×2 A+B; colo $1,398/mo 5-yr · $1,798/mo 3-yr · $1,998/mo 1-yr
    per GPU: $9,584.95 (8 GPUs)
    page: https://ai.upstation.io/p/UPS-GPU-5090-8

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# GPU hosts without GPUs (build your own; 8 slots)

> UpStation catalog · Home › GPU hosts without GPUs (build your own; 8 slots) · https://ai.upstation.io/gpu-hosts

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

GPU hosts without GPUs: the table below is the same as on the [homepage](https://ai.upstation.io/); each product is then listed in full, and each has its own page (linked from its name). Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

| Host | Price | Slots | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8) | $47,884.46 | no GPUs, 8 slots | 2× 72C Intel | 512GB | 2× 960GB | 2× 25G | 1.1 → 1.38 | SYS-422GA-NRT | Supermicro | 3-5 wk |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-X13) | $37,937.10 | no GPUs, 8 slots | 2× 28C Intel | 512GB | 2× 960GB | 2× 25G | 1.7 → 2.19 | SYS-521GE-TNRT | Supermicro | 3-5 wk |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-AMD) | $39,179.76 | no GPUs, 8 slots | 2× 32C AMD | 512GB | 1× 960GB | 2× 25G | 1.0 → 1.25 | AS-5126GS-TNRT | Supermicro | 3-5 wk |
| [8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN5) | $32,919.60 | no GPUs, 8 slots | 2× 32C AMD | 512GB | 2× 960GB | 2× 25G | 1.1 → 1.31 | 4U-E10G-GENOAX2 | Octoserver | 2-3 wk |
| [8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN4) | $15,146.20 | no GPUs, 8 slots, PCIe 4.0 | 2× 32C AMD | 512GB DDR4 | 2× 960GB | 2× 25G | 1.1 → 1.34 | 4U-E10G-ROMEX2 | Octoserver | 2-3 wk |

## [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8): $47,884.46

- $47,884.46 — no GPUs, 8 slots · 2× 72C Intel · 512GB · 2× 960GB · 2× 25G · 1.1 kW nameplate — SYS-422GA-NRT, Supermicro — 3-5 wk
    spec: 4U X14 PCIe 5.0 dual-root switch, 8x double-width 600W slots, 8x E3.S/2.5in NVMe bays; 2x Xeon 6944P 72C 1.8GHz (144 cores); Xeon 6 supports Intel TDX, the trusted-VM basis for NVIDIA confidential computing on RTX PRO 6000; 512GB DDR5-6400 (16x32GB, 8 slots free); 2x 960GB M.2 7450 PRO NVMe boot; 2x 25GbE SFP28 (CX6 add-in) + 2x 10GbE RJ45 on-board; 4x 3200W Titanium (3+1); 3-yr parts and labor + 1-yr cross-ship; No GPUs
    equivalents: Intel 5th Gen (TDX) Supermicro SYS-521GE-TNRT (2x Xeon Gold 5520+ 28C), host −$9,947.36; AMD Supermicro AS-5126GS-TNRT (2x EPYC 9354 32C), host −$8,704.70
    options (price delta): CPU 2x 6944P -> 2x 6960P 72C 2.7GHz +$2,496; CPU 2x 6944P -> 2x 6980P 128C +$5,992; Memory 512GB -> 1TB (16x64GB) +$16,896; +1x 3.84TB E3.S NVMe +$1,975; +1x 100G CX6-DX 2x QSFP56 +$1,228; +1x 400G CX7 OSFP +$1,531; TPM 2.0 +$66; On-site Silver NBD, 3 yr (approx 16% of system) +$7,662
    power: 1,100 W nameplate; provision 1.38 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $241/mo 5-yr · $310/mo 3-yr · $345/mo 1-yr
    page: https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8

## [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-X13): $37,937.10

- $37,937.10 — no GPUs, 8 slots · 2× 28C Intel · 512GB · 2× 960GB · 2× 25G · 1.7 kW nameplate — SYS-521GE-TNRT, Supermicro — 3-5 wk
    spec: 5U X13 PCIe 5.0 dual-root switch, 10x double-width slots (8 GPUs), 24x 2.5in bays, 32 DIMM slots; 2x Xeon Gold 5520+ 28C 2.2GHz (56 cores), 5th Gen (Emerald Rapids); Intel TDX, the trusted-VM basis for NVIDIA confidential computing on RTX PRO 6000 (NVIDIA CC guide lists Emerald Rapids); 16 DIMMs fitted as TDX requires 8 per socket; 512GB DDR5-5600 (16x32GB, 16 slots free; 5th Gen tops out at 5600); 2x 960GB M.2 7450 PRO NVMe boot; 2x 25GbE SFP28 (AIOM CX-6 LX) + 2x 10GbE RJ45 on-board; 4x 2700W Titanium redundant PSU; 3-yr parts and labor + 1-yr cross-ship; DCMS licence included; No GPUs
    equivalents: Intel Xeon 6 (TDX) Supermicro SYS-422GA-NRT (2x Xeon 6944P 72C), host +$9,947.36; AMD Supermicro AS-5126GS-TNRT (2x EPYC 9354 32C), host +$1,242.66; Intel 5th Gen, 4U Supermicro SYS-421GE-TNRT (2x Xeon 5th Gen), host quote
    options (price delta): CPU 2x 5520+ -> 2x Platinum 8562Y+ 32C (TDX) +$6,258; CPU 2x 5520+ -> 2x Platinum 8558 48C (TDX) +$6,152; Memory: DDR5-6400 parts instead of 5600 (run at 5600) +$2,560; Memory 512GB -> 1TB (16x64GB) +$19,456; +1x 3.84TB U.2 NVMe (P5430 Gen4) +$2,784; +1x 100G CX6 2x QSFP56 PCIe +$1,465.10; +1x 400G CX7 +$1,530.65; TPM 2.0 +$49; Warranty 3 yr + 2 yr cross-ship (7.5% of system) +$2,845.28; On-site Silver NBD, 3 yr +$6,196.48
    power: 1,748 W nameplate; provision 2.19 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $383/mo 5-yr · $493/mo 3-yr · $548/mo 1-yr
    page: https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-X13

## [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-AMD): $39,179.76

- $39,179.76 — no GPUs, 8 slots · 2× 32C AMD · 512GB · 1× 960GB · 2× 25G · 1.0 kW nameplate — AS-5126GS-TNRT, Supermicro — 3-5 wk
    spec: 5U PCIe 5.0 dual-root switch, 8x double-width slots, 24x 2.5in NVMe bays; 2x EPYC 9354 32C 3.25GHz (64 cores); 512GB DDR5-6400 (16x32GB, 8 slots free); 1x 960GB M.2 7450 PRO NVMe boot (config #1 has one M.2 slot; second boot drive as a U.2 option); 2x 25GbE SFP28 (CX-6 LX) + 2x 10GbE on-board; 6x 2700W Titanium (4+2); 3-yr warranty; No GPUs
    equivalents: Intel Xeon 6 (TDX) Supermicro SYS-422GA-NRT (2x Xeon 6944P 72C), host +$8,704.70; Intel 5th Gen (TDX) Supermicro SYS-521GE-TNRT (2x Xeon Gold 5520+ 28C), host −$1,242.66
    options (price delta): +1x 960GB U.2 PM9A3 as second boot drive (config 1 has one on-board M.2) +$1,167; 2nd 960GB M.2 (API accepts qty 2; configurator menu caps M.2 at 1 - verify in cart) +$1,149; 2x 100G CX6-DX instead of 25G +$940.20; CPU 2x 9354 -> 2x 9355 32C 3.55GHz +$2,558; Memory 512GB -> 1TB (16x64GB) +$16,896; +1x 3.84TB U.2 PS1010 Gen5 +$3,314; TPM 2.0 +$66; DCMS licence +$180; On-site Silver NBD, 3 yr +$6,391.20; On-site Bronze 3-business-day, 3 yr +$4,187.34
    power: 1,000 W nameplate; provision 1.25 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $219/mo 5-yr · $281/mo 3-yr · $312/mo 1-yr
    page: https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-AMD

## [8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN5): $32,919.60

- $32,919.60 — no GPUs, 8 slots · 2× 32C AMD · 512GB · 2× 960GB · 2× 25G · 1.1 kW nameplate — 4U-E10G-GENOAX2, Octoserver — 2-3 wk
    spec: 4U, dual SP5, PCIe 5.0, 8x dual-slot x16 GPU slots (10 total), 2x U.2 + 2x M.2; 2x EPYC 9354 32C 3.3GHz (64 cores); 512GB DDR5-4800 (16x32GB, 8 usable slots free); 2x 960GB M.2 PM9A3 NVMe boot; 2x 25GbE SFP28 + 2x 1GbE on-board, IPMI; 6400W (3+1 1600W Platinum); 3-yr parts warranty, rails and cables included; No GPUs
    options (price delta): +1x 100G 2x QSFP28 NIC +$578.20; Ubuntu Server 24.04 LTS +$20
    power: 1,050 W nameplate; provision 1.31 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $229/mo 5-yr · $295/mo 3-yr · $328/mo 1-yr
    page: https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN5

## [8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN4): $15,146.20

- $15,146.20 — no GPUs, 8 slots, PCIe 4.0 · 2× 32C AMD · 512GB DDR4 · 2× 960GB · 2× 25G · 1.1 kW nameplate — 4U-E10G-ROMEX2, Octoserver — 2-3 wk
    spec: 4U, dual SP3 (EPYC 7003), PCIe 4.0, 8x dual-slot x16 GPU slots (slots 9-10 at x8), 2x U.2 + 1x M.2; 2x EPYC 7543 32C 2.8GHz (64 cores); 512GB DDR4-3200 (16x32GB, 16 slots free); 960GB M.2 PM9A3 + 960GB U.2 PM9A3 NVMe boot (chassis has one M.2 slot); 2x 25GbE SFP28 + 2x 10GbE on-board, IPMI; 6400W (3+1 1600W Platinum); 3-yr parts warranty, rails and cables included; No GPUs
    power: 1,074 W nameplate; provision 1.34 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $234/mo 5-yr · $302/mo 3-yr · $335/mo 1-yr
    page: https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN4

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# Switches, cables and optics (FS.com, PicOS included, generic optics)

> UpStation catalog · Home › Switches, cables and optics (FS.com, PicOS included, generic optics) · https://ai.upstation.io/switches

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

Switches, cables and optics: the table below is the same as on the [homepage](https://ai.upstation.io/); each product is then listed in full, and each has its own page (linked from its name). Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

| Ports | Price | $/port | W nameplate | Provision kW | Part | Lead time | Spec |
|---|---|---|---|---|---|---|---|
| [64× 800G OSFP](https://ai.upstation.io/p/UPS-NET-800G-64) | $44,600 | $697 | 1,800 | 2.25 | N8650-64OD, PicOS | ~2 wk | 64x 800G OSFP (or 128x 400G breakout); Broadcom Tomahawk 5, 51.2T; PicOS: RoCEv2, EVPN-VXLAN, MLAG; 5-yr warranty; options: Same spec on Dell Z9864F-ON / Enterprise SONiC quote; Same spec on Arista 7060X6-64PE / EOS quote; NVIDIA Spectrum-4 SN5600 / Cumulus quote |
| [32× 800G OSFP](https://ai.upstation.io/p/UPS-NET-800G-32) | $37,499 | $1,172 | 1,000 | 1.25 | N8650-32OD, PicOS | ~2 wk | 32x 800G OSFP (or 64x 400G breakout); Broadcom Tomahawk 5, 25.6T; PicOS; options: Dell / Arista / NVIDIA equivalent quote |
| [64× 400G QSFP-DD](https://ai.upstation.io/p/UPS-NET-400G-64) | $23,999 | $375 | 1,000 | 1.25 | N8610-64DB, PicOS | ~2 wk | 64x 400G QSFP-DD; Broadcom Tomahawk 4, 25.6T; PicOS; options: Dell Z9664F-ON / Enterprise SONiC quote; Arista 7060DX5-64S / EOS quote |
| [32× 400G QSFP-DD](https://ai.upstation.io/p/UPS-NET-400G-32) | $19,999 | $625 | 600 | 0.75 | N8610-32D, PicOS | ~2 wk | 32x 400G QSFP-DD; Broadcom Tomahawk 4, 12.8T; PicOS; options: FS N5680-32D (Tomahawk 3) budget variant −$700; Dell Z9432F-ON / Enterprise SONiC quote; Arista 7060DX5-32 / EOS quote |
| [24× 200G + 8× 400G](https://ai.upstation.io/p/UPS-NET-200G-32) | $12,400 | $517 | 600 | 0.75 | N8510-24CD8D, PicOS | Backorder to 12/2026 | 24x 200G QSFP56 + 8x 400G QSFP-DD; Broadcom Trident 4-X9; PicOS; options: No 32/64-port pure-200G box at FS; use 400G-32 with breakout quote |
| [64× 100G QSFP28](https://ai.upstation.io/p/UPS-NET-100G-64) | $11,900 | $186 | 550 | 0.69 | N5630-64C, PicOS | ~2 wk | 64x 100G QSFP28; Broadcom Tomahawk 2, 6.4T; PicOS; options: Dell Z9264F-ON / Enterprise SONiC quote; Arista 7060CX2-32S class quote |
| [32× 100G QSFP28](https://ai.upstation.io/p/UPS-NET-100G-32) | $7,919 | $247 | 250 | 0.31 | N8560-32C, PicOS | ~2 wk | 32x 100G QSFP28; Broadcom Trident 3-X7, 3.2T; PicOS; options: Dell S5232F-ON / Enterprise SONiC quote; Arista 7050CX3-32S / EOS quote |
| [48× 25G + 8× 100G](https://ai.upstation.io/p/UPS-NET-25G-48) | $6,049 | $126 | 150 | 0.19 | S5470-48Y, PicOS | ~2 wk | 48x 25G SFP28 + 8x 100G QSFP28; PicOS, MLAG; options: FS S5580-48Y (Trident 3, EVPN-VXLAN) +$1,700; Dell S5248F-ON / Enterprise SONiC quote; Arista 7050SX3-48YC8 / EOS quote |
| [48× 10G + 8× 100G](https://ai.upstation.io/p/UPS-NET-10G-48) | $4,569 | $95 | 120 | 0.15 | S5860-48SC | ~2 wk | 48x 10G SFP+ + 8x 100G QSFP28; FS OS rather than PicOS (PicOS 10G options: S5860-20SQ, S5440-12S) |
| [48× 1G RJ45 + 4× 10G](https://ai.upstation.io/p/UPS-NET-1G-48) | $1,089 | $23 | 60 | 0.07 | S3410-48TS, PicOS | ~2 wk | 48x 1GbE RJ45 + 4x 10G SFP+ uplinks; PicOS L2+ managed (VLAN, LACP, MLAG); out-of-band / BMC / IPMI layer; Fanless-class 1U, redundant PSU not standard; 5-yr warranty; options: Budget: S3540L-48T, 48x 1G + 4x 10G, PicOS L2+ (369 in stock) −$480; L3: S5810-48TS, 48x 1G + 4x 10G, PicOS L3 (US warehouse) +$490; 24-port: S3410-24TS, 24x 1G + 4x 10G, PicOS L2+ −$290; PoE+: S3410-48TS-P, 48x 1G PoE+ + 2x 10G + 2x 1G +$530; 48x 1G SFP fiber: S5810-48FS, PicOS L3 +$1,440; Non-PicOS (FS OS) same hardware: S3910-48TS L2+ $1,049 / S5810-48TS L3 $1,509 quote |

### Cables and optics (FS, generic-coded; reaches are rated maxima, OM4 for multimode)

| Speed | 2 m passive DAC | Passive reach | Active DAC / AEC | AOC 10 m (max length) | Short-reach optic, each (reach) | Long-reach optic, each (reach) | Breakout cable |
|---|---|---|---|---|---|---|---|
| 800G | $239 | 3 m | AEC 5 m $1,424 (max 5 m) | $3,011 (50 m) | SR8 $879 (50 m) | DR8 $1,319 (500 m) · 2×FR4 $1,869 (2 km) | 800G→2×400G 2 m $299 |
| 400G | $174 | 3 m | Active DAC 5 m $611 (max 5 m) | $1,055 (30 m) | SR8 $219 (100 m) | DR4 $549 (500 m) · FR4 $719 (2 km) · LR4 $999 (10 km) | 400G→4×100G 2 m $199 |
| 200G | $120 | 3 m | Active DAC 7 m $325 (max 7 m) | $1,034 (50 m) | SR4 $299 (100 m) | FR4 $1,249 (2 km) · LR4 $2,624 (10 km) | 200G→2×100G 2 m $119 |
| 100G | $54 | 5 m | Active DAC 9 m $300 (max 9 m) | $221 (30 m) | SR4 $99 (100 m) | CWDM4 $209 (2 km) · LR4 $399 (10 km) | 100G→4×25G 3 m $136 |
| 25G | $39 | 5 m | Active DAC 9 m $73 (max 9 m) | $73 (30 m) | SR $49 (100 m) | LR $74 (10 km) | — |
| 10G | $18 | 7 m | Active DAC 10 m $105 (max 10 m) | $39 (30 m) | SR $25 (300 m) | LR $34 (10 km) | — |
| 1G | RJ45, Cat6 patch | — | — | — | 10G SFP+ uplinks: see 10G row |  | — |

### Fibre patch (generic FS)

| Cable | 1 m | 3 m | 10 m | 30 m | 50 m | Used with |
|---|---|---|---|---|---|---|
| LC-LC duplex OM4 | $8.40 | $10 | $20 | $47 | — | SR (10G, 25G) |
| MTP-12 OM4 | $59 | $89 | $150 | $301 | $452 | SR4, SR8, DR4, DR8 |
| LC-LC duplex OS2 | $8.40 | — | — | — | — | FR4, CWDM4, LR, LR4 |

### Link by distance

| Distance | Use | Cost per link |
|---|---|---|
| Inside passive reach | 2 m passive DAC | 1 cable |
| To active max | Active DAC or AEC | 1 cable |
| To AOC max, or to short-reach optic reach | AOC, or 2× short-reach optic + multimode fibre | 1 cable, or 2 optics + fibre |
| Beyond | 2× long-reach optic + single-mode fibre | 2 optics + fibre |
| Any | Connected cost per port = switch $/port + link |  |

Choosing a link by distance: 2 m passive DAC inside passive reach; active DAC or AEC up to its maximum; beyond that an AOC (10 m priced; longer lengths cost more) or two short-reach optics plus multimode fibre out to the optic's reach; past that, two long-reach optics over single-mode. At 100G and up the optics, not the fibre, set the cost. Connected cost per port = switch $/port + one cable, or + two optics plus fibre. GPU-fabric links are estimated all-optical: GPU node density (1–3 HGX nodes per 40 kW rack) puts most nodes past passive-copper reach of their leaf, since a 3 m DAC spans at most the adjacent rack. Mixing copper in is realistic — expect 25% or more of fabric cables to land as passive DAC once installation is engineered (peer links, co-racked leaf↔spine, nodes in or beside the leaf's rack) — but which links qualify depends on rack elevations, so estimates stay optical and copper is taken as a savings at engineering time.

| PicOS | 100G | 400G | 800G |
|---|---|---|---|
| Licence | included | included | included |
| 1-yr support | $369 | $439 | $659 |
| Dell/SONiC or Arista, same ports | quote | quote | quote |

## [800G switch, 64x OSFP (Tomahawk 5)](https://ai.upstation.io/p/UPS-NET-800G-64): $44,600

- $44,600 — 64× 800G OSFP · 1.8 kW nameplate — FS N8650-64OD, PicOS, FS.com — ~2 wk
    spec: 64x 800G OSFP (or 128x 400G breakout); Broadcom Tomahawk 5, 51.2T; PicOS: RoCEv2, EVPN-VXLAN, MLAG; 5-yr warranty
    options (price delta): Same spec on Dell Z9864F-ON / Enterprise SONiC quote; Same spec on Arista 7060X6-64PE / EOS quote; NVIDIA Spectrum-4 SN5600 / Cumulus quote
    power: 1,800 W nameplate; provision 2.25 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $394/mo 5-yr · $506/mo 3-yr · $562/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-800G-64

## [800G switch, 32x OSFP (Tomahawk 5)](https://ai.upstation.io/p/UPS-NET-800G-32): $37,499

- $37,499 — 32× 800G OSFP · 1.0 kW nameplate — FS N8650-32OD, PicOS, FS.com — ~2 wk
    spec: 32x 800G OSFP (or 64x 400G breakout); Broadcom Tomahawk 5, 25.6T; PicOS
    options (price delta): Dell / Arista / NVIDIA equivalent quote
    power: 1,000 W nameplate; provision 1.25 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $219/mo 5-yr · $281/mo 3-yr · $312/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-800G-32

## [400G switch, 64x QSFP-DD (Tomahawk 4)](https://ai.upstation.io/p/UPS-NET-400G-64): $23,999

- $23,999 — 64× 400G QSFP-DD · 1.0 kW nameplate — FS N8610-64DB, PicOS, FS.com — ~2 wk
    spec: 64x 400G QSFP-DD; Broadcom Tomahawk 4, 25.6T; PicOS
    options (price delta): Dell Z9664F-ON / Enterprise SONiC quote; Arista 7060DX5-64S / EOS quote
    power: 1,000 W nameplate; provision 1.25 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $219/mo 5-yr · $281/mo 3-yr · $312/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-400G-64

## [400G switch, 32x QSFP-DD (Tomahawk 4)](https://ai.upstation.io/p/UPS-NET-400G-32): $19,999

- $19,999 — 32× 400G QSFP-DD · 0.6 kW nameplate — FS N8610-32D, PicOS, FS.com — ~2 wk
    spec: 32x 400G QSFP-DD; Broadcom Tomahawk 4, 12.8T; PicOS
    options (price delta): FS N5680-32D (Tomahawk 3) budget variant −$700; Dell Z9432F-ON / Enterprise SONiC quote; Arista 7060DX5-32 / EOS quote
    power: 600 W nameplate; provision 0.75 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $131/mo 5-yr · $169/mo 3-yr · $188/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-400G-32

## [200G switch, 24x 200G + 8x 400G (Trident 4)](https://ai.upstation.io/p/UPS-NET-200G-32): $12,400

- $12,400 — 24× 200G + 8× 400G · 0.6 kW nameplate — FS N8510-24CD8D, PicOS, FS.com — Backorder to 12/2026
    spec: 24x 200G QSFP56 + 8x 400G QSFP-DD; Broadcom Trident 4-X9; PicOS
    options (price delta): No 32/64-port pure-200G box at FS; use 400G-32 with breakout quote
    power: 600 W nameplate; provision 0.75 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $131/mo 5-yr · $169/mo 3-yr · $188/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-200G-32

## [100G switch, 64x QSFP28 (Tomahawk 2)](https://ai.upstation.io/p/UPS-NET-100G-64): $11,900

- $11,900 — 64× 100G QSFP28 · 0.6 kW nameplate — FS N5630-64C, PicOS, FS.com — ~2 wk
    spec: 64x 100G QSFP28; Broadcom Tomahawk 2, 6.4T; PicOS
    options (price delta): Dell Z9264F-ON / Enterprise SONiC quote; Arista 7060CX2-32S class quote
    power: 550 W nameplate; provision 0.69 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $121/mo 5-yr · $155/mo 3-yr · $172/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-100G-64

## [100G switch, 32x QSFP28 (Trident 3)](https://ai.upstation.io/p/UPS-NET-100G-32): $7,919

- $7,919 — 32× 100G QSFP28 · 0.2 kW nameplate — FS N8560-32C, PicOS, FS.com — ~2 wk
    spec: 32x 100G QSFP28; Broadcom Trident 3-X7, 3.2T; PicOS
    options (price delta): Dell S5232F-ON / Enterprise SONiC quote; Arista 7050CX3-32S / EOS quote
    power: 250 W nameplate; provision 0.31 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $54/mo 5-yr · $70/mo 3-yr · $78/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-100G-32

## [25G leaf, 48x SFP28 + 8x 100G](https://ai.upstation.io/p/UPS-NET-25G-48): $6,049

- $6,049 — 48× 25G + 8× 100G · 0.1 kW nameplate — FS S5470-48Y, PicOS, FS.com — ~2 wk
    spec: 48x 25G SFP28 + 8x 100G QSFP28; PicOS, MLAG
    options (price delta): FS S5580-48Y (Trident 3, EVPN-VXLAN) +$1,700; Dell S5248F-ON / Enterprise SONiC quote; Arista 7050SX3-48YC8 / EOS quote
    power: 150 W nameplate; provision 0.19 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $33/mo 5-yr · $43/mo 3-yr · $48/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-25G-48

## [10G leaf, 48x SFP+ + 100G uplinks](https://ai.upstation.io/p/UPS-NET-10G-48): $4,569

- $4,569 — 48× 10G + 8× 100G · 0.1 kW nameplate — FS S5860-48SC, FS.com — ~2 wk
    spec: 48x 10G SFP+ + 8x 100G QSFP28; FS OS rather than PicOS (PicOS 10G options: S5860-20SQ, S5440-12S)
    power: 120 W nameplate; provision 0.15 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $26/mo 5-yr · $34/mo 3-yr · $38/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-10G-48

## [1G management switch, 48x RJ45 + 4x 10G SFP+](https://ai.upstation.io/p/UPS-NET-1G-48): $1,089

- $1,089 — 48× 1G RJ45 + 4× 10G · 0.1 kW nameplate — FS S3410-48TS, PicOS, FS.com — ~2 wk
    spec: 48x 1GbE RJ45 + 4x 10G SFP+ uplinks; PicOS L2+ managed (VLAN, LACP, MLAG); out-of-band / BMC / IPMI layer; Fanless-class 1U, redundant PSU not standard; 5-yr warranty
    options (price delta): Budget: S3540L-48T, 48x 1G + 4x 10G, PicOS L2+ (369 in stock) −$480; L3: S5810-48TS, 48x 1G + 4x 10G, PicOS L3 (US warehouse) +$490; 24-port: S3410-24TS, 24x 1G + 4x 10G, PicOS L2+ −$290; PoE+: S3410-48TS-P, 48x 1G PoE+ + 2x 10G + 2x 1G +$530; 48x 1G SFP fiber: S5810-48FS, PicOS L3 +$1,440; Non-PicOS (FS OS) same hardware: S3910-48TS L2+ $1,049 / S5810-48TS L3 $1,509 quote
    power: 60 W nameplate; provision 0.07 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $12/mo 5-yr · $16/mo 3-yr · $18/mo 1-yr
    page: https://ai.upstation.io/p/UPS-NET-1G-48

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# Compute nodes

> UpStation catalog · Home › Compute nodes · https://ai.upstation.io/compute-nodes

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

Compute nodes: the table below is the same as on the [homepage](https://ai.upstation.io/); each product is then listed in full, and each has its own page (linked from its name). Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

| Product | Price | Form | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [1U compute node](https://ai.upstation.io/p/UPS-CN-INT-1U) | $33,298.33 | 1U | 2× 32C Intel | 512GB | diskless, 12 NVMe bays | 2× 25G | 1.0 → 1.2 | SYS-122C-TN | Supermicro | 3-5 wk |

## [1U compute node, dual socket, 64 cores, 512GB, diskless NVMe (virtualisation host)](https://ai.upstation.io/p/UPS-CN-INT-1U): $33,298.33

- $33,298.33 — 1U · 2× 32C Intel · 512GB · diskless, 12 NVMe bays · 2× 25G · 1.0 kW nameplate — SYS-122C-TN, Supermicro — 3-5 wk
    spec: 1U CloudDC X14, dual socket Xeon 6700P/6500P (8 channels each), 32 DIMM slots; 2x Xeon 6530P 32C 2.3GHz (64 cores, 225W); 6737P 2.9GHz is an option, 6730P is not offered on this chassis; 512GB DDR5-6400 (16x32GB, one DIMM per channel, 16 slots free); Diskless (config #3): 12x 2.5in hot-swap NVMe/SATA/SAS bays + M.2, all empty; 2x 25GbE SFP28 (AIOM CX-6 LX, OCP 3.0) + 1GbE; 3 PCIe 5.0 slots free; 2x 1000W Titanium redundant PSU; 3-yr parts and labor + 1-yr cross-ship; rail kit; Use: virtualisation / management tier beside a GPU cluster; AMD equivalent (AS-1126HS-TN, 2x EPYC 9354) on request; Intel/AMD is not a price decision, memory and drives set the price
    equivalents: Intel (base) Supermicro SYS-122C-TN (2x Xeon 6530P 32C), host same price; AMD Supermicro AS-1126HS-TN (2x EPYC 9354 32C), host quote
    options (price delta): CPU 2x 6530P -> 2x 6737P 32C 2.9GHz +$4,062; CPU 2x 6530P -> 2x 6520P 24C −$2,482; +2x 100G CX6 2x QSFP56 PCIe NIC +$2,930.20; 25G as PCIe Intel E810-XXVAM2 (AOC-S25GC-I2S-P) instead of the AIOM −$4; PSU 1000W -> 1600W pair (1200W pair unavailable today) +$315.90; Memory 512GB -> 1TB (32x32GB) +$23,584; +2x 960GB M.2 boot +$2,298; Config #1 (4 bays) instead of #3 (12 bays) −$400.42; TPM 2.0 +$54.60; DCMS licence +$180; On-site Silver NBD, 3 yr +$5,431.79; On-site Bronze 3-business-day, 3 yr +$3,558.76
    power: 960 W nameplate; provision 1.2 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $210/mo 5-yr · $270/mo 3-yr · $300/mo 1-yr
    page: https://ai.upstation.io/p/UPS-CN-INT-1U

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# Storage nodes

> UpStation catalog · Home › Storage nodes · https://ai.upstation.io/storage-nodes

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

Storage nodes: the table below is the same as on the [homepage](https://ai.upstation.io/); each product is then listed in full, and each has its own page (linked from its name). Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

| Product | Price | Form | CPUs | Memory | Storage | Network | kW nameplate → provisioned | Part | Vendor | Lead time |
|---|---|---|---|---|---|---|---|---|---|---|
| [2U all-NVMe storage node](https://ai.upstation.io/p/UPS-SN-INT-NVME) | $182,994.59 | 2U | 2× 32C Intel | 512GB | 24× 7.68TB NVMe, 184TB | 2× 100G | 1.3 → 1.65 | SYS-222C-TN | Supermicro | 3-5 wk |
| [2U HDD storage node](https://ai.upstation.io/p/UPS-SN-AMD-HDD) | $51,743.45 | 2U | 1× 16C AMD | 128GB | 24× 24TB HDD, 576TB | 2× 100G | 0.6 → 0.78 | ASG-2015S-E1CR24H | Supermicro | 3-5 wk |

## [2U all-NVMe storage node, dual socket, 24x 7.68TB NVMe (184TB raw)](https://ai.upstation.io/p/UPS-SN-INT-NVME): $182,994.59

- $182,994.59 — 2U · 2× 32C Intel · 512GB · 24× 7.68TB NVMe, 184TB · 2× 100G · 1.3 kW nameplate — SYS-222C-TN, Supermicro — 3-5 wk
    spec: 2U CloudDC X14, dual socket Xeon 6700P/6500P, 32 DIMM slots; 2x Xeon 6530P 32C 2.3GHz (64 cores, 225W); 512GB DDR5-6400 (16x32GB, one DIMM per channel); 24x 2.5in hot-swap NVMe bays (config #3): 24x 7.68TB Samsung PM9D3a PCIe 5.0 NVMe (184TB raw); 2x 960GB M.2 NVMe boot; 1x 100GbE CX6-DX 2x QSFP56 PCIe + 1GbE; AIOM slot free; 2x 2000W Titanium redundant PSU; 3-yr parts and labor + 1-yr cross-ship; rail kit; AMD equivalent (AS-2126HS-TN, 2x EPYC 9354, 24x NVMe) on request; drives set the price (~$788/TB at Supermicro's NVMe pricing)
    equivalents: Intel (base) Supermicro SYS-222C-TN (2x Xeon 6530P 32C), host same price; AMD Supermicro AS-2126HS-TN (2x EPYC 9354 32C), host quote
    options (price delta): Drives: 24x 7.68TB Solidigm D7-PS1010 Gen5 instead of PM9D3a −$7,296; NIC 2x 100G -> 2x 25G SFP28 AIOM CX-6 LX (frees the PCIe slot) −$948.20; +1x 100G CX6-DX (4 ports total) +$1,228.20; +AIOM 25G CX-6 LX alongside the 100G card +$280; CPU 2x 6530P -> 2x 6737P 32C 2.9GHz +$4,062; Memory 512GB -> 1TB (16x64GB) +$16,896; Drives: 12x 7.68TB (92TB raw) instead of 24 −$72,576; Drives: 24x 15.3TB Kioxia CD8-P (368TB raw) +$117,456; TPM 2.0 +$54.60; DCMS licence +$180; On-site Silver NBD, 3 yr +$18,528.20; On-site Bronze 3-business-day, 3 yr +$12,009.02
    power: 1,320 W nameplate; provision 1.65 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $289/mo 5-yr · $371/mo 3-yr · $412/mo 1-yr
    page: https://ai.upstation.io/p/UPS-SN-INT-NVME

## [2U HDD storage node, single socket, 24x 24TB SAS (576TB raw)](https://ai.upstation.io/p/UPS-SN-AMD-HDD): $51,743.45

- $51,743.45 — 2U · 1× 16C AMD · 128GB · 24× 24TB HDD, 576TB · 2× 100G · 0.6 kW nameplate — ASG-2015S-E1CR24H, Supermicro — 3-5 wk
    spec: 2U, 24x 3.5in SAS3 hot-swap, single socket; EPYC 9115 16C; 128GB DDR5 (4x32GB); 24x 24TB WD Ultrastar DC HC580 SAS3 7.2K (576TB raw); 2x 960GB M.2 boot; 2x 100GbE (AIOM BCM57508 2x QSFP28); Intel equivalent (SSG-521E-E1CR24H) and a 12-bay hybrid (AS-2015CS-TNR) on request; drives set the price at ~$63/TB
    equivalents: AMD (base) Supermicro ASG-2015S-E1CR24H (1x EPYC 9115 16C), host same price; Intel Supermicro SSG-521E-E1CR24H (1x Xeon Gold 6526Y 16C), host quote; AMD, smaller Supermicro AS-2015CS-TNR (1x EPYC 9115 16C), host quote; Intel, larger Supermicro SSG-641E-E1CR36H (1x Xeon), host quote
    options (price delta): NIC: 2x 100G CX6-DX PCIe card instead of the 100G AIOM +$366.20; 2x 25G (AIOM CX-6 LX) instead of 100G −$587; Drives: 23x 24TB (configurator menu lists max 23; the API accepted 24) −$1,507; Drives: 24x 22TB HC580 SAS (528TB raw) −$5,808; Drives: 24x 20TB Seagate Exos X24 SAS (480TB raw) −$12,696; Drives: 12x 24TB (288TB raw) −$18,084; CPU 9115 -> 9354 32C +$910; Memory 128GB -> 256GB (8x32GB) +$5,896; TPM 2.0 +$49; DCMS licence +$180; On-site Silver NBD, 3 yr +$5,239.02; On-site Bronze 3-business-day, 3 yr +$3,395.66
    power: 620 W nameplate; provision 0.78 kW (nameplate ÷ 0.8); circuit 208V 1-ph 20A (4.2 kW rated) ×2 A+B; colo $136/mo 5-yr · $176/mo 3-yr · $195/mo 1-yr
    page: https://ai.upstation.io/p/UPS-SN-AMD-HDD

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# Parts rate card (marginal; apply to a base build)

> UpStation catalog · Home › Parts rate card (marginal; apply to a base build) · https://ai.upstation.io/rate-card

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

Apply these marginal rates to a base build to resize memory, CPU or drives; alternative configurations are not priced separately. Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

| Item | Rate | Typical |
|---|---|---|
| DDR5-6400 RDIMM 32GB (Supermicro) | $46.06/GB | $1,474 per DIMM (MEM-DR532L-SL02-ER64), identical on every chassis probed; DDR5-5600 32GB $1,314 ($41.06/GB) on X13 |
| DDR5-6400 RDIMM 64GB (Supermicro) | $39.53/GB | $2,530 per DIMM; cheapest per GB |
| DDR5-6400 RDIMM 96GB (Supermicro) | $46.08/GB | $4,424 per DIMM |
| DDR5-6400 RDIMM 128GB (Supermicro) | $47.23/GB | $6,045 per DIMM; 256GB $10,311 ($40.28/GB, 09-03) |
| DDR5-4800 RDIMM 32GB (Octoserver) | $39.53/GB | $1,265 per DIMM |
| DDR4-3200 RDIMM 32GB (Octoserver Rome host) | $13.44/GB | $430 per DIMM |
| AMD EPYC 9005/9004, mainstream 16-96C (Supermicro) | $31–95/core | $66/core |
| Intel Xeon 6700P, mainstream 32-64C (Supermicro) | $101–166/core | $124/core |
| Intel Xeon 6900P, 72-128C (Supermicro) | $58–82/core | $65/core |
| AMD EPYC 9004/9005 at Octoserver | $29–51/core | $42/core |
| AMD EPYC 7003 at Octoserver (PCIe 4.0 host) | $13–29/core | $20/core |
| Nearline HDD 16-30TB SAS/SATA (Supermicro) | $49–63/TB | $58/TB |
| NVMe Gen4 U.2 TLC 1.9-15TB (Supermicro) | $690–1390/TB | $1006/TB |
| NVMe Gen5 U.2/E1.S/E3.S (Supermicro) | $715–1029/TB | $821/TB |
| SATA/SAS SSD (Supermicro) | $678–1050/TB | $800/TB |
| Exotic parts: Xeon 6768P/6788P/6748P, 3DS DIMMs, dual-port drives | quote |  |

Example: 1TB → 2TB with 64GB DIMMs = 1,024 GB × $39.53 = $40,478.72 (the H200 line lists this option at +$40,480). Marginal rates from non-exotic SKUs, taken from the eStore configurator (the price its cart charges). Within a family the price rises almost linearly with cores and capacity, so alternative configurations are not priced separately: apply these rates to the base build. Supermicro's drive pricing is high: buyers who source NVMe or HDD elsewhere should price drives from the market and the chassis from here. Exotic parts (Xeon 6768P/6788P/6748P at $250-320/core, 3DS 256GB DIMMs, dual-port or FIPS drives) sit outside these bands.

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# Power and colo

> UpStation catalog · Home › Power and colo · https://ai.upstation.io/power

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

Power model and colocation rates, with the computed provisioned kW, circuit and monthly colo for every listed product. Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

| Rule | Value |
|---|---|
| Provisioned kW | nameplate W ÷ 0.8 (no circuit above 80%) |
| Feeds | A + B, each sized for the whole box |
| Colo, 5-year term | $175/kW/month on provisioned kW, US or Canada |
| Colo, 3-year term | $225/kW/month on provisioned kW, US or Canada |
| Colo, 1-year term | $250/kW/month on provisioned kW, US or Canada |

| Circuit | Rated kW |
|---|---|
| 208V 1-ph 20A | 4.2 |
| 208V 1-ph 30A | 6.2 |
| 208V 3-ph 30A | 10.8 |
| 208V 3-ph 50A | 18.0 |
| 208V 3-ph 60A | 21.6 |
| 415V 3-ph 30A | 21.6 |
| 415V 3-ph 60A | 43.1 |
| 415V 3-ph 100A | 71.9 |

Nameplate max for PCIe GPU servers = (GPU TDP x count + CPU TDP x 2 + 8 W per DIMM + 15 W per drive + 400 W fans/board/NICs) / PSU efficiency (Titanium 0.94, Platinum 0.92); HGX servers use the vendor configurator total, which matches NVIDIA system guidance. Vendor configurator totals for PCIe boxes are kept in each product's basis as a cross-check only (Supermicro pads chassis overhead by 1.5-2.4 kW; Octoserver omits fans and PSU loss). Minimum circuit kW = nameplate / 0.8 so the circuit never runs above 80% (NEC continuous). Pick the smallest standard circuit whose rated kW (V x A x sqrt3 for 3-phase) >= that minimum; provision it twice (A+B) so either feed carries the box alone. Colo kW to buy = nameplate / 0.8; monthly colo = that kW x the term rate ($175 / $225 / $250 per kW-month for 5 / 3 / 1-year terms). Quoted to UpStation 2026-09-08, available in Canada and the US: $175/kW/mo on a 5-year term, $225 on 3-year, $250 on 1-year. Billed on the provisioned kW (nameplate / 0.8).

### Per box (computed)

| Item | Nameplate W | Provision kW | Circuit (A+B) | Colo/mo 5-yr | Colo/mo 3-yr | Colo/mo 1-yr |
|---|---|---|---|---|---|---|
| [8x NVIDIA B300 (HGX B300 NVL8), Intel](https://ai.upstation.io/p/UPS-GPU-B300-8) | 13,566 | 16.96 | 208V 3-ph 50A ×2 | $2,968 | $3,816 | $4,240 |
| [8x NVIDIA H200 (HGX H200), Intel](https://ai.upstation.io/p/UPS-GPU-H200-8) | 9,069 | 11.34 | 208V 3-ph 50A ×2 | $1,984 | $2,552 | $2,835 |
| [8x NVIDIA RTX PRO 6000 Blackwell Server Edition, Intel](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8) | 6,780 | 8.47 | 208V 3-ph 30A ×2 | $1,482 | $1,906 | $2,118 |
| [8x NVIDIA RTX PRO 6000 Blackwell Server Edition, AMD](https://ai.upstation.io/p/UPS-GPU-RTXPRO-8-T2) | 6,692 | 8.37 | 208V 3-ph 30A ×2 | $1,465 | $1,883 | $2,092 |
| [8x NVIDIA RTX 5090 (passive server edition), AMD](https://ai.upstation.io/p/UPS-GPU-5090-8) | 6,389 | 7.99 | 208V 3-ph 30A ×2 | $1,398 | $1,798 | $1,998 |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8) | 1,100 | 1.38 | 208V 1-ph 20A ×2 | $241 | $310 | $345 |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX)](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-X13) | 1,748 | 2.19 | 208V 1-ph 20A ×2 | $383 | $493 | $548 |
| [8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-AMD) | 1,000 | 1.25 | 208V 1-ph 20A ×2 | $219 | $281 | $312 |
| [8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN5) | 1,050 | 1.31 | 208V 1-ph 20A ×2 | $229 | $295 | $328 |
| [8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD](https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN4) | 1,074 | 1.34 | 208V 1-ph 20A ×2 | $234 | $302 | $335 |
| [64× 800G OSFP switch](https://ai.upstation.io/p/UPS-NET-800G-64) | 1,800 | 2.25 | 208V 1-ph 20A ×2 | $394 | $506 | $562 |
| [32× 800G OSFP switch](https://ai.upstation.io/p/UPS-NET-800G-32) | 1,000 | 1.25 | 208V 1-ph 20A ×2 | $219 | $281 | $312 |
| [64× 400G QSFP-DD switch](https://ai.upstation.io/p/UPS-NET-400G-64) | 1,000 | 1.25 | 208V 1-ph 20A ×2 | $219 | $281 | $312 |
| [32× 400G QSFP-DD switch](https://ai.upstation.io/p/UPS-NET-400G-32) | 600 | 0.75 | 208V 1-ph 20A ×2 | $131 | $169 | $188 |
| [24× 200G + 8× 400G switch](https://ai.upstation.io/p/UPS-NET-200G-32) | 600 | 0.75 | 208V 1-ph 20A ×2 | $131 | $169 | $188 |
| [64× 100G QSFP28 switch](https://ai.upstation.io/p/UPS-NET-100G-64) | 550 | 0.69 | 208V 1-ph 20A ×2 | $121 | $155 | $172 |
| [32× 100G QSFP28 switch](https://ai.upstation.io/p/UPS-NET-100G-32) | 250 | 0.31 | 208V 1-ph 20A ×2 | $54 | $70 | $78 |
| [48× 25G + 8× 100G switch](https://ai.upstation.io/p/UPS-NET-25G-48) | 150 | 0.19 | 208V 1-ph 20A ×2 | $33 | $43 | $48 |
| [48× 10G + 8× 100G switch](https://ai.upstation.io/p/UPS-NET-10G-48) | 120 | 0.15 | 208V 1-ph 20A ×2 | $26 | $34 | $38 |
| [48× 1G RJ45 + 4× 10G switch](https://ai.upstation.io/p/UPS-NET-1G-48) | 60 | 0.07 | 208V 1-ph 20A ×2 | $12 | $16 | $18 |
| [1U compute node, dual socket, 64 cores, 512GB, diskless NVMe (virtualisation host)](https://ai.upstation.io/p/UPS-CN-INT-1U) | 960 | 1.2 | 208V 1-ph 20A ×2 | $210 | $270 | $300 |
| [2U all-NVMe storage node, dual socket, 24x 7.68TB NVMe (184TB raw)](https://ai.upstation.io/p/UPS-SN-INT-NVME) | 1,320 | 1.65 | 208V 1-ph 20A ×2 | $289 | $371 | $412 |
| [2U HDD storage node, single socket, 24x 24TB SAS (576TB raw)](https://ai.upstation.io/p/UPS-SN-AMD-HDD) | 620 | 0.78 | 208V 1-ph 20A ×2 | $136 | $176 | $195 |

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# How solutions are sized

> UpStation catalog · Home › How solutions are sized · https://ai.upstation.io/sizing

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

UpStation sizes a deployment in this order: GPUs → nodes → networks → storage → racks and power → bill of materials. The rules below are planning conventions (the same data is in catalog.json under sizing); figures marked estimate are not vendor specifications. Two worked designs computed with these rules are listed under Reference builds.

### Before sizing, settle

- Workload: training, fine-tuning, inference, HPC (default training when 'AI at scale' is implied)
- GPU count or node count (8 GPUs per node), or model size and precision to derive it
- Parallelism: single GPU, multi-GPU in one node, multi-node (decides HGX vs PCIe and whether a GPU fabric exists)
- Growth: day-1 full (default) or +N GPUs over a horizon
- Storage: usable TB now and at horizon, dataset character (text, compressed image/video, uncompressed vision), checkpoint size and pause, protection level
- Site: per-rack kW limit, rack space, whether racks/PDUs/whips exist (customer, colo, or to be supplied)
- Colo term (1, 3 or 5 years) and whether InfiniBand is a hard requirement (not listed; 800G RoCE Ethernet at 1:1 usually satisfies the intent)

### GPUs to nodes

- Size a GPU cluster on accelerators (GPU count and aggregate GPU memory), never on host cores. Every listed GPU server has 8 GPUs: nodes = ceil(GPUs / 8). Nodes ship fully populated; partial population is not offered.
- Model memory: required GPU memory (GB) = parameters (billions) × bytes per parameter × 1.2 (weights plus overhead) + KV cache (grows with context length × batch). Bytes per parameter: FP16/BF16 2, FP8 1, INT4/AWQ 0.5. GPU count = ceil(required GB / GB per GPU), rounded up to a tensor-parallel width of 1, 2, 4 or 8 inside one node. Quantization is the biggest lever on GPU count: 70B at FP16 is about 168 GB (2× H200, or 1× B300); 70B at FP8 is about 84 GB (1× H200 or 1× RTX PRO 6000 for weights, plus KV cache).
- GPU memory per card: NVIDIA B300 (HGX NVL8) 288 GB, NVIDIA H200 (HGX) 141 GB, NVIDIA RTX PRO 6000 Server Edition 96 GB, NVIDIA RTX 5090 (passive server edition) 32 GB.
- HGX or PCIe: Does the model at its precision fit one GPU's memory, and can each GPU work alone? Yes: PCIe. No: HGX. Distributed training is never served by PCIe servers. B300 and H200 (NVLink, HGX): the GPUs must act as one (model does not fit one GPU, tensor or pipeline parallel, all-reduce across GPUs or nodes; distributed training at any node count). RTX PRO 6000 and RTX 5090 (PCIe): GPUs run independently (inference replicas that each fit one card, batch jobs, VDI, rendering); no GPU-to-GPU fabric and no back-end network by default.
- What an HGX cluster carries: An HGX cluster of 2 or more nodes carries a dedicated non-blocking GPU fabric (1 port per GPU), a storage tier sized to feed the GPUs, and up to 2× the power per node of a PCIe server; a PCIe inference deployment carries neither the fabric nor the parallel-file tier.
- Host memory: Platform standard (the listed base build, 1–2 TB on an 8-GPU node) is the default. Host RAM ≥ aggregate GPU memory (up to 1.5–2×) is a costed option taken only for CPU/NVMe offload, unified-memory oversubscription or host-staged checkpointing; price it with the rate card and snap to whole DIMM fills (16 or 32 DIMMs). CPUs: Report the listed CPUs as an assumption; they are not a sizing input. Local NVMe: Several TB of local NVMe per node for scratch and checkpoints (listed options); datasets live on shared storage.

### Networks

| Plane | Carries | Ports per node | Oversubscription |
|---|---|---|---|
| GPU fabric | Collectives only (all-reduce, GPUDirect RDMA); never storage | 1 port per GPU at the GPU generation's speed: H200 400G (listed option, 8× 400G), B300 800G (on board, 8× 800G). None on PCIe GPU servers unless the buyer asks | 1:1 non-blocking |
| Front-end (converged) | Client access, orchestration, in-band management, and storage unless a storage plane is dedicated | 2 ports per node, one to each switch of a pair (25G floor; 100G or 200G when the base build or option provides it) | 3:1 (storage-facing 4:3 or better) |
| Storage (dedicated) | Parallel file system or NAS data path at scale | 2 ports per storage node and per GPU node on the storage plane, only when that plane is dedicated (see 'Dedicated storage plane when'); otherwise storage rides the front-end plane | 1:1 |
| OOB | BMC and switch management | 1× 1G BMC port per node plus 1× 1G management port per switch, on a 48× 1G switch (1 per rack or per 48 ports) | n/a |

- One switch: One switch is enough when the plane's port demand fits one switch (GPU fabric: ≤ 32 or ≤ 64 GPU ports on the 32- or 64-port 800G/400G switches). Note it is a single point of failure; a leaf-spine pair or fat-tree removes it.
- Fat-tree: R-port switch at the host speed: R/2 ports down to hosts, R/2 up to spines. leaves = ceil(host ports ÷ (R/2)); uplinks = leaves × R/2; spines = max(2, ceil(uplinks ÷ R)); each spine port terminates one uplink. Rail-optimized: Default for GPU fabrics: 8 rails (one per GPU position); leaves per rail = ceil(nodes ÷ (R/2)); leaves = 8 × leaves per rail; hosts connect to leaves in a network rack (20 m links, optical — see the cable-medium rule under Links), leaves to spines at 20–50 m. Examples with 64-port switches: 8 nodes (64 GPU ports) → 1 switch or 2 leaves + 2 spines; 32 nodes (256 ports) → 8 leaves + 4 spines, 256 host links + 256 uplinks; 128 nodes (1,024 ports) → 32 leaves + 16 spines. Three tiers: When leaves > R a 2-tier fat-tree cannot reach every leaf: pods of R/2 leaves with their own spines and a super-spine tier (above about 1,000–2,000 GPUs on 64-port switches).
- Front-end leaf-spine: Access leaves (48× 25G + 8× 100G): 48 host ports; uplinks = ceil(48 × 25 ÷ 3 ÷ 100) = 4× 100G per leaf at 3:1; spines = max(2, ceil(uplinks ÷ spine ports)). A single MLAG pair (2 leaves) or two pairs (4 leaves, full-mesh 4-link square between pairs) needs no spine; 3 or more pairs go leaf-spine. Peer link: Each switch pair reserves 2 ports at its uplink speed for the peer link (in-rack DAC) plus a 1G heartbeat over OOB.
- Dedicated storage plane when: Sustained storage traffic approaches 40 GB/s per node or several hundred GB/s in aggregate (parallel file system at scale); otherwise storage nodes and GPU-node storage traffic share the front-end plane, with PFC/ECN enabled and storage-facing oversubscription 4:3 or better. A port-speed mismatch (100G storage node on a 200G switch) is a breakout, not a reason to split. The GPU fabric never carries storage: checkpoint flows head-of-line-block collectives and need different congestion tuning; storage must scale independently of GPU count.
- Links: 1 cable per link: host to leaf, leaf to spine, peer links. Transceivers: DAC/AOC 0; fibre point-to-point 2 per link (both ends); breakout 1 parent + N children. Default lengths: host to leaf 10–20 m (leaves sit in a network rack; 2 m only when a node genuinely shares the leaf's rack), leaf to spine 10–30 m; pick the medium from the cables table by reach. Breakout: 800G→2×400G, 400G→4×100G, 200G→2×100G, 100G→4×25G: one breakout cable connects N slower endpoints to one switch port; cages used = ceil(endpoints ÷ N).
- Cable medium for estimates: GPU node density forces inter-rack cabling: 1–3 HGX nodes fill a 40 kW rack, and a 3 m passive DAC reaches at most the adjacent rack (switch mid-rack of a 3-rack span). Even the best case — one 64-port switch fully serving 8 nodes (8 ports each) — would need about 3 nodes per rack inside that span, beyond normal datacenter power densities; a fat-tree spreads nodes further still. So estimate every GPU-fabric link (host to leaf and leaf to spine) as optical — 2× short-reach optics + multimode fibre, or AOC — never as passive DAC. Mixing copper in is entirely realistic: which links can drop to passive DAC (or 5 m AEC) depends on how racks, switches and elevations are actually installed — switch peer links, co-racked leaf↔spine, and nodes in or adjacent to the leaf's rack all qualify, and 25% of fabric cables landing as copper DAC is a reasonable minimum expectation. Take that as a savings at engineering time, not in the estimate.

### Storage

- Capture: Usable capacity needed (current + growth, TB), throughput (GB/s), protection level, media, protocol (file for AI), checkpoint size and pause, dataset size; raw capacity is a design output.
- Read demand (aggregate GB/s per GPU): LLM/text 0.15; compressed image/video 0.25–0.5; uncompressed vision/multimodal 0.5–0.8; cap 0.8. Aggregate cluster-sustained rates (published pod reference tiers). Never multiply a single-node peak (4–5 GB/s per GPU) by the cluster's GPU count; a 2,048-GPU pod needs about 1 TB/s, not 8–10 TB/s.
- Write demand: max(0.5 × read GB/s, checkpoint burst). Checkpoint burst GB/s = parameters (billions) × 16 GB ÷ acceptable pause (s); full mixed-precision optimizer state is about 16 bytes per parameter. Asynchronous checkpointing drops the sustained write need to 50–200 GB/s regardless of model size.
- Capacity: dataset + 3–5 retained checkpoints × (parameters (billions) × 16 GB ÷ 1,000).
- Plannable capacity, per-node RAID: plannable per node = raw × RAID efficiency × 0.85 (fill ceiling). A node is one failure domain: losing it takes its data offline unless the software also replicates across nodes. Distributed (Ceph): plannable = raw × (N − 1) ÷ N (self-heal reserve) × efficiency × 0.85; hosts ≥ k + m for erasure coding (≥ 3 for replica 3), plus 1 recommended; parity m = 2 by default; k ≤ 8. Scale-out NAS: usable = raw × efficiency (N + 2 across nodes: 0.6 at 5 nodes, 0.8 at 10, 0.89 at 16+) − 1 virtual hot-spare drive per node, × 0.97 file-system overhead, then × 0.85 fill. Fill ceiling 0.85 in every case.
- Node count: nodes = max(ceil(read GB/s ÷ per-node GB/s), ceil(capacity ÷ per-node plannable), protection floor). No extra headroom nodes; ceil() already rounds up. If plannable capacity exceeds 2× the need, the feed rate is a single-node peak or the tier is wrong: re-derive.

| Protection | Efficiency (usable ÷ raw before fill ceiling) |
|---|---|
| Replica 3 | 0.333 |
| EC 2+2 | 0.5 |
| EC 4+2 | 0.667 |
| EC 8+2 | 0.8 |
| EC 8+3 | 0.727 |
| EC 8+4 | 0.667 |
| RAID 6 10+2 | 0.833 |
| RAID 6 8+2 | 0.8 |
| RAID 6 6+2 | 0.75 |
| RAID 10 | 0.5 |

| Storage node | Raw TB | Ports | Line rate GB/s | Planning GB/s per node | RAID 6 10+2 plannable TB per node | Note |
|---|---|---|---|---|---|---|
| 2U all-NVMe storage node, dual socket, 24x 7.68TB NVMe (184TB raw) | 184 | 2× 100G (4 with the listed option) | 25 | 20 | 130.3 | line rate 2 × 100 Gb/s ÷ 8 = 25 GB/s; plan 20 GB/s per node (estimate, software dependent); 50 GB/s line rate with 4× 100G |
| 2U HDD storage node, single socket, 24x 24TB SAS (576TB raw) | 576 | 2× 100G | 25 | 5 | 407.8 | 24 nearline drives at about 250 MB/s each ≈ 6 GB/s streaming (estimate); plan 5 GB/s; add NVMe DB/WAL (1–4% of HDD capacity) via the rate card for Ceph |

- Ceph node rules: All-NVMe: 1 OSD per drive, 4–6 cores and 8 GB RAM per OSD, 2× 100G; HDD: 1 core and 5 GB RAM per OSD plus 16 GB base, 1 NVMe DB/WAL per 6 HDDs, 2× 25G or 100G; 3 monitors (5 at scale), 2 managers, ≥ 2 metadata servers; public and cluster networks, cluster ≈ public × replica size.
- Tiers: Aggregate read below about 1 TB/s: scale-out NAS or Ceph-class file on the listed storage nodes. 1–4 TB/s: still NAS-class by default; a parallel file system (WEKA, BeeGFS, Lustre, IBM Storage Scale) converged on the GPU nodes' NVMe or on ≥ 8 NVMe nodes when the buyer wants one. Above about 4 TB/s: a standalone parallel file system on all-NVMe nodes. Storage software is not in this catalog: the buyer supplies and supports it (Ceph, BeeGFS, WEKA, ZFS/NFS, ...); UpStation quotes the nodes.
- GPU nodes and storage: GPU nodes reach storage over the front-end plane (converged) or a dedicated storage plane; the GPU fabric never carries storage.

### Racks, power and management

- nodes per rack = min(floor(40 U ÷ RU per node), floor(rack kW ÷ provisioned kW per node)). Rack kW = the facility's stated limit, else 40 kW (a common air-cooled compute-rack envelope). Report whether power or space binds. Every listed server is air-cooled.
- Feeds: A + B feeds, each alone carrying the whole rack; per-rack circuits = smallest standard circuit whose rated kW ≥ rack provisioned kW, ×2, or several pairs when a rack exceeds one circuit.
- Placement: Leaves for GPU fabrics sit in a network rack (or mid-row); front-end pair and OOB switch in the rack they serve; spines in a central rack; storage nodes in their own racks.
- Management: Clusters of 4 or more GPU nodes add 3 management nodes (the listed 1U compute node: provisioning/PXE, cluster manager, scheduler control plane, monitoring; 2 for an active/passive pair at small scale, 1 for a lab) and one 48× 1G OOB switch per rack; the management nodes attach to the front-end plane.

| GPU server | RU | Provisioned kW | Per 40 kW rack | Rack kW | Binds on |
|---|---|---|---|---|---|
| 8x NVIDIA B300 (HGX B300 NVL8), Intel | 8 | 16.96 | 2 | 33.92 | power |
| 8x NVIDIA H200 (HGX H200), Intel | 8 | 11.34 | 3 | 34.02 | power |
| 8x NVIDIA RTX PRO 6000 Blackwell Server Edition, Intel | 4 | 8.47 | 4 | 33.88 | power |
| 8x NVIDIA RTX PRO 6000 Blackwell Server Edition, AMD | 4 | 8.37 | 4 | 33.48 | power |
| 8x NVIDIA RTX 5090 (passive server edition), AMD | 4 | 7.99 | 5 | 39.95 | power |
| 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX) | 4 | 1.38 | 10 | 13.8 | space |
| 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX) | 5 | 2.19 | 8 | 17.52 | space |
| 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD | 5 | 1.25 | 8 | 10.0 | space |
| 8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD | 4 | 1.31 | 10 | 13.1 | space |
| 8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD | 4 | 1.34 | 10 | 13.4 | space |

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt


# Reference builds and worked designs

> UpStation catalog · Home › Reference builds and worked designs · https://ai.upstation.io/designs

Every price on this site is on the [homepage](https://ai.upstation.io/). Each product type and each product has its own page with the full details (spec, options, Intel/AMD equivalents, power and colo): [GPU servers](https://ai.upstation.io/gpu-servers) · [GPU hosts without GPUs](https://ai.upstation.io/gpu-hosts) · [Switches, cables and optics](https://ai.upstation.io/switches) · [Compute nodes](https://ai.upstation.io/compute-nodes) · [Storage nodes](https://ai.upstation.io/storage-nodes) · [Parts rate card](https://ai.upstation.io/rate-card) · [Power and colo](https://ai.upstation.io/power) · [How solutions are sized](https://ai.upstation.io/sizing) · [Reference builds and worked designs](https://ai.upstation.io/designs). AI assistants: the map is [llms.txt](https://ai.upstation.io/llms.txt); the data is [catalog.json](https://ai.upstation.io/catalog.json); every page is also Markdown (append .md, or send Accept: text/markdown).

Every build below is computed from the catalog data and the [sizing rules](https://ai.upstation.io/sizing); use them to check arithmetic. Prices in USD as of 2026-09-09; final price confirmed on order. Continental US; Canada quoted case by case. Orders and quotes: [book a meeting](https://calendly.com/alecglover/intro-call). Every page: [llms.txt](https://ai.upstation.io/llms.txt) (map) · [homepage](https://ai.upstation.io/) (all tables) · [catalog.json](https://ai.upstation.io/catalog.json) (data).

### One 8× H200 node, 3-year colo

| Item | Qty | Unit price | Extended | Provisioned kW |
|---|---|---|---|---|
| 8x NVIDIA H200 (HGX H200), Intel — SYS-821GE-TNHR | 1 | $319,421.64 | $319,421.64 | 11.34 |
| Hardware total |  |  | $319,421.64 | 11.34 |
| Colo, 3-year term, $225/kW/month × 11.34 kW |  |  | $2,552/month |  |
| Colo over 36 months |  |  | $91,872 |  |
| Hardware + colo over 36 months |  |  | $411,293.64 |  |

### Four 8× B300 nodes, 5-year colo

| Item | Qty | Unit price | Extended | Provisioned kW |
|---|---|---|---|---|
| 8x NVIDIA B300 (HGX B300 NVL8), Intel — SYS-822GS-NB3RT-01-G2 | 4 | $572,364 | $2,289,456 | 67.84 |
| Hardware total |  |  | $2,289,456 | 67.84 |
| Colo, 5-year term, $175/kW/month × 67.84 kW |  |  | $11,872/month |  |
| Colo over 60 months |  |  | $712,320 |  |
| Hardware + colo over 60 months |  |  | $3,001,776 |  |

### Two 8× RTX PRO 6000 nodes, one 48× 25G switch, 1-year colo

| Item | Qty | Unit price | Extended | Provisioned kW |
|---|---|---|---|---|
| 8x NVIDIA RTX PRO 6000 Blackwell Server Edition, Intel — SYS-422GA-NRT | 2 | $191,888.10 | $383,776.20 | 16.94 |
| +1× 25G 2× SFP28 CX6 NIC (listed option, one per node) | 2 | $391 | $782 | — |
| 48× 25G + 8× 100G switch — S5470-48Y, PicOS | 1 | $6,049 | $6,049 | 0.19 |
| 25G 2 m passive DAC (2 per node) | 4 | $39 | $156 | — |
| Hardware total |  |  | $390,763.20 | 17.13 |
| Colo, 1-year term, $250/kW/month × 17.13 kW |  |  | $4,282/month |  |
| Colo over 12 months |  |  | $51,384 |  |
| Hardware + colo over 12 months |  |  | $442,147.20 |  |

### Worked design: 4× 8× B300 training cluster (32 GPUs), 800G fabric, 200G front-end, 5× NVMe storage nodes, 5-year colo

Inputs: 32 GPUs for multi-node training; compressed image/video data (0.5 GB/s per GPU aggregate); 70B-parameter model, synchronous checkpoints with a 60 s pause; 200 TB dataset, 4 checkpoints retained; buyer-supplied Ceph, replica 3; 40 kW racks; 5-year colo

| Item | Qty | Unit price | Extended | Provisioned kW |
|---|---|---|---|---|
| 8x NVIDIA B300 (HGX B300 NVL8), Intel — SYS-822GS-NB3RT-01-G2 | 4 | $572,364 | $2,289,456 | 67.84 |
| +2× 200G CX7 north-south (listed option, one per node: front-end ports) | 4 | $3,848 | $15,392 | — |
| 32× 800G OSFP switch — N8650-32OD, PicOS (GPU fabric, 1:1) | 1 | $37,499 | $37,499 | 1.25 |
| 800G link, 2× SR8 optic + MTP-12 OM4 10 m (1 per GPU port; every fabric link estimated optical) | 32 | $1,908 | $61,056 | — |
| 24× 200G + 8× 400G switch — N8510-24CD8D, PicOS (front-end pair, MLAG) | 2 | $12,400 | $24,800 | 1.5 |
| 200G AOC 10 m (GPU nodes to the front-end pair, 2 per node) | 8 | $1,034 | $8,272 | — |
| 200G→2×100G breakout 2 m (storage nodes to the front-end pair, 2 nodes per cable) | 6 | $119 | $714 | — |
| 400G 2 m passive DAC (peer link between the pair) | 2 | $174 | $348 | — |
| 2U all-NVMe storage node, dual socket, 24x 7.68TB NVMe (184TB raw) — SYS-222C-TN | 5 | $182,994.59 | $914,972.95 | 8.25 |
| 48× 1G RJ45 + 4× 10G switch — S3410-48TS, PicOS (OOB; Cat6 patches not priced) | 1 | $1,089 | $1,089 | 0.07 |
| Hardware total |  |  | $3,353,598.95 | 78.91 |
| Colo, 5-year term, $175/kW/month × 78.91 kW |  |  | $13,809/month |  |
| Colo over 60 months |  |  | $828,540 |  |
| Hardware + colo over 60 months |  |  | $4,182,138.95 |  |

- GPUs → nodes: 32 ÷ 8 = 4 nodes. Multi-node training means the GPUs act as one, so HGX class (B300, 288 GB per GPU, 2.3 TB per node); a PCIe server would not serve it. Host memory stays at the listed 2TB (no offload stated).
- GPU fabric: 1 port per GPU = 32 × 800G. 32 ≤ 32, so one 32× 800G OSFP switch is non-blocking at 1:1 (single point of failure, accepted; growth past 4 nodes needs the 64-port switch or a fat-tree: at 32 nodes, 8 leaves + 4 spines on 64-port switches). Links: all 32 GPU ports priced optical (2× SR8 + MTP-12 fibre, $1,908 per link) — at 2 nodes per 40 kW rack the cluster spans 2 racks, so most links leave the switch's rack and 800G passive DAC (3 m reach) cannot be assumed; swapping the 16 genuinely in-rack links to DAC is a final-engineering saving, not part of the estimate. No storage on this fabric.
- Front-end (converged with storage): each node takes the listed +2× 200G option and connects one port to each switch of an MLAG pair of 24× 200G + 8× 400G switches, 3:1 or better; the storage nodes' 2× 100G ports land on the same pair through 200G→2×100G breakouts (5 nodes × 2 ports = 10 ports, 3 breakout cables per switch); peer link 2× 400G DAC. Uplinks to the buyer's core are not priced. Storage stays converged because 16 GB/s across 4 nodes is far below the 40 GB/s-per-node dedicated-fabric threshold.
- Storage demand: read = 32 GPUs × 0.5 GB/s = 16 GB/s (aggregate, not the 4–5 GB/s single-node peak); checkpoint burst = 70 × 16 GB ÷ 60 s = 18.7 GB/s; write = max(½ × 16, 18.7) = 18.7 GB/s; capacity = 200 TB + 4 × 1.12 TB = 204.5 TB.
- Storage supply: 2U all-NVMe storage node, dual socket, 24x 7.68TB NVMe (184TB raw) = 184 TB raw, 20 GB/s planning per node. Replica 3 on N nodes plans to 184 × N × (N−1)/N × 0.333 × 0.85: N = 4 gives 156.2 TB (short), N = 5 gives 208.3 TB ≥ 204.5 TB. Throughput 5 × 20 = 100 GB/s ≥ 16. Erasure coding 4+2 would need 6 hosts and plan to 521.6 TB, 2.6× the need, so replica 3 on 5 nodes is the answer. Software (Ceph) is buyer-supplied and not priced.
- OOB: 4 GPU nodes + 5 storage nodes + 3 switches = 12 × 1G ports on one 48× 1G RJ45 + 4× 10G switch.
- Racks at 40 kW: B300 = 8 U and 16.96 kW provisioned, so 2 per rack (power binds). Rack A: 2× B300 + GPU switch + OOB switch = 35.24 kW; rack B: 2× B300 = 33.92 kW; rack C: 5× storage + front-end pair = 9.75 kW, 12 U. Racks, PDUs and cabling labour are quoted separately.
- Power and colo: 78.91 kW provisioned in total; 5-year colo $175/kW/month × 78.91 kW = $13,809/month. Management nodes (3× the listed 1U compute node) are recommended at this size and not included here.

### Worked design: 2 PB usable archive on 5× HDD storage nodes (RAID 6), 100G front-end pair, 5-year colo

Inputs: 2,000 TB usable, archive / cold data, RAID 6 protection, no stated throughput; buyer-supplied file software (ZFS/NFS or similar); 40 kW racks; 5-year colo

| Item | Qty | Unit price | Extended | Provisioned kW |
|---|---|---|---|---|
| 2U HDD storage node, single socket, 24x 24TB SAS (576TB raw) — ASG-2015S-E1CR24H | 5 | $51,743.45 | $258,717.25 | 3.9 |
| 32× 100G QSFP28 switch — N8560-32C, PicOS (front-end pair, MLAG) | 2 | $7,919 | $15,838 | 0.62 |
| 100G 2 m passive DAC (storage nodes to the pair, 2 per node) | 10 | $54 | $540 | — |
| 100G 2 m passive DAC (peer link between the pair) | 2 | $54 | $108 | — |
| 48× 1G RJ45 + 4× 10G switch — S3410-48TS, PicOS (OOB; Cat6 patches not priced) | 1 | $1,089 | $1,089 | 0.07 |
| Hardware total |  |  | $276,292.25 | 4.59 |
| Colo, 5-year term, $175/kW/month × 4.59 kW |  |  | $803/month |  |
| Colo over 60 months |  |  | $48,180 |  |
| Hardware + colo over 60 months |  |  | $324,472.25 |  |

- Per node: 2U HDD storage node, single socket, 24x 24TB SAS (576TB raw) = 576 TB raw. RAID 6 as two 12-drive groups (10+2) = 0.833 efficiency; plannable = 576 × 0.833 × 0.85 (fill ceiling) = 407.8 TB per node.
- Nodes: ceil(2,000 ÷ 407.8) = 5 nodes = 2,039.0 TB plannable (1.02× the need, inside the 2× overshoot check). Throughput 5 × 5 GB/s ≈ 25 GB/s streaming (estimate); no stated demand, so capacity binds. Each node is one failure domain: a node outage takes its 407.8 TB offline until repaired.
- Alternative, distributed across nodes with Ceph erasure coding 4+2 (0.667 efficiency, 6-host floor): plannable = 576 × N × (N−1)/N × 0.667 × 0.85; N = 7 gives 1,959.4 TB (short), N = 8 gives 2,285.9 TB, so 8 nodes ($413,947.60 for the nodes) plus NVMe DB/WAL drives at 1–4% of HDD capacity via the rate card. Survives a whole node.
- Network: each node has 2× 100G; one port to each switch of an MLAG pair of 32× 100G QSFP28 switches (10 ports of 32 per switch), peer link 2× 100G DAC, uplinks to the buyer's core not priced; the listed −$874 option swaps to 2× 25G if the core is 25G. OOB: 5 nodes + 2 switches on one 48× 1G RJ45 + 4× 10G switch.
- Rack and power: 5 × 2 U + 3 U of switches = 13 U, 4.59 kW provisioned: one rack, one 208V 1-ph 30A circuit pair. 5-year colo $803/month.

Generated 2026-09-22 from UpStation's catalog data; prices as of 2026-09-09; re-checked monthly. Book a meeting: https://calendly.com/alecglover/intro-call. Map of every page: https://ai.upstation.io/llms.txt