# 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. Prices in USD as of 2026-09-09; final price confirmed on order. Delivery: Continental US; Canada quoted case by case. Orders and quotes: book a meeting at https://calendly.com/alecglover/intro-call, or email alec@upstation.io.

This file is the map. Each page below is the same page a person reads; fetch only the ones a question needs. Every page is also plain Markdown at the same URL with `Accept: text/markdown`, or with `.md` appended (the homepage is `/index.md`).

## Conventions

- Price basis: vendor public price (Supermicro eStore, Octoserver, FS.com US), USD, dated 2026-09-09. Final pricing is confirmed by UpStation on order.
- Tiers for the same GPU: Tier 1 is an OEM factory build (one SKU, one warranty). Tier 2 is a system builder (cheaper, in stock). Build-your-own is a host with 8 slots and no GPUs; the buyer sources the cards.
- One priced base build per class. Intel and AMD equivalents are listed with a host delta; the CPU brand is not a price decision. Alternative configurations are not priced separately: resize memory, CPU or drives with the rate card (marginal $/GB, $/core, $/TB) and say so; non-linear items (NICs, drives, warranty, service) use the listed option deltas.
- Status: every line is confirmed unless marked draft, estimate or price pending. Warranty is 3 years parts and labour unless stated; on-site and extended warranty are options.
- Networking: FS.com Broadcom switches with PicOS included; generic-coded optics and cables from FS. Connecting a port: 2 m passive DAC inside passive reach, then active DAC/AEC, then AOC or two short-reach optics plus fibre, then long-reach optics over single-mode (switches page has every reach and price).
- Power: provisioned kW = nameplate W ÷ 0.8 so no circuit runs above 80%; the listed circuit is provisioned twice (A and B feeds). Colo is billed on provisioned kW at $175/kW/month (5-year term), $225/kW/month (3-year term), $250/kW/month (1-year term), US or Canada (power page).
- A quote is a bill of materials: one line per item with quantity, unit price and extended price; then a networking summary (switches, cables, optics), a power summary (kW provisioned, circuits, monthly colo at the chosen term), totals, and the assumptions made. Anything marked quote, outside the rate-card bands, or listed under 'Not in this catalog' is quoted by UpStation rather than estimated. Sizing a deployment (nodes, switches, storage nodes, racks) follows How solutions are sized; the reference builds and worked designs are computed from this data and can be used to check arithmetic.

## Pages

- [UpStation](https://ai.upstation.io/): start here: every price, spec, power and colo table, the rate card, a sizing digest, reference-build totals, Q&A and the quoting conventions (~7,900 tokens as Markdown; HTML about 1.5×)
- [GPU servers (8 GPUs each)](https://ai.upstation.io/gpu-servers): the GPU server table plus every build in full (spec, options, equivalents, power, colo, per-GPU) (~2,500 tokens as Markdown; HTML about 1.5×)
- [GPU hosts without GPUs (build your own; 8 slots)](https://ai.upstation.io/gpu-hosts): 8-slot hosts without GPUs, in full, with the GPU cards that fit them (~2,500 tokens as Markdown; HTML about 1.5×)
- [Switches, cables and optics (FS.com, PicOS included, generic optics)](https://ai.upstation.io/switches): switches in full, every cable and optic with price and reach, fibre patch, link-by-distance rule, PicOS support (~3,700 tokens as Markdown; HTML about 1.5×)
- [Compute nodes](https://ai.upstation.io/compute-nodes): 1U compute node in full (~1,000 tokens as Markdown; HTML about 1.5×)
- [Storage nodes](https://ai.upstation.io/storage-nodes): NVMe and HDD storage nodes in full (~1,500 tokens as Markdown; HTML about 1.5×)
- [Parts rate card (marginal; apply to a base build)](https://ai.upstation.io/rate-card): marginal $/GB, $/core, $/TB for resizing a base build (~1,000 tokens as Markdown; HTML about 1.5×)
- [Power and colo](https://ai.upstation.io/power): power model, circuits, colo rates and the computed kW, circuit and colo for every product (~1,800 tokens as Markdown; HTML about 1.5×)
- [How solutions are sized](https://ai.upstation.io/sizing): the sizing rules: GPUs to nodes, HGX vs PCIe, fabric planes and switch counts, storage feed rates and protection math, racks, management nodes, questions to settle (~3,900 tokens as Markdown; HTML about 1.5×)
- [Reference builds and worked designs](https://ai.upstation.io/designs): reference builds and two worked designs with the arithmetic shown (~2,600 tokens as Markdown; HTML about 1.5×)
- Product pages: one page per product at https://ai.upstation.io/p/<id> (~300–600 tokens each); the id is in the tables and in catalog.json:
  - https://ai.upstation.io/p/UPS-GPU-B300-8: 8x NVIDIA B300 (HGX B300 NVL8), Intel — $572,364
  - https://ai.upstation.io/p/UPS-GPU-B200-8: 8x NVIDIA B200 (HGX B200), Intel or AMD — not stocked, see B300
  - https://ai.upstation.io/p/UPS-GPU-H200-8: 8x NVIDIA H200 (HGX H200), Intel — $319,421.64
  - https://ai.upstation.io/p/UPS-GPU-RTXPRO-8: 8x NVIDIA RTX PRO 6000 Blackwell Server Edition, Intel — $191,888.10
  - https://ai.upstation.io/p/UPS-GPU-RTXPRO-8-T2: 8x NVIDIA RTX PRO 6000 Blackwell Server Edition, AMD — $163,909.60
  - https://ai.upstation.io/p/UPS-GPU-5090-8: 8x NVIDIA RTX 5090 (passive server edition), AMD — $76,679.60
  - https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8: 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel Xeon 6 (TDX) — $47,884.46
  - https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-X13: 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, Intel 5th Gen Xeon (TDX) — $37,937.10
  - https://ai.upstation.io/p/UPS-GPU-BARE-ENT-8-AMD: 8-GPU host, Tier 1 (Supermicro), PCIe 5.0, AMD — $39,179.76
  - https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN5: 8-GPU host, Tier 2 (Octoserver), PCIe 5.0, AMD — $32,919.60
  - https://ai.upstation.io/p/UPS-GPU-BARE-T2-GEN4: 8-GPU host, Tier 2 (Octoserver), PCIe 4.0 budget, AMD — $15,146.20
  - https://ai.upstation.io/p/UPS-NET-800G-64: 800G switch, 64x OSFP (Tomahawk 5) — $44,600
  - https://ai.upstation.io/p/UPS-NET-800G-32: 800G switch, 32x OSFP (Tomahawk 5) — $37,499
  - https://ai.upstation.io/p/UPS-NET-400G-64: 400G switch, 64x QSFP-DD (Tomahawk 4) — $23,999
  - https://ai.upstation.io/p/UPS-NET-400G-32: 400G switch, 32x QSFP-DD (Tomahawk 4) — $19,999
  - https://ai.upstation.io/p/UPS-NET-200G-32: 200G switch, 24x 200G + 8x 400G (Trident 4) — $12,400
  - https://ai.upstation.io/p/UPS-NET-100G-64: 100G switch, 64x QSFP28 (Tomahawk 2) — $11,900
  - https://ai.upstation.io/p/UPS-NET-100G-32: 100G switch, 32x QSFP28 (Trident 3) — $7,919
  - https://ai.upstation.io/p/UPS-NET-25G-48: 25G leaf, 48x SFP28 + 8x 100G — $6,049
  - https://ai.upstation.io/p/UPS-NET-10G-48: 10G leaf, 48x SFP+ + 100G uplinks — $4,569
  - https://ai.upstation.io/p/UPS-NET-1G-48: 1G management switch, 48x RJ45 + 4x 10G SFP+ — $1,089
  - https://ai.upstation.io/p/UPS-CN-INT-1U: 1U compute node, dual socket, 64 cores, 512GB, diskless NVMe (virtualisation host) — $33,298.33
  - https://ai.upstation.io/p/UPS-SN-INT-NVME: 2U all-NVMe storage node, dual socket, 24x 7.68TB NVMe (184TB raw) — $182,994.59
  - https://ai.upstation.io/p/UPS-SN-AMD-HDD: 2U HDD storage node, single socket, 24x 24TB SAS (576TB raw) — $51,743.45

## Data files

- https://ai.upstation.io/catalog.json: everything as JSON with computed fields (list_price, price_per_gpu, provision_kw, circuit, colo_usd_per_month), the sizing rules, reference builds and Q&A (~24,700 tokens); per product: https://ai.upstation.io/p/<id>.json
- https://ai.upstation.io/llms-full.txt: every page above in one Markdown file, for a single fetch or a paste (~27,800 tokens)
- https://ai.upstation.io/brief.txt: under 8,000 characters for small paste boxes

## Which pages for which question

| Question | Fetch |
|---|---|
| Price or spec of one product | its /p/<id> page |
| Compare builds of one GPU, or pick a tier | /gpu-servers |
| Cable, optic or switch choice and cost | /switches |
| Resize memory, CPU or drives | /rate-card |
| Colo cost, circuits, kW | /power (per-product figures) or the product page |
| Design a cluster: nodes, switches, cables, storage, racks | /sizing, then /designs for a worked example, then /gpu-servers, /switches and /storage-nodes for prices |
| Size a storage cluster (capacity, RAID, erasure coding, node count) | /sizing (Storage) and /storage-nodes |
| Anything not listed | the homepage 'Not in this catalog' list, then book a meeting (https://calendly.com/alecglover/intro-call) |

Generated 2026-09-22; prices as of 2026-09-09; re-checked monthly.