Cost Breakdown of a 1GW AI Data Center
teortaxesTex · x · 2026-07-13
Citing an Epoch AI breakdown, the post discusses the upfront costs of a 1GW AI data center, which are around $37.2 billion. The most expensive components are servers and GPUs, rather than land and power.
Key points include:
- Servers account for about $21 billion, with GPUs taking the lion's share
- Facilities/buildings cost around $11 billion
- Network infrastructure is roughly $4.9 billion
- Land and power access combined are only about $330 million, less than 1%
The author argues that at $21,000 per kilowatt, such an AI data center is "actually quite cheap," and believes there is room for NVIDIA-related costs to double or triple in the future.
More from Infra
- Resin lets an agent build standalone LLM inference engines; vLLM and SGLang optimize DeepSeek kernels differently — AccBalanced · 2026-10-27
- Local benchmark: Flint matches Qwen3.6 quality at 50/52 while running 3.6x faster on M4 Pro — RichardSocher · 2026-10-27
- Alchemy adds effect/Workflow support across Durable Objects, Workflows and Lambda Durable Functions — samgoodwin89 · 2026-10-27
- Alchemy's Lambda Web Functions run plain Node web servers on Lambda, no OpenNext needed — samgoodwin89 · 2026-10-27
- Foldkit Adds SSR Support With a Few Lines of Config on Cloudflare and AWS — samgoodwin89 · 2026-10-27
- Alchemy Cuts Cloudflare Bindings CPU Time by Up to 166× Via Profiling — samgoodwin89 · 2026-10-27