数据库分类 · 14 记录
数据中心
The facilities: hyperscale campuses, colocation halls and GPU-cloud sites, with reported capacities, tenants, cooling and power sources. Figures are as reported by operators or credible press — each record labels its basis.
现场笔记
这里的每个设施都以不同方式回答同三个问题：电力从哪里来、热量如何排出、谁是主力租户？读两三条记录之后，你会在点击之前就开始猜答案——这正是这里的意义所在。
从这三个开始
SA
Stargate Abilene
这个时代的标志性建设——八栋楼，自备涡轮机，OpenAI入驻其中。
XC
xAI Colossus
12.2万张GPU，122天建成。每个建设者都在追赶的速度纪录。
MF
Microsoft Fairwater
范本：闭环冷却、GB200集群，如今正逐城复制。
Hyperscale & AI campuses
9
SA
Stargate Abilene
The flagship of the $500B Stargate program — OpenAI's first purpose-built campus.
XC
xAI Colossus
100,000 GPUs in 122 days — the build that reset the industry's clock.
AN
AWS New Carlisle (Project Rainier)
Anthropic's training ground: a Trainium2 mega-campus in the Indiana corn belt.
MH
Meta Hyperion
A campus the size of Manhattan: 5 GW announced scale in rural Louisiana.
MF
Microsoft Fairwater
Microsoft's 'most powerful AI datacenter' — and a template it is replicating.
MP
Meta Prometheus
Meta's first gigawatt-class cluster, slated online in 2026.
SU
Stargate UAE
The 5 GW Gulf out-station of the Stargate program.
TC
Tesla Cortex
The GPU cluster at Giga Texas training Full Self-Driving.
XC
xAI Colossus 2
The gigawatt sequel: xAI's push toward a million-GPU class site.
Colocation & wholesale
2
EA
Equinix Ashburn (Data Center Alley)
The most connected place on the internet — where AI meets the network.
SC
Switch Citadel Campus
One of the world's largest planned colocation campuses — and Oklo's first customer site region.
