中
◀ Back to feed
TechCrunch

Crusoe sells troubled Wyoming data center project to Google

Published 2026-11-06 Collected 2026-11-07 12:00 AI news #ai#industry
Full text

Crusoe sells troubled Wyoming data center project to Google

Marina Temkin · November 6, 2026 · TechCrunch

Crusoe, an eight-year-old, Denver-based data center developer recently valued at $30.9 billion, confirmed Friday it has sold an unfinished data center project in southeastern Wyoming to Google.

Google, which Crusoe had tried and failed to land as a customer for the project, is taking over the development after reportedly raising concerns about its cost and construction timeline under Crusoe’s management.

Crusoe initially announced the project last year with Blackstone-backed infrastructure company Tallgrass, planning to eventually scale it to 10 gigawatts of power capacity.

Google did not disclose how much it paid Crusoe for the unfinished project, but it announced plans to invest $17 billion to develop the data center campus.

The deal highlights how difficult it can be for independent data center developers to deliver massive projects on time and within budget, given the challenges of securing equipment and power. Growing opposition from local communities can further complicate development.

Despite taking over the Wyoming facility, Google hasn’t walked away from working with Crusoe entirely. Crusoe is developing another data center for Google near Amarillo, Texas.

The deal to sell the site to Google was first reported by Bloomberg in June, but the Wyoming setback didn’t stop Crusoe from raising a $3.9 billion round in September.

Crusoe makes money by leasing data center space to customers that bring their own GPUs, by renting out its own GPUs, and by selling compute power used to run AI models, known as inference.

Crusoe’s largest projects serve some of the biggest names in AI. Its campus in Abilene, Texas, is being built in phases for Oracle and OpenAI, and the company is also constructing a separate 900-megawatt campus next door for Microsoft, with its first building expected to come online in mid-2027. Meta is also a customer, and Bloomberg reported that the trading firm Jane Street signed a $13 billion, five-year cloud deal with Crusoe.

Source
Title
Crusoe sells troubled Wyoming data center project to Google
Author
Marina Temkin
Site
TechCrunch
Published
2026-11-06
Collected
2026-11-07 12:00 · 2026-11-07-12-ai.json
Link
https://techcrunch.com/2026/11/06/crusoe-sells-troubled-wyoming-data-center-project-to-google/
Open original ↗

Reposted. Will be removed on request. All rights belong to the original author.

More from this batch

12:00batch · 2026-11-0712 items
>

Measuring how AI agents handle everyday wet-lab science

OpenAI+Ligo BioBasicsBench: 100 raw wet-lab tasks; GPT-6 Astra ~66.7% at xhigh vs 43.8% for GPT-5.6 Sol.

OpenAI
>

OpenAI×Frontier Model Forum:给「绕过防护是否可能造成严重网络伤害」定级框架

官方帖:与 FMF 成员机构一起发布框架,帮助判断绕过 AI 防护是否可能促成严重网络危害,并优先加强防护。属安全评测方法论,非新模型发布;可与 Vals CyberOffense 等攻防评测对照。

OpenAI / X
>

vork by Voyager: a headless 3D and VFX engine for agents

Research-preview headless 3D/VFX engine: agents write scenes as code (~2.5k-token API), render on GPU, inspect, revise.

Voyager
>

Claude Duet 手机端可监控与审批 Claude Code 会话(本地与云端)

多条社区与日报称:Claude Duet App 已能查看 Claude Code 会话状态、审批改动、发送更新,覆盖本地与云会话。对「人在回路」远程盯编码 Agent 有直接产品含义;以 Anthropic 客户端实际能力为准。

社区 / X 趋势
>

Linus Torvalds 批强制披露「是否用了 AI」:结果比工具更重要

X 趋势话题:Torvalds 在编码相关讨论中称强制 AI 使用披露「utterly stupid」,强调看结果而非工具。属社区言论摘要,细节以原帖/邮件列表为准;对企业合规「AI 生成标记」政策是反面对照。

X 趋势 / 内核社区
>

建设者把重心放在 Agent Harness(工具/记忆/子代理/校验),而非「完美模型」

趋势摘要:多位建设者强调用 harness 组合让 Agent 能稳定跑数小时任务。可与本期 Harness-S 论文、ECC 等仓库对照;属舆论主题而非单一产品发布。

X 趋势
>

Harness-S: Learning from Sibling Harnesses for Generalizable Agent Training

Selective sibling-harness rescue transfers complementary behaviors into one policy that still works under the original harness.

arXiv / HF
>

VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding

Parallel primitive decoding cuts trajectory error and latency; VFlash-conditioned WAM hits 70.2% on MolmoSpaces suites.

arXiv / NVIDIA 等
>

Decoupled Predictive Decision Making:共享条件预测器服务多 Agent 决策

改进什么:多 Agent 各自学预测易冗余、难共享。怎么做:解耦出共享条件预测器,决策头按需使用。HF↑114,适合对照多智能体世界模型/规划栈。

HF Papers
>

SPECMEM:用跨会话活记忆加速工具调用的投机解码草稿

改进什么:工具调用仍自回归解码,延迟高;现有系统无跨会话草稿库或建完不更新。怎么做:按用户重复行为维护可更新的草稿记忆,服务投机解码。面向部署侧工具 Agent 延迟。

arXiv
>

Tools Need Not Speak in Text:工具结果用隐式上下文压缩,少占对话 token

改进什么:工具返回全进文本上下文又贵又吵。怎么做:对工具观测做隐式压缩再注入 Agent。适合长工具链企业 Agent 控成本。

HF Papers
>

Ask, Don't Equip:问答驱动的编码 Agent,做深层内核优化

改进什么:给 Agent 堆满工具不一定等于会做系统级优化。怎么做:以提问驱动探索与验证,面向内核等深层优化任务。对照「工具越多越好」直觉。

HF Papers