Notes on the xAI/Anthropic data center deal

7th May 2026
There weren’t a lot of big new announcements from Anthropic at yesterday’s Code w/ Claude event, but the biggest by far was the deal they’ve struck with SpaceX/xAI to use “all of the capacity of their Colossus data center”.
As I mentioned in my live blog of the keynote, that’s the one with the particularly bad environmental record. The gas turbines installed to power the facility initially ran without Clean Air Act permits or pollution control devices, which they got away with by classifying them as “temporary”. Credible reports link it to increases in hospital admissions relating to low air quality.
Andy Masley, one of the most prolific voices pushing back against misleading rhetoric about data centers (see The AI water issue is fake and Data center land issues are fake), had this to say about Colossus:
I would simply not run my computing out of this specific data center
I get that Anthropic are severely compute-constrained, but in a world where the very existence of “AI data centers” is a red-hot political issue (see recent news out of Utah for a fresh example), signing up with this particular data center is a really bad look.
There was a lot of initial chatter about how this meant xAI were clearly giving up on their own Grok models, since all of their capacity would be sold to Anthropic instead. That was a misconception—Anthropic are getting Colossus 1, but xAI are keeping their larger Colossus 2 data center for their own work.
As an interesting side note, the night before the Anthropic announcement, xAI sent out a deprecation notice for Grok 4.1 Fast and several other models providing just two weeks’ notice before shutdown, reported here by @xlr8harder from SpeechMap:
This is terrible @xai. I just spent time and money to migrate to grok 4.1 fast, and you’re disabling it with less than two weeks notice, after releasing it in November, with no migration path to a fast/cheap alternative.
I will never depend on one of your products again.
Here’s SpeechMap’s detailed explanation of how they selected Grok 4.1 Fast for their project in March.
Were xAI serving those models out of Colossus 1?
xAI owner Elon Musk (who previously delighted in calling Anthropic “Misanthropic”) tweeted the following:
By way of background for those who care, I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed. [...]
After that, I was ok leasing Colossus 1 to Anthropic, as SpaceXAI had already moved training to Colossus 2.
And then shortly afterwards:
Just as SpaceX launches hundreds of satellites for competitors with fair terms and pricing, we will provide compute to AI companies that are taking the right steps to ensure it is good for humanity.
We reserve the right to reclaim the compute if their AI engages in actions that harm humanity.
Presumably the criteria for “harm humanity” are decided by Elon himself. Sounds like a new form of supply chain risk for Anthropic to me!
More recent articles
- Live blog: Code w/ Claude 2026 - 6th May 2026
- Vibe coding and agentic engineering are getting closer than I'd like - 6th May 2026

Facts Only

* Anthropic struck a deal with SpaceX/xAI to use the capacity of the Colossus data center.
* The Colossus facility has an environmental record related to gas turbines, which initially ran without Clean Air Act permits or pollution control devices.
* Reports link the facility's air quality issues to increases in hospital admissions.
* Andy Masley stated he would not run computing out of this specific data center due to the environmental record.
* xAI was keeping their larger Colossus 2 data center for their own work, while Anthropic acquired Colossus 1.
* xAI sent a deprecation notice for Grok 4.1 Fast, providing only two weeks’ notice before shutdown.
* Elon Musk stated they reserve the right to reclaim compute if their AI engages in actions that harm humanity.

Executive Summary

Anthropic struck a deal with SpaceX/xAI to utilize the full capacity of the Colossus data center. This facility has drawn criticism regarding its environmental record, specifically concerning gas turbines that initially operated without Clean Air Act permits or pollution control devices. A voice critical of the data center practice argued that running computing out of this facility is problematic given the environmental record and the political context surrounding AI data centers. The information also details internal tensions within xAI regarding model deprecation, specifically the shutdown notice for Grok 4.1 Fast, which was issued with minimal notice. Elon Musk commented on the deal, suggesting the right to reclaim compute if the AI engages in actions that harm humanity.

Full Take

The narrative frames the transfer of high-capacity compute as a transactional exchange, juxtaposing technological progress with regulatory and environmental costs. The underlying pattern relies on positioning the AI infrastructure as a zero-sum resource subject to external, often selectively applied, ethical criteria. This creates a systemic risk by subordinating environmental accountability and data center location to the operational needs of the AI ecosystem. The conflict between the stated goal of ensuring AI is "good for humanity" and the practical reality of resource allocation (as articulated by Musk) highlights a potential divergence between aspirational rhetoric and concrete supply chain risk. This structure exploits the authority of technological leaders to establish new, self-serving criteria for resource distribution, which can function as a mechanism for evasion of external accountability. The focus on the deprecation of Grok models alongside the infrastructure deal introduces a secondary pattern: the prioritization of proprietary model development over transparent, immediate migration paths, leveraging urgency to mitigate criticism rather than address it. The implied cost is borne by the environment and public health, while the benefit is concentrated among the entities controlling the compute.

Sentinel — Human

Confidence

This text is a piece of highly specific, opinionated commentary that synthesizes external reports and quotes, indicating a strong human editorial presence.

Signals Detected

Erratic sentence length and highly varied rhetorical flow, consistent with a personal blog or commentary.

Presence of strong, idiosyncratic emphasis and passionate framing, typical of a specific authorial voice rather than neutral aggregation.

Heavy reliance on specific, cited sources (names, handles, specific quotes) points to human research and synthesis, not template matching.

Specific, highly detailed, and interconnected claims regarding model deprecation and internal company dynamics suggest specific, non-generic knowledge.

Human Indicators

The text features a distinct, opinionated, and highly engaging voice that weaves complex, often contradictory, claims together.

The inclusion of specific, cited sources (@xlr8harder, SpeechMap) and direct quotes from figures like Elon Musk and Andy Masley indicates original research and commentary rather than generalized AI output.

The argument is framed around synthesizing internal corporate events and external political/environmental concerns, requiring contextual understanding beyond simple data recitation.