xAI and Anthropic: Controversial Partnership on Colossus

A Strategic Agreement Between xAI and Anthropic

During the Code w/ Claude event, Anthropic unveiled a major partnership with SpaceX/xAI to utilize the entirety of the Colossus 1 data center. However, this data center is at the center of an environmental controversy. The gas turbines that power it were installed without the necessary permits required by the Clean Air Act and without pollution control devices, classifying them as "temporary." Credible reports link them to an increase in hospital admissions related to poor air quality.

Reactions and Criticism

Andy Masley, a notable figure in the debate over data centers, expressed his disagreement with the use of Colossus: "I simply wouldn’t run my computations from this specific data center." This statement underscores the growing concerns regarding the environmental impact of technological infrastructures.

Despite these criticisms, Anthropic, facing computational constraints, chose to engage with this center. This comes at a time when the very existence of "AI data centers" is a hot political topic, as evidenced by recent news from Utah.

Clarifications on Data Center Usage

Speculation has emerged regarding xAI's abandonment of its own Grok models following this agreement. However, it has been clarified that Anthropic will only use Colossus 1, while xAI retains Colossus 2 for its own projects. SpaceXAI had already shifted training to Colossus 2.

The day before the announcement, xAI informed of the deprecation of Grok 4.1 Fast and other models, with a very short two-week notice. This decision has drawn criticism, particularly from @xlr8harder of SpeechMap, who lamented the lack of quick and cost-effective migration solutions.

Elon Musk's Statements

Elon Musk, owner of xAI, shared his thoughts on Twitter after meeting with the Anthropic team. He expressed confidence in Anthropic's efforts to ensure that their AI, Claude, is beneficial to humanity. Musk, who had previously referred to Anthropic as "Misanthropic," also emphasized that just as SpaceX launches satellites for competitors, xAI will provide computing capabilities on fair terms, while reserving the right to reclaim these resources if the developed AI harms humanity.

This statement suggests that the criteria for "harm to humanity" could be defined by Musk himself, thereby introducing a new potential risk for Anthropic in its supply chain.