Musk: Suing OpenAI while leasing computing power to Claude

Last night, SpaceXAI announced a computing resource partnership with Anthropic, granting access to the full capacity of Colossus 1 — a Memphis data center with over 220,000 NVIDIA GPUs — delivering approximately 300 megawatts of new compute for Claude within a month. The added power will expand Claude Pro, Claude Max, Claude Code, API, and Opus service limits, with both parties exploring future gigawatt-class orbital AI compute.

The deal unfolds against two backdrops: xAI has been acquired by SpaceX, consolidating rocket, satellite, and AI capabilities under SpaceXAI; and Elon Musk’s ongoing lawsuit against OpenAI, challenging its shift to a for-profit model. By supplying compute to OpenAI’s key competitor, Musk exerts pressure on both legal and commercial fronts.

Yet beyond rivalry, the partnership signals a strategic pivot. With training workloads moved to the newer Colossus 2, Colossus 1 becomes a commercial asset, monetizing idle capacity. For SpaceXAI, this boosts utilization, creates a new revenue stream, and burnishes its credentials as a top-tier infrastructure provider. It marks a shift from model-centric competition (Grok) toward controlling the AI infrastructure layer — akin to an arms dealer — aiming to be an indispensable upstream player irrespective of which model wins.

Compiled by: PANews Big Pliers

Last night, SpaceX AI and Anthropic reached a computing resource partnership.

SpaceX AI will open up the full computing power of Colossus 1 to Anthropic—this data center in Memphis, Tennessee, equipped with more than 220,000 NVIDIA GPUs (including H100, H200 and GB200), which will bring about 300 megawatts of new computing power to Claude within a month.

For Anthropic, this computing power will be used to expand the service capacity of Claude Pro and Claude Max, and to increase the usage limits of Claude Code, Claude API, and Claude Opus. Both parties also revealed their intention to jointly explore future gigawatt-level orbital AI computing capabilities.

This collaboration took place against two unavoidable backgrounds.

First, xAI has been incorporated into SpaceX.

In February 2026, SpaceX completed its acquisition of xAI, integrating its rocket and satellite businesses with the AI company behind Grok into a single system. The transaction was valued at approximately $1 trillion for SpaceX and approximately $250 billion for xAI.

Second, the legal dispute between Musk and OpenAI is still ongoing.

Musk sued OpenAI in 2024, accusing it of improperly shifting towards for-profit and deviating from its philanthropic mission, and demanding that it return to a non-profit structure. The case is currently in its second week of hearings.

Putting these two contexts together, it's not hard to discern a certain logic: SpaceX AI providing its full computing power to Anthropic—OpenAI's most important competitor in enterprise services, programming tools, and developer scenarios—will objectively enhance Claude's service capabilities and increase competitive pressure on OpenAI in the enterprise and developer markets.

Musk is putting pressure on OpenAI on two levels simultaneously: legally, by challenging its organizational structure and commercialization path; and commercially, by supplying computing power to its core competitors.

But "avenging OpenAI" is only half of this collaboration.

Musk provided the other half of the explanation on X: xAI has moved its training to the newer Colossus 2, so Colossus 1 can be rented out.

The meaning of this statement is straightforward: Colossus 1 has transformed from an internal asset into a commercial product.

The operating costs of large GPU clusters are continuous—electricity, cooling, networking, and maintenance are incurred daily. If they are only used to serve a single model product, asset utilization and returns will be suppressed. Leasing computing power to leading AI companies is the most direct way to convert fixed costs into sustainable revenue.

Anthropic is an ideal customer: they have clear computing power needs, strong purchasing power, and sufficient capacity to fill the entire capacity of Colossus 1.

For SpaceXAI, this deal offers at least four benefits:

Improved utilization : Colossus 1 no longer idles;
Establish new revenue streams : Computing power leasing becomes an independent business;
Strengthening its infrastructure positioning : demonstrating SpaceX AI's ability to serve top AI companies;
Increased pressure on OpenAI : Claude's computing power boosts, altering the competitive landscape.

SpaceX AI's business boundaries are quietly shifting.

When xAI operates independently, its core objective is singular: to develop models and use Grok to compete with ChatGPT and Claude at the application layer.

After being incorporated into SpaceX, xAI was placed into a completely different system—data centers, GPU clusters, energy, the Starlink network, launch capabilities, and future orbital AI computing, all on the same balance sheet.

The logic behind this system is not to "win the model war," but to control the upstream of the AI industry .

Grok continues to compete in the model race, while Colossus begins to take on the infrastructure business. Even if Grok does not immediately surpass ChatGPT in model capabilities or user scale, SpaceX AI can still occupy an upstream position through computing power services—just as Nvidia did not need to win the AI application competition back then, but only needed to become a supplier that every player had to rely on.

Claude's access to SpaceX AI's computing power, while ostensibly a competitive move, is essentially a signal of transformation:

SpaceXAI is no longer just a model player betting on Grok to beat ChatGPT. It's trying to become the computing power arms dealer of the AI industry.

Musk's integration of rockets, satellites, data centers, and AI into a single system may not be aimed at simply winning a model war, but rather at becoming the indispensable player in the next round of AI infrastructure competition.