Detailed Analysis
Elon Musk's decision to lease Anthropic access to the world's largest supercomputer — just three months after publicly labeling the AI lab "evil," "misanthropic," and hostile to Western civilization — represents one of the more striking pivots in the recent history of technology dealmaking. The arrangement, announced in May 2026, will see Anthropic use computing infrastructure owned by SpaceX, with analysts estimating it will generate between $3 billion and $4 billion in annual revenue for Musk's rocket and aerospace company. The deal coincides with Musk's merger of his AI venture xAI into SpaceX to form "SpaceXAI," consolidating his AI ambitions under the corporate umbrella he is preparing to take public.
The financial logic underlying the deal is straightforward, even if the optics are jarring. SpaceX has already borne the capital costs of constructing the supercomputer facility, meaning the incremental cost of leasing it to Anthropic is largely limited to electricity and staffing. According to Antoine Chkaiban of New Street Research, this structure produces over $2.5 billion in cash profit annually — margins that appear extreme on the surface but are entirely consistent with the economics of already-depreciated data center infrastructure. Musk's about-face on Anthropic, in this reading, is not an ideological conversion but a straightforward asset utilization decision: idle GPU clusters are expensive, and a marquee tenant transforms a sunk cost into a high-margin revenue stream.
The deal's deeper significance lies in what it does for SpaceX's forthcoming IPO. With a confidential S-1 filed in early April 2026 and a public roadshow expected to begin within weeks, SpaceX is targeting a valuation between $1.75 trillion and $2 trillion. Before the Anthropic announcement, the company's nascent cloud-infrastructure business lacked the kind of credible, high-profile customer that institutional investors typically demand before assigning premium multiples to a new revenue segment. Anthropic — one of the most heavily funded and closely watched AI safety companies in the world, backed by billions from Google and Amazon — provides exactly that validation. The timing of the announcement, synchronized with the xAI merger, suggests deliberate IPO narrative construction rather than organic commercial development.
For Anthropic, the arrangement speaks to the acute and ongoing scarcity of AI compute at the frontier. The company, founded in 2021 by former OpenAI researchers including Dario and Daniela Amodei, has consistently required massive computational resources to train and run its Claude model family. Securing a dedicated pipeline to the world's largest supercomputer — regardless of the identity or prior statements of its owner — addresses a fundamental constraint on the company's ability to remain competitive with OpenAI, Google DeepMind, and Meta AI. That Anthropic would transact with a counterparty who had recently characterized it as civilizationally hostile underscores how thoroughly compute access functions as a strategic imperative that overrides reputational considerations in the current AI race.
Taken together, the episode illustrates several converging dynamics in the AI industry as of mid-2026: the increasing financialization of AI infrastructure, in which data centers and GPU clusters function as yield-generating real estate assets; the blurring of competitive and cooperative relationships among AI's major players; and the extent to which IPO preparation is shaping corporate behavior and deal flow across the sector. Musk's rhetorical reversal on Anthropic is less a reflection of changed views about the company than a demonstration of how the imperatives of capital markets can override ideological posturing when billions of dollars in annualized profit are on the table.
Read original article →