Elon Musk Dismisses $120B Anthropic Exit Talk as SpaceX Plans $74B AI Push

Musk’s $74 Billion AI Expansion Targets Texas and Arizona

Elon Musk’s SpaceX (Nasdaq: SPCX) announced plans on Nov. 12 to invest more than $74 billion in two AI supercomputing facilities in Texas and Arizona. The expansion followed Musk’s response to speculation about Anthropic’s $120.5 billion computing agreement, linking the company’s ambitious spending plans to questions about future customer demand.

SpaceX outlined its planned computing expansion, allocating more than $24 billion to its Starbase facility in Texas and approximately $50 billion to a proposed campus in Cochise County, Arizona. The Texas project will generate its own electricity rather than rely on the public grid. The company expects more than 500 permanent jobs there, while the Arizona development could create approximately 1,000 permanent positions.

The proposed Arizona facility faces regulatory and safety challenges. Cochise County’s development application describes a 747-acre site with up to nine computing buildings and an estimated electricity demand of 6 gigawatts at full capacity. The project requires county approval and has drawn public opposition.

At a Nov. 12 meeting, the county’s Local Emergency Planning Committee reportedly approved six recommendations addressing fire protection, battery storage, and emergency response capabilities. The advisory panel sought stronger safeguards before development proceeds.

One Customer Accounted for 34.5% of SpaceX’s Quarterly Revenue

The multibillion-dollar investment plans come as SpaceX’s AI operations account for a growing share of its business. Its third-quarter SEC filing showed that the Super Intelligence segment generated $7.56 billion, approximately 52% of total quarterly revenue of $14.58 billion. The segment, which includes computing services, software, and advertising, reported $431 million in operating income, reversing a $1.72 billion loss a year earlier.

Much of that revenue is concentrated among major customers. Bloomberg identified Anthropic as SpaceX’s largest customer, citing a person familiar with the relationship. SpaceX disclosed that a single customer, labeled Customer B, accounted for 34.5% of third-quarter revenue.

That reliance has attracted scrutiny following unverified claims that Anthropic could decline to extend its computing agreement beyond mid-2027. Responding to speculation surrounding the agreement on Nov. 11, Musk stated on X:

“We would be happy to take the compute back. No problem. But I think the opposite is going to happen.”

The speculation emerged as Anthropic prepared for a major initial public offering (IPO). Its Nov. 10 filing proposed selling 825 million shares at $70 to $78 each, representing a potential offering of $57.75 billion to $64.35 billion before underwriting fees. The company applied to list on Nasdaq under the ticker ANT, subject to completion of the offering.

Anthropic’s amended Securities and Exchange Commission filing, submitted Nov. 10, detailed computing agreements with SpaceX.AI LLC worth up to $120.5 billion over terms extending into 2029. Monthly payments are expected to reach approximately $3.1 billion once all contracted capacity becomes available, currently anticipated in December. Either company can terminate the relevant agreements with 90 days’ written notice after initial noncancelable periods expire in January and March 2027.

Massive AI Investments Face Competing Demand Expectations

The expansion and Anthropic agreement reflect SpaceX’s bet on sustained demand for AI computing power. Musk previously defended SpaceX’s long-term ambitions amid criticism of its AI computing arrangements. Venture capitalist Chamath Palihapitiya suggested constrained supplies could strengthen the company’s negotiating position, while Musk acknowledged that cheaper, more abundant capacity could reduce revenue.

Anthropic also maintains substantial computing relationships with Google, Amazon, Microsoft, and other providers. Beyond these centralized suppliers, decentralized AI computing networks offer an alternative model that distributes processing capacity among independent operators. Such networks remain distinct from the large corporate facilities supporting Anthropic’s agreements.

SpaceX is separately pursuing Musk’s plans for space-based AI computing, including orbital data centers designed to operate beyond terrestrial infrastructure. The company faces two significant tests: whether customers continue purchasing sufficient computing capacity and whether its proposed facilities secure regulatory approvals. The Arizona project faces its next review on Dec. 9, when county officials will consider the development amid public opposition and emergency preparedness concerns.