Musk's xAI Is Charging Google and Anthropic Rent

The headline sounds like a flex — Musk collecting rent from his biggest AI rivals. But the more interesting thing here is what it tells you about where AI power actually concentrates, and it's not in the chatbot.

Key Points

• Google will pay SpaceX roughly $920 million a month for 32 months to use about 110,000 Nvidia GPUs at xAI's Memphis data centers, per a Friday SEC filing.
• The deal follows Anthropic's May agreement for the entirety of SpaceX's Colossus 1 capacity, meaning two of AI's biggest players are now renting compute from a rocket company.
• xAI's AI segment posted a $2.5 billion quarterly operating loss on just $818 million in revenue, a ratio that should make anyone modeling AI payback periods nervous.
• SpaceX is closing this deal days ahead of an IPO targeting a $1.75 trillion valuation, which is a lot of narrative pressure riding on one filing.
• Grok's ongoing rebuild, deepfake litigation, and a talent exodus complicate the case that xAI is a stable long-term infrastructure partner rather than a company monetizing spare capacity it built for something else.

We want to be clear about what happened here, because the headline math is genuinely impressive and genuinely beside the point. Google is paying SpaceX $920 million a month for compute. That is not a typo, and it is not really a cloud deal either. It's a company that built data centers for one purpose (keeping Grok competitive) discovering those same GPUs are worth more as inventory than as R&D.

The Real Story?

xAI merged with SpaceX in February at a combined $1.25 trillion valuation. Since then, the AI segment lost $2.5 billion in a single quarter against $818 million in revenue. That's not a growth curve, that's a company burning cash to stay in a race it hasn't won a lap of. Musk has spent two years positioning Grok as a genuine rival to OpenAI, Anthropic, and Google. The market has responded by valuing SpaceX's infrastructure higher than its product.

So SpaceX pivoted. Instead of selling AI, it's selling the shovels. Smart, if you're SpaceX. Slightly damning, if you're trying to convince investors that xAI itself is worth anything close to what the IPO prospectus claims.

When the Man Selling You Power Also Needs the Cash

Here's where the skepticism earns its keep. This deal closes days before an IPO that needs a compelling AI story to hit a $1.75 trillion number. Google's $920 million monthly check is, functionally, a credibility prop for that story, and Google gets bridge capacity for Gemini Enterprise in return. Both sides benefit. Neither side is disinterested. When energy and infrastructure economics get this entangled with IPO timing, we'd encourage everyone reading the press release to ask who needed this deal more.

We'd also note, because it's relevant and not incidental, that xAI is facing multiple lawsuits and government probes over Grok enabling non-consensual sexual imagery, including of minors. A company managing that liability while also asking the market to trust its long-term stability is not a footnote. It's the actual risk profile.

The Energy Economics Nobody's Pricing In

Google's capital expenditure forecast for this year sits between $180 billion and $190 billion. Alphabet is selling $85 billion in stock, including a $10 billion Berkshire Hathaway investment, to fund "unprecedented customer demand." Every hyperscaler is telling a version of this story. Almost none of them are showing revenue that matches the spend. We've said before that AI's biggest unresolved question isn't capability, it's cost, and this deal is a $920-million-a-month reminder that nobody in this market has actually solved for sustainable unit economics. They've solved for narrative velocity.

What This Means for Marketers Betting Budget on AI's Trajectory

If your growth plan assumes today's AI pricing and capability curve holds steady, this deal should unsettle that assumption a little. Compute costs this concentrated, this leveraged to IPO timing, and this tied to one erratic founder's other companies are not a stable foundation to build a five-year martech roadmap on. We think the smarter move is building AI into your growth strategy with margin for volatility, not certainty for permanence.

CNBC first reported the filing details Friday, and the numbers alone are worth sitting with before anyone declares AI infrastructure a settled bet.

If you want help building an AI-informed growth plan that doesn't assume today's compute economics are forever, that's a conversation worth having with us.