Your AI Vendor's Landlord Might Be Its Rival

Elon Musk stacked his companies into one tower this year. SpaceX bought xAI, renamed the AI unit SpaceXAI, and went public at a valuation now around $2.1 trillion. That is a headline. The part that hits your roadmap is quieter: the same company selling AI infrastructure to Anthropic and Google also competes with them. Power is bunching up in a few hands, fast. Let me catch you up on what that means for the tools your product leans on.

When the landlord is also the competitor

Anthropic signed a deal in May to buy the entire output of xAI's Colossus 1 data center near Memphis. The terms are steep: $1.25 billion a month through 2029, worth about $40 billion to SpaceX. Google signed too, at $920 million a month. So two of the biggest names in AI now rent their compute from a company that also builds models.

Users on X asked the obvious question: could Musk boot a rival off his servers to hurt it? He called that "not my style" and pointed to Tesla opening its Supercharger network as proof. Maybe. But he also admitted under oath that "generally AI companies distill other AI companies," meaning they probe each other's models to learn how they work. Hosting a rival's compute gives you a close look at how it runs.

If your product sits on top of Anthropic or Google, your supply chain now runs through their competitor's building. That is a new layer of risk you did not sign up for.

The device that could move your users

SpaceX showed investors a prototype of a "handset-like" AI device, slimmer than an iPhone and running its own operating system with xAI baked in. Musk called the report "utterly false," which is the same denial he gave last year before this prototype reportedly showed up in front of investors.

The point is the pattern. SpaceX has the factories and the chips to mass produce hardware, and OpenAI is chasing the same idea with Jony Ive. A proprietary OS means your app would live inside someone else's platform again, with their AI as the default. The Humane and Rabbit graveyard says consumers may not bite. But if one of these lands, your distribution rules change overnight.

Owning the whole stack, chip to satellite

Musk is not just renting out data centers. He is buying the pieces that make them faster. SpaceX moved to acquire Mesh Optical Technologies, a startup from three former SpaceX engineers building light-based hardware that moves data quicker and cheaper than the old electrical kind. The FTC fast-tracked the review.

This is the real game underneath the drama. Owning the chips, the connections, and the buildings lets you set the price and the terms for everyone renting from you. And Musk is not alone. Meta starts making its own AI chips in September to cut its Nvidia bill, on top of $125 to $145 billion in planned spending this year. OpenAI and Anthropic are building their own silicon too. The companies you depend on are all trying to own more of the stack.

Everyone wants off the middleman

Apple is playing the same hand from a different seat. It signed a $30 billion deal with Broadcom for 15 billion U.S.-made wireless chips, with $1.5 billion going to expand a Colorado plant. Part of it answers political pressure. Part of it is the same instinct as Meta and Musk: control your own parts so no one else can squeeze you.

That instinct trickles down to you. The vendors under your product are working to depend on fewer outside players. You should be asking the same question about them.

The deep cut

The risk is not that Musk flips a switch and kills Anthropic tomorrow. Contracts and $40 billion make that unlikely. The risk is slower and cheaper to ignore: your product's whole supply chain may run through two or three companies that compete with each other and with the tool you actually build on. When one of them owns the chips, the data center, and maybe the phone in your user's hand, they set the price and the rules, and you find out after the fact.

So map it now. Write down which vendor powers your core AI feature, who owns the compute under that vendor, and what happens to your product if the price doubles or the terms change. If that chain has a single point where a rival controls the pipe, that is your exposure. Bring the map to your next review before someone else draws it for you.

Three questions for your team

1. If our main AI vendor's compute is hosted by its own competitor, what is our fallback, and how fast could we switch?
2. Which single company, if it changed prices or terms tomorrow, would break our product economics, and do we have a second source lined up?
3. If a new AI-native device took even 5 percent of our users off the platforms we ship on today, what would we lose, and what would we do about it?