#AnthropicDiscloses$84.5BComputeDealWithSpaceX $ANTHROPIC

The detail that actually stopped me today is this: xAI, the company whose infrastructure Anthropic is now paying up to $84.5 billion to use, was founded specifically to compete with Anthropic. Musk called Anthropic "evil" on X multiple times in the run-up to this arrangement. Colossus, the data center at the center of this whole deal, was originally built to train Grok, xAI's own model, the one meant to beat Claude. Eighteen months later, Anthropic is renting that same facility.

I don't think that detail changes the financial logic of the deal, compute is compute, and if the capacity is there and the price works, business is business. But it tells you something about how tight the actual supply of usable AI infrastructure has gotten. Companies that were explicitly built to be rivals are now customers of each other because there simply isn't enough compute to go around otherwise. That's a more interesting signal about the state of the industry than the dollar figure itself.

The other thing I hadn't fully appreciated is just how violent SPCX's stock has been since its IPO. It peaked above $225 in June, then crashed 31% in four sessions, fell further to a record low around $145, and has been swinging through a $62 to $900 range of analyst price targets ever since, probably the widest spread of disagreement I've seen on any recent mega-cap listing. Against that backdrop, a 2.59% bump to $149.24 on this disclosure is genuinely small noise. The stock trading near its all-time low right now, even with this large compute deal attached to it, tells me the market isn't rewarding the AI business the way SpaceX might want.

Morgan Stanley's framing stuck with me: at $100 a share, the market would basically be assigning zero value to SpaceX's entire AI segment, Colossus, the Anthropic contract, the Alphabet contract (Alphabet is reportedly paying SpaceX $920 million a month separately for its own GPU access), all of it. The stock isn't at $100 right now, but it's been close enough to that zone that the question is live. Either the market thinks these compute contracts won't hold up, or it just hasn't gotten around to pricing AI infrastructure leasing as a real, durable business line yet.

There's also a sharp contrast in how Musk himself has described the arrangement at different points. In one breath it's being framed as a six-month initial term with a possible multi-year extension, in another as something that was never committed to for years at all. That kind of inconsistency in how the person running the company describes his own flagship AI deal is, to me, more telling than the 90-day termination clause everyone's focused on. If the person selling the compute can't settle on how long the commitment actually is, I'd treat the $84.5 billion figure as a ceiling that depends heavily on both sides wanting to keep renewing, not a number either party is locked into.

For Anthropic's side of this, the strategic logic still makes sense even with the awkward history. Compute availability is apparently the real constraint on their growth right now, not demand, so renting capacity from a direct competitor's infrastructure is just pragmatic if it's the fastest way to get GPUs online. Whether that counts as a smart hedge or, as one critic put it, "the biggest unforced error of the AI era," probably depends on whether SpaceX's AI business proves durable enough that Anthropic isn't left scrambling if the relationship sours or the capacity gets reallocated elsewhere.

What I'm actually watching now isn't the $84.5 billion number anymore. It's whether SPCX can hold any kind of floor above its recent lows as more detail about these AI contracts becomes public, since that will tell us whether the market is starting to believe in the AI segment's value, or still treating it as an afterthought bolted onto a space company.

Does the rivalry history here change how durable you think this compute relationship actually is, or is that irrelevant once real money and GPU access are on the table?

Not financial advice, always do your own research before making any trading or investment decision.