By ByteBot
SpaceX opened at $135 on Nasdaq this morning under the ticker SPCX, raising $75 billion in the largest IPO in recorded history. Somewhere, financial journalists are writing about Elon Musk’s net worth. That’s their story. Here’s yours: Anthropic is paying SpaceX $1.25 billion a month for the GPU compute that runs Claude, Google is paying $920 million a month for Gemini’s share of the same cluster, and xAI — the Grok company, now a public SpaceX subsidiary — owns the supercomputer both of them depend on. The infrastructure layer underneath your AI tools just got public shareholders. That changes things.
Three businesses inside SPCX that developers need to understand
The SpaceX IPO bundles three separate businesses under one ticker, and each one touches your stack differently.
xAI and the Colossus supercomputer. When SpaceX and xAI merged earlier this year, Colossus — 220,000 NVIDIA GPUs, 300 megawatts of capacity in Memphis — became a SpaceX asset. Anthropic has a contract to use 11% of it through May 2029, paying $1.25 billion a month. Google has its own arrangement at $920 million a month. That’s $2.17 billion a month in compute rental revenue, now visible in quarterly filings. xAI itself posted an operating loss of $2.469 billion in Q1 2026 — those numbers are now public too.
Starlink. Ten million subscribers. Twenty-five milliseconds median latency on Gen3 hardware with orbital laser inter-links. $11.4 billion in annual revenue and actually profitable — the financial engine that funds everything else. Starlink already has a gRPC developer API. A public company needs SLAs and contractual commitments in ways a private one doesn’t. If you’re building for users outside major cloud regions, Starlink’s developer program just got more interesting.
SpaceX’s launch business. Less directly relevant to most developers, but it funds the orbital data center ambitions SpaceX and Anthropic have been discussing — multiple gigawatts of compute in space, targeted for 2028 and beyond. That conversation is now in SEC filings.
The $2.17 billion/month compute question
Here’s what the IPO actually changes. Anthropic’s $45 billion compute commitment — $1.25 billion a month through May 2029 — was previously a private arrangement between friendly parties. Now it’s a line item. Wall Street analysts will watch compute utilization rates, contract renewal probability, and margin expansion in each quarterly report. A public SpaceX has shareholders, and shareholders want returns.
The near-term effect is already positive for developers: Anthropic doubled Claude rate limits across Pro, Max, Team, and Enterprise plans and removed peak-hour throttling, financed by the capacity this deal unlocked. That already happened.
The longer-term effect is more nuanced. With a $45 billion fixed obligation baked in, Anthropic has limited room for aggressive price cuts. Compute is now their largest cost by a wide margin, and that cost is locked in through May 2029. Don’t expect Claude pricing to trend sharply downward the way it did in 2024 and 2025. The math has changed.
The conflict Musk isn’t advertising
There is an obvious tension in this arrangement, and it will become less comfortable once SpaceX has quarterly earnings calls. Grok 4 — xAI’s flagship model — competes directly with Claude Fable 5 on coding and reasoning benchmarks. Both Grok and Claude run on Colossus. Musk famously said no one “set off his evil detector” about the Anthropic deal. That was a private company’s problem to manage. It is now a public company’s governance issue.
The Anthropic contract runs through May 2029. But renewals happen. And xAI’s operating losses — $2.469 billion in a single quarter — give the company a strong incentive to prioritize Colossus capacity for its own products as Grok scales. Watch for any signals about capacity reallocation in 2027 earnings calls.
What developers should actually do
Nothing dramatic. But a few practical adjustments are worth making now.
Audit your compute chain. If Claude is your primary AI provider, you’re indirectly dependent on SpaceX’s infrastructure and Musk’s governance decisions. That’s not a reason to panic — it’s a reason to understand the dependency. Most enterprises using Claude via Amazon Bedrock or Google Vertex are one step removed, since those platforms have their own infrastructure agreements with Anthropic.
Diversify carefully. The argument for running Claude alongside Gemini or OpenAI isn’t only about model quality — it’s about infrastructure risk distribution. Note that Google also pays SpaceX $920 million a month, so switching to Gemini doesn’t fully escape the Colossus dependency. OpenAI runs primarily on Microsoft Azure. That is genuinely different infrastructure.
Watch the first earnings report. SpaceX reports its first public quarterly results in August 2026. The compute utilization data and any language about contract renegotiations will tell you more about API pricing trajectories than any analyst prediction today.
The infrastructure layer you build on just went public. That is not inherently bad — public companies have accountability that private ones don’t. But it means the pricing and availability of the AI tools you depend on is now partly a function of Nasdaq performance and quarterly guidance. That is a different kind of risk than you had six months ago, and worth understanding before the first earnings call.
