SpaceX Goes Public at $2 Trillion While Quietly Becoming AI's Biggest Landlord

The biggest IPO of the decade just landed, and it came with a plot twist nobody predicted. SpaceX opened at $150 per share, giving the company a market capitalization above $2 trillion and making Elon Musk the world's first trillionaire. But buried beneath the IPO frenzy is a story that matters far more for the AI industry: SpaceX has quietly become one of the largest AI compute providers on Earth.

The Colossus Problem

SpaceX built Colossus 1, its massive AI training cluster in Memphis, Tennessee, to train Grok AI models. The plan was ambitious: connect three data center campuses into one unified training grid. It did not work.

According to Bloomberg reporting, SpaceX hit severe latency issues when trying to connect Colossus 1 with two other sites located more than 10 miles apart. Aging network infrastructure compounded the problem. Training frontier AI models requires extraordinarily low-latency connections between GPUs, and the Memphis setup simply could not deliver.

Rather than let the compute capacity sit idle, SpaceX started renting it out. The deals are staggering.

The Numbers Behind the Compute Deals

Anthropic signed a deal worth approximately $15 billion annually for compute capacity on Colossus. Google followed with a contract worth roughly $920 million per month. That is over $11 billion per year from Google alone.

To put these numbers in perspective: SpaceX's total revenue from launches and satellite services was estimated at roughly $14 billion in 2025. The AI compute rental business could surpass that within its first full year of operation.

This tells us something important about the AI industry in 2026. Compute scarcity is not a future problem. It is a right-now problem, and companies are willing to pay almost any price to secure it.

Why Not Just Fix the Network?

You might wonder why SpaceX does not simply upgrade its network infrastructure and use Colossus for its own Grok training. The answer is a mix of physics, economics, and timing.

Connecting data center campuses separated by more than 10 miles at the speeds required for distributed AI training is genuinely hard. The round-trip latency of fiber optic connections over those distances, combined with switching equipment overhead, creates bottlenecks that current networking technology struggles to overcome. NVIDIA's InfiniBand and similar technologies work brilliantly within a single campus but degrade quickly over inter-campus distances.

Upgrading the infrastructure would require laying dedicated dark fiber routes, installing specialized optical networking equipment, and potentially acquiring right-of-way permits across Memphis. That is a multi-year project, not a quick fix.

Meanwhile, the compute rental deals generate revenue immediately. For a company that just went public and faces intense scrutiny from public market investors, that revenue stream is enormously valuable.

What This Means for AI Competition

The fact that Anthropic and Google are paying SpaceX for compute tells us that traditional cloud providers cannot keep up with demand. AWS, Google Cloud, and Microsoft Azure have all aggressively expanded their GPU fleets, but it has not been enough.

This creates an interesting dynamic. SpaceX is effectively subsidizing Grok's competitors by renting capacity to companies building rival AI systems. But from SpaceX's perspective, Grok is an xAI product, not a SpaceX product. The data center business is pure infrastructure revenue.

It also suggests that the next wave of AI infrastructure may not come from traditional tech companies at all. If a aerospace company can become one of the most important compute providers in the world, the market is wide open for unconventional players.

The IPO Angle

SpaceX's decision to go public now, while its compute business is generating massive revenue, is well-timed. Public investors are desperate for AI exposure that is not just another chip stock or cloud provider. SpaceX offers a unique value proposition: launch services, satellite internet via Starlink, and AI compute infrastructure all in one company.

The $2 trillion valuation reflects all three businesses. But the AI compute angle may be the most underappreciated by the market. If the Colossus deals are sustainable and expandable, SpaceX's infrastructure revenue could eventually rival its launch and connectivity businesses.

What Comes Next

Several questions remain unanswered. Will SpaceX build more data center capacity specifically for rental? Will the company eventually solve its inter-campus networking issues and pull Colossus back for Grok training? How will AWS and Microsoft respond to losing compute customers to an aerospace company?

The AI industry has been characterized by unexpected developments since ChatGPT launched in late 2022. SpaceX becoming a compute landlord is among the strangest. But strange is the new normal in AI, and the companies that adapt fastest to bizarre market conditions will win.

For now, SpaceX sits at the center of the AI infrastructure conversation, collecting billions from rivals while its own AI ambitions wait for better networking. It is a uniquely Muskian business strategy: turn every technical failure into a revenue stream.

One thing is certain. The AI compute market is far more complex and interconnected than most people realize. Your Claude query might run on GPUs in a Memphis data center built by a rocket company. The future is weird, and it is already here.