Opinion
Elon Musk just dropped a $1.5 trillion reality check on the AI race
"Musk is assembling the railroad, the network and the factories of the AI economy in one stack. If you care about AI strategy, this is a reality check on the physical limits of the AI race, and who ends up owning the infrastructure those limits depend on," writes Renana Ashkenazi, managing partner at Grove Ventures.
Last week it was reported that Elon Musk is targeting a $1.5 trillion valuation for SpaceX, not only making it the largest IPO in history but effectively pricing the company higher than most G7 economies. At almost the same moment, he moved to merge xAI into SpaceX, creating the most valuable private company on Earth and putting rockets, satellites and AI under one roof.
This isn't a sci-fi space company. While most headlines focus on rockets and Mars, the smart money is pricing something else entirely. Musk is assembling the railroad, the network and the factories of the AI economy in one stack. If you care about AI strategy, this is a reality check on the physical limits of the AI race, and who ends up owning the infrastructure those limits depend on.
$1.5 trillion signals AI has hit a terrestrial cap
On paper, this looks like a space play: SpaceX, already valued in the high hundreds of billions, merging with xAI at a combined valuation around $1.25 trillion. In practice, it's a thesis about where AI hits the wall. Capital markets don't put a trillion-plus price tag on a space dream; they do it when they think they're looking at future infrastructure.
For the last two years, most AI conversations have been about models: which frontier model to pick, how to plug in agents, how to control hallucinations. All those still matter, but the constraint now driving real capital allocation is more basic: power, cooling, land and local politics.
You can see it where AI already lives at scale. Data center hubs are colliding with grid limits and community pushback: Even if you can get GPUs, you can't always get the megawatts, permits and neighbors to go with them. At some point, Earth itself starts looking like a cap on how far you can push the current approach. A trillion-dollar bet on a space-AI utility means the bottleneck is no longer intelligence. It's physics.
Physics arbitrage: Why Musk is betting against Earth
At its core, folding xAI into SpaceX is a physics arbitrage play. Data centers on Earth have two expensive problems: they need enormous power, and they generate enormous heat. Both get worse as models grow larger and industries move from pilot to production.
Orbit changes that equation. In space, you get 24/7 solar exposure instead of cloudy grids and permitting fights. You also get the ultimate free chiller: the cold vacuum around you. 'Floating data centers' sounds wild until you realize Musk is betting orbital energy and cooling will beat fighting over land, power and permits on Earth.
From space company to full-stack utility
To see why this matters for everyone else in AI, shift metaphors - from rockets to railroads. In the 19th century, big money wasn't in finding gold but in owning the tracks that moved everyone there.
Today, SpaceX is the only functioning railroad to orbit at meaningful scale. If you want to launch serious hardware into space, you pay Musk's toll. That's logistics dominance. On top of that railroad sits Starlink, the emerging global connectivity grid. It sells bandwidth where terrestrial infrastructure is fragile, expensive or politicized, where it becomes the backbone of information flow.
Now fold in xAI. Musk isn't just launching other people's payloads; he's running 'AI factories' at the end of his own railroad. He owns transport (rockets), network (Starlink) and intelligence (xAI): vertical integration most AI competitors will never have.
Warning shot to competitors: While others fight for land, nuclear deals and grid access on the ground, Musk is looking up, trying to own the alternative route entirely. If that stack proves viable, every purely terrestrial data center suddenly looks constrained by design.
One man's infrastructure: The geopolitical reality check
There's another lesson here: concentration of power. When terrestrial networks failed in Ukraine, Starlink became a lifeline. Connectivity, targeting, coordination - all ran through a private system. When Ukraine planned to strike Russia's Crimea fleet, one CEO's choice about network coverage shaped battlefield reality.
Extend that from connectivity to compute. As AI becomes mission-critical for healthcare, logistics, defense and finance, 'who controls the infrastructure?' stops being academic and becomes a sovereignty issue. This $1.5 trillion reality check isn't about needing new pipes, but that those pipes may end up more centralized and privately governed than many nation-state systems.
What operators and investors should do next
You don't need your own rockets to respond. Update how you think about AI infrastructure instead.
- Treat infrastructure as board-level risk, not IT detail. Map where critical AI workloads run, which providers they depend on, how concentrated that exposure is. Ask what happens if energy prices spike, a key region hits permitting walls, or one provider changes terms.
- Design for infrastructure volatility. Build hybrid architectures that fail over between on-prem, multiple clouds and connectivity layers (fiber, 5G, satellite) without breaking. In real deployments, resilience and latency beat another benchmark point from your model.
- Hunt value in the 'boring' layers. Durable moats will emerge in grid optimization, cooling, workload routing to cheapest/cleanest energy, traffic compression and cross-environment orchestration. These sit where physics, regulation and capital collide.
Musk's move is extreme by design. But the message is simple: AI has outgrown the fantasy it's just software in the cloud. The race now belongs to whoever masters the infrastructure - and physics - that make it possible.
Renana Ashkenazi is a managing partner at Grove Ventures.
