The power problem behind AI just got a price tag, and it’s coming from inside Elon Musk’s own empire. SpaceX spent $329 million on Tesla Megapacks so far this year, according to earnings figures reported by TechCrunch AI on Tuesday. Most of that, $295 million, landed in the second quarter alone. What stands out here is where those batteries are almost certainly headed: xAI’s data centers.
Situation Report
- SpaceX bought $329M of Tesla Megapacks year to date, $295M of it in Q2, per SpaceX’s own earnings report cited by TechCrunch AI.
- The likely destination is xAI’s data centers. Earlier this year SpaceX absorbed xAI, which had itself acquired X in 2025.
- Before that merger, xAI had already bought $430M worth of Megapacks. In Q1 it spent just $34M. The jump to a $295M quarter signals a serious ramp.
- Separately, SpaceX reported $131M in Tesla Cybertrucks acquired at MSRP as of December 2025.
Why This Matters
This is a clean look at how tightly Musk’s companies now feed each other. He’s CEO and largest shareholder of SpaceX. He runs Tesla. His AI outfit, xAI, sits under the SpaceX umbrella. So Tesla builds the batteries, SpaceX buys them, and xAI runs the GPUs they support. Money moves in a circle, and the numbers are public because of SpaceX’s regulatory filing.
The bigger signal is what it says about AI infrastructure. Training and inference don’t pull power at a steady rate. Demand spikes hard, then drops, then spikes again depending on what the models are doing. Those peaks are the enemy. They rack up utility charges and can overwhelm on-site generators.
How Batteries Solve It
Megapacks are Tesla’s industrial-scale battery storage units. In a data center they pull double duty:
- Backup power that can kick in under a second if the main supply drops.
- Peak smoothing that feeds extra juice to GPUs the moment they demand it, so the site doesn’t lean on the grid or overload its own turbines.
The result is lower cost and steadier operation. That’s why batteries matter even for a company like xAI that leans heavily on natural gas. TechCrunch AI notes xAI has run dozens of unpermitted gas turbines at a Mississippi site near its Colossus project. Gas keeps the lights on. Batteries keep the power clean and responsive.
What Comes Next
Expect more of this, not less. The AI buildout is running into a wall that has nothing to do with chips and everything to do with electricity. Grid connections take years. Permits take longer. Batteries are one of the few pieces you can buy and deploy fast, and Musk happens to own the factory that makes them.
A few things to watch:
- Vertical integration as a moat. Rivals buying Megapacks pay a competitor. Musk buys from himself. That’s a structural cost edge if the numbers hold.
- Scrutiny on the loop. Related-party spending at this scale invites questions from regulators and investors about pricing and disclosure.
- Power, not compute, as the bottleneck. The companies that lock down energy and storage early will train the biggest models. The ones that don’t will wait in line.
My read: this $329M line item is a preview of where the AI arms race actually gets fought. Not just on GPU orders, but on who can keep them fed. More detail is available in the original TechCrunch AI report.