This Is SpaceX's Biggest Risk, and It's Not What You Think

All stocks carry risk, and Space Exploration Technologies (NASDAQ: SPCX) is no exception. However, its biggest risk is not what you likely think. It's not its high valuation, impending share lock-ups, or even failure to reach some of its more ambitious goals, like orbital data centers, within CEO Elon Musk's stated time frames. Instead, its biggest risk is actually its current artificial intelligence (AI) infrastructure build-out.

Terrestrial AI infrastructure is currently SpaceX's biggest growth driver. The company is renting out excess compute capacity every month and getting some outstanding rates. This is helping it achieve some of the best economics in the space, with payback on its chip and networking investments in about a year.

Over $500 billion in 2027 capex

Given the robust returns it is seeing on its investments, SpaceX is looking to put the pedal to the metal and greatly ramp up its AI infrastructure spending. During SpaceX's first-ever earnings call following its Q2 results, Musk said SpaceX would have around 2 gigawatts of compute power by the end of this year and closer to 10 gigawatts than 5 gigawatts by the end of 2027. That was an aggressive but reasonable pace.

However, a quarter later on SpaceX's Q3 earnings call, Musk decided to take this into hyperdrive. The company will now have up to 3 gigawatts of compute power before year-end 2026, with plans to reach 13 gigawatts by the end of 2027. Meanwhile, he said its moonshot goal was to add 20 gigawatts of compute power next year.

SpaceX has decided to exclusively build its data centers on Nvidia's next-generation platform. At its developer conference in October in Berlin, Nvidia CEO Jensen Huang said that preassembled AI data centers built on its Vera Rubin platform cost around $50 billion per gigawatt. Using that figure, SpaceX appears to be looking to spend $500 billion on AI infrastructure capex next year to add 10 GW of compute power.

That's just a staggering number, much greater than what the big three cloud computing companies spend. For example, Amazon plans to spend up to $220 billion on capital expenditures this year, with Alphabet at up to $205 billion and Microsoft at around $175 billion. While those numbers are projected to rise meaningfully next year, they would still be nowhere near what SpaceX plans to spend.

Meanwhile, the big cloud providers also generate robust operating cash flow that can cover the bulk of their AI infrastructure spending, while SpaceX's operating cash flow is minimal by comparison. That means it's going to have to raise substantial debt and perhaps issue equity to fund its spending.

Image source: The Motley Fool.

High debt and a spending treadmill

While this does add risk, piling on debt isn't the sole issue. SpaceX is getting very strong returns on its AI infrastructure because we are currently in a compute-starved world. However, with much more capacity set to come online next year, it could significantly weaken the economics for a company renting out capacity on short-term leases. That's just simple supply and demand.

Meanwhile, given the short five- to six-year economic lifespan of graphics processing units (GPUs), it essentially puts cloud computing companies on a permanent spending treadmill. That's fine when end-user demand and token monetization are climbing, but when incremental revenue growth starts to plateau, the continual massive hardware upgrade cycle will pressure free cash flow and squeeze margins.

Adding a massive foundry in Terafab would only exacerbate this. Foundries need to run near full capacity to be profitable, and if they don't, it can crush the business. You only have to look at Intel and its money-losing foundry business to see this. Because of the high up-front costs, required technological expertise, and capacity utilization risks, most semiconductor companies today operate a fabless model and outsource manufacturing to a company like Taiwan Semiconductor Manufacturing.

While none of this needs to play out, and AI infrastructure will be a huge revenue driver for SpaceX, it is a risk that investors should be aware of before investing in the stock.

Geoffrey Seiler has positions in Alphabet and Amazon. The Motley Fool has positions in and recommends Alphabet, Amazon, Intel, Microsoft, Nvidia, and Taiwan Semiconductor Manufacturing. The Motley Fool has a disclosure policy.