SpaceX AI Data Center Shake-Up: Leadership Out, Power Crisis Looms

SpaceX is overhauling its AI data center operation after a string of reliability failures, while Elon Musk used a G20 stage to argue that a power crunch will define the AI race in 2026 and beyond.

Key points

• SpaceX replaced several data center leaders with executives from its rocket and satellite internet divisions, following operational problems at facilities in Tennessee and Mississippi.
• During the company’s rapid infrastructure buildout, clusters ran for extended periods without backup cooling or power systems, materially raising the risk of outages, according to sources cited by The Information via Digitimes.
• Musk told G20 technology ministers in Chapel Hill on September 2 that there will be “a significant power shortfall next year,” warning that the energy crunch is handing China a competitive advantage in AI.
• He also said SpaceX is building its own power generation capacity to accompany its data centers, suggesting the company intends to address the energy problem in-house rather than waiting on the grid.
• SPCX carries a price-to-sales valuation that GuruFocus describes as “sky-high”, making execution risk in the data center segment a live investor concern.

What went wrong at the Tennessee and Mississippi sites?

The core problem, according to reporting cited by Digitimes, was speed over resilience. SpaceX pushed to stand up AI computing clusters quickly, and in doing so operated them without backup cooling or power redundancy for extended stretches. That is an unusual posture for data center infrastructure, where uptime guarantees typically require redundant systems from day one. The consequence was elevated outage risk, and apparently enough operational pain to trigger a leadership reshuffle.

The incoming executives come from SpaceX’s rocket and satellite internet divisions, per GuruFocus. That is a notable choice. Rocket and Starlink operations are both defined by hard real-time reliability constraints, which may be precisely the culture Musk wants applied to the data center buildout. Whether that cross-domain transplant works in civil engineering and power infrastructure is an open question, and the sources do not yet offer a verdict.

What did Musk argue at G20, and what does it mean for SpaceX?

Speaking virtually at the G20 technology ministers’ meeting in Chapel Hill, Musk framed the AI infrastructure problem in stark terms: “There will be a significant power shortfall next year, not the distant future,” he said. He tied that directly to geopolitical competition, arguing the energy gap gives China an advantage.

Musk also stated that SpaceX is building the power generation required to support its own data centers. No financial figures or project timelines were provided in the available sources, so the scale of that commitment remains unclear. Still, the statement signals that SpaceX views vertical integration in energy as part of its AI infrastructure strategy, not just a temporary workaround.

Mark Zuckerberg, speaking in the same session, emphasized a different bottleneck: a shortage of skilled tradespeople to physically build out data centers at the pace demand requires. The two arguments together sketch a picture of AI infrastructure constrained from multiple directions at once, which may explain why SpaceX’s own buildout cut corners on redundancy earlier in the process.

Musk used the G20 platform to criticize European AI regulation, arguing that innovation requires things to be “default legal as opposed to default illegal.” The remarks drew an audience that included technology ministers from France, Germany, and Italy, as well as the EU’s tech chief.

How does the valuation picture interact with operational risk?

SPCX’s price-to-sales multiple is, by GuruFocus’s characterization, elevated relative to peers. That kind of valuation is typically supported by expectations of high-margin, scalable revenue streams. AI data center services could theoretically fit that description, but the Tennessee and Mississippi problems suggest the business is still in a loss-absorbing buildout phase, not a steady-state operating one.

Leadership reshuffles at a critical growth division introduce their own uncertainty. New executives need time to assess what they have inherited, restructure workflows, and rebuild any damaged client or partner confidence. The sources do not indicate whether any customers experienced outages or whether commercial agreements were affected by the reliability issues. That gap in the available reporting is worth monitoring.

For investors, the core tension is straightforward: a sky-high multiple requires the AI data center bet to pay off at scale, and the events of the past week are a reminder that execution in this segment is harder than the headline ambition suggests.