SpaceX has put Starlink senior executive Michael Nichols in charge of AI data centre operations. The company is seen as moving to stabilise operations as key staff keep leaving the data centre unit and some facility outages are blamed for disrupting AI model training.

Cryptopolitan, a blockchain media outlet, reported on Monday local time that Nichols took over as head of xAI infrastructure after several senior data centre executives left the company in recent weeks. SpaceX must demonstrate stable data centre operations as it has secured major compute customers including Google and Anthropic.

SpaceX's data centre business has highlighted fast buildouts and low costs as strengths. The company said it operated its first two computing clusters at its second data centre at about a quarter of typical industry costs, but did not disclose what costs were included.

A problem has been heavy reliance on temporary equipment during rapid expansion. xAI's Macrohard site began operations using mobile gas turbines, Tesla Megapack batteries and more than 100 portable cooling units. A person involved in the project claimed outages from such temporary equipment interfered with AI model training. The facility's uptime was also reported to have consistently fallen short of the company's internal target of 99.9%.

Some Tennessee and Mississippi facilities were reported to have operated for months without backup cooling or backup power as they rushed to expand computing capacity. SpaceX later slowed plans for new data centres outside the Memphis hub, and the schedule for a Texas candidate site has also been delayed.

These setbacks could also weigh on SpaceX's goal of securing more than 2 gigawatts of computing capacity by the end of 2026. Online capacity disclosed by the company in June was about 1.4 gigawatts, meaning further expansion is needed to meet the target.

SpaceX is also reshuffling staff internally to fill vacancies. After infrastructure chief Jake Palmer left the company in July, SpaceX moved personnel from its rocket and Starlink organisations to the data centre unit. Wesley Salandro, a senior production manager on the Falcon and Dragon rocket lines, and Logan McConnell, head of product support at Texas Starbase, joined the data centre team, while the company is also recruiting engineers from Neuralink and hiring civil engineers.

Data centre operating stability is also directly tied to revenue. SpaceX's AI revenue rose to $2.6 billion in the second quarter of 2026 from $737 million a year earlier. Google was reported to have signed a $920 million-a-month compute contract covering about 110,000 Nvidia GPUs, and Anthropic was also reported to be leasing SpaceX's 'Colossus' system for up to $1.25 billion a month.

With major customers secured, continued data centre outages could also directly affect revenue. SpaceX may have to spend more to fix problems or cut the amount of computing it can provide to paying customers.

Elon Musk separately said he is building a foundry in Bastrop, Texas, to produce its own gas turbine blades and vanes. Musk claimed this could bring forward the timing of natural gas turbine deployments by up to 18 months. The NAACP, meanwhile, criticised SpaceX, saying it operated turbines at the Memphis site without proper permits.

Ultimately, SpaceX's AI data centre business now faces proving stable operations rather than simply the speed of expansion as its key task. As the AI business grows rapidly on contracts with major customers including Google and Anthropic, how quickly Nichols can address staffing gaps and facility outages is expected to be an important variable for future expansion of computing infrastructure.