SpaceX is heavily promoting xAI's ability to build data centers rapidly and at low cost, positioning this as a core advantage over competing AI labs and a key selling point as the company prepares for its Initial Public Offering (IPO).

For instance, a draft of SpaceX's IPO prospectus reveals that xAI deployed the first two compute clusters at its second data center for approximately one-quarter of the industry's typical construction cost.

However, this speed and low upfront cost involve a hidden trade-off: the use of temporary power and cooling solutions introduces reliability risks, which could increase overall operational expenses in the long term.

For SpaceX, ensuring the efficient operation of xAI's data centers is critical. In 2025, the combined revenue of SpaceX and xAI exceeded $18.5 billion, but they reported a combined loss of approximately $5 billion. Notably, xAI's capital expenditure reached nearly $13 billion, which is 50% higher than the investment in SpaceX's satellite and rocket businesses.

xAI achieves rapid data center deployment partly by retrofitting existing structures, such as abandoned appliance factories and warehouses, rather than building from scratch. The company also utilizes temporary power and cooling equipment, avoiding delays associated with connecting to the main power grid and completing permanent infrastructure, which significantly shortens the construction timeline.

The prospectus indicates that the two major data centers in Memphis, Macrohard and another local xAI site, have a combined computing capacity of 700 megawatts, with plans for further expansion, serving as evidence of xAI's speed and scale. SpaceX also led analysts on a site tour of the Memphis campus in late April.

According to the prospectus, xAI brought the first compute cluster at Macrohard online in just 91 days, compared to a standard industry timeline of up to two years. The cost was as low as $2.7 million per megawatt, significantly below the industry benchmark of $12.3 million per megawatt. However, the prospectus does not specify which cost items are included in this calculation.

Sources familiar with the project disclosed that the Macrohard campus currently relies primarily on mobile gas turbines for on-site power, supplemented by Tesla Megapack battery storage. Furthermore, over one hundred mobile cooling units are used to cool the GPU servers.

These temporary solutions result in significantly lower operational stability compared to traditionally built data centers. The sources reported that the temporary infrastructure has already caused multiple campus outages, interrupting the training of new AI models. The site's operational uptime is far below the internal target of at least 99.9%.

In recent weeks, CEO Elon Musk has installed a SpaceX management team to take full control and overhaul xAI's operations, appointing Senior Vice President of Starlink, Michael Nichols, to lead the effort. In an internal memo from April, Nichols stated that xAI must improve the utilization efficiency of the hundreds of thousands of Nvidia GPUs in its data centers. He explicitly noted that the company's Model FLOPs Utilization (MFU) was a disappointingly low 11% and demanded that the team increase utilization to 50% within the coming months. This memo was first reported by Business Insider. Both xAI and SpaceX declined to comment.

**Executive Departures and Safety Incidents** Public records and sources familiar with the project indicate that alongside operational pressures, xAI is facing the departure of several executives and frequent safety incidents during construction, leading to subsequent adjustments in safety protocols for ongoing projects.

At least three data center executives have left the Memphis campus, including Dan Roland and Brent Mayo, who led the construction of the Macrohard and xAI's first data center, Colossus. Jack Palmer, a former executive at crypto mining firm Hut 8 who joined xAI last year, has been promoted to oversee operations at the Memphis data center.

Another source revealed that xAI shelved an $8 million water recycling plant project, originally intended to reduce water consumption at the Colossus campus, after which the project's lead engineer, Mark Carroll, departed. In recent weeks, xAI has also laid off several mid- and junior-level employees, including staff from the water reuse project and data center engineering teams.

The campus has experienced multiple construction safety incidents in recent months. Records from the Tennessee Occupational Safety and Health Administration show that in January, a temporary contractor at Macrohard had his right index finger crushed while operating a heat fusion machine to join pipes, ultimately requiring amputation. The worker was employed by Darana Hybrid, a labor subcontractor working with xAI. Following the accident, the contractor was mandated to implement additional safety measures, including retraining operators and requiring a two-person verification that hands are clear of the danger zone during machine operation. These regulatory documents had not been previously disclosed.

In April, the administration received an anonymous complaint stating that several employees at the Macrohard campus had experienced symptoms such as headaches and breathing difficulties. The complaint pointed to the extensive use of diesel generators by xAI-affiliated company CTC Property LLC for temporary site lighting. In response, xAI's representative, Brianna Hamilton, informed regulators that the company continuously monitors carbon monoxide levels, which were within safe limits, and that building doors and windows were kept open for ventilation. The administration ultimately decided to close the case without further action.

Beyond incidents this year, a serious safety event occurred last September: two employees of xAI's partner, Solaris Energy Infrastructure, suffered an electric shock while working on nearby power facilities. The Memphis Daily News reported that one worker was severely shocked by a high-voltage line and hospitalized for at least 53 days.