Elon Musk: xAI targets 7x data center growth and $500B revenue

The Billion-Dollar Blueprint: How SpaceX is Building the AI Supercomputer of Tomorrow

In the high-stakes world of artificial intelligence infrastructure, some predictions sound like science fiction, yet they are being hammered out in data centers across the globe. One such ambitious forecast comes from Elon Musk, who has laid out a staggering vision for the computational power that SpaceX‘s xAI division is poised to unleash.

Musk estimates that by late 2027, the power capacity of the xAI data centers will explode by seven times, reaching 10 gigawatts (GW). If this monumental infrastructure goal is met, Musk suggests that these facilities could generate between 300 billion to 500 billion dollars in annual revenue for the company. This kind of projection brings immediate attention, especially as the market capitalization of SpaceX has experienced significant volatility in recent months.

The ambition isn’t just about raw electricity; it’s about computational density. While the nameplate power draw for current xAI centers in Memphis and Southaven is 1.4 GW, the true measure of capability lies not in Watts but in floating-point operations per second (FLOPS). To understand the scale, we must look beyond the simple power figures.

When analyzing data center performance, it’s crucial to account for efficiency. A typical facility with a Power Usage Effectiveness (PUE) of around 1.2 means that much of the energy goes toward operational systems like cooling and lighting. This often leaves the actual compute capacity—the accelerators processing the AI workloads—as the key metric.

Experts suggest that for every gigawatt consumed by facilities, the resulting accelerator power dedicated to AI processing could be substantial. Assuming 70% to 80% of total IT power translates into computational accelerators, this suggests that a massive data center can house nearly a gigawatt of processing power.

Looking ahead to 2027, based on the projection of 10 GW of compute capacity running on advanced systems, the potential performance leap is staggering. If xAI deploys these systems using next-generation hardware, such as Nvidia’s Rubin-based accelerators, the cluster could achieve performance levels far surpassing current public benchmarks.

Estimates suggest that this combined capability would place the xAI cluster in a league entirely separate from current supercomputers on the Top 500 list. Potential training and inference speeds are projected to reach hundreds of ExaFLOPS, signaling the ability to train models of unprecedented complexity.

This level of computational power opens the door for dramatically more sophisticated AI models than currently possible. While achieving those multi-hundred-billion-dollar revenue targets will depend on market penetration and fierce competition from other tech giants, the scale of infrastructure being built by SpaceX underscores a massive shift in how the world approaches artificial intelligence.