Mark Zuckerberg signals commercialization of computing power, Meta's infrastructure may enter a "self-use + external leasing" dual-track model.

Mars Financial News, July 1 — According to Bloomberg, Zuckerberg said during a shareholder phone call in May: "Almost every week, different companies come to us, either hoping we’ll build an API service or asking if we have computing power we can sell them at a price higher than what we paid. We haven’t done that yet because we have our own demand for these computing resources," Zuckerberg said at the time. "But obviously, if we realize we’ve overbuilt, then that’s an option, and that partly gives us confidence to invest in scaling up construction."

In the rapidly evolving AI race, Zuckerberg has repeatedly stated that he believes the entire industry faces a bottleneck in computing power, and that Meta should accumulate as much computing power as possible before deciding how to use it. Despite numerous complex challenges, Meta CEO Mark Zuckerberg has signaled to investors that he is willing to sell excess computing infrastructure, and even launch so-called API services, allowing customers to pay based on AI usage—a business typically measured in "tokens" (the amount of data generated and consumed during a customer query).

Discussions around Meta Platforms potentially commercializing some of its AI computing power have led the market to worry about an "oversupply" of computing resources, but analysts believe this judgment is oversimplified. Currently, Meta does not have significant spare computing power available for sale. On one hand, it continues to expand AI infrastructure investments and sign large-scale computing agreements with companies like Crusoe. On the other hand, existing H100/H200 resources are still primarily used for internal recommendation systems and AI model training, with demand remaining tight. Therefore, the so-called "selling computing power" is more of a future optional asset utilization strategy rather than a current reality. Overall, it reflects that AI computing power is transitioning from a single shortage to multi-generational tiered usage and long-term capacity planning.