Mark Zuckerberg Signals Commercialization of Computing Power, Meta's Infrastructure May Adopt a Dual-track Mode of "Self-use + Leasing"

BlockBeats News, July 1st, according to Bloomberg, Mark Zuckerberg said during a shareholder conference call in May, "Almost every week, a different company comes to us, either wanting us to build an API service or asking if we have computing power to sell to them at a higher price than what we bought it for. We haven't done this yet because we have our own need for this computing power," Zuckerberg said at the time. "But obviously, if we realize we've built too much, that is an option, and it partly gives us the confidence to invest in expanding our infrastructure."

In the rapidly evolving AI competition, Zuckerberg has repeatedly stated that he believes the entire industry is facing a bottleneck in computing power, and Meta should accumulate as much computing power as possible before deciding how to use it. Despite facing many complex challenges, Meta CEO Mark Zuckerberg has sent a signal to investors: he is willing to sell excess computing infrastructure, and even launch the so-called API service, allowing customers to pay based on AI usage -- a business that is usually measured in "tokens" (the amount of data generated and consumed by customer queries).

In the discussion surrounding the potential commercialization of part of Meta Platforms' AI computing power, the market was once concerned about a possible computing power "surplus." However, analysts believe this assessment is oversimplified; Meta does not currently have a significant surplus of computing power available for sale overall. On the one hand, it is still continuously expanding its investment in AI infrastructure and signing large-scale computing agreements with companies such as Crusoe. On the other hand, the existing H100/H200 resources are still mainly used for internal recommendation systems and AI model training, and the demand remains high. Therefore, the so-called "selling of computing power" is more of a future optional asset utilization method rather than a current reality, reflecting more the transition of AI computing power from a single scarcity to multi-generational layered use and long-term capacity planning.