Summary of Key Points
Meta has announced the sale of its "excess" AI computing power, challenging the market consensus that AI computing power will always be in short supply. This move has caused a collective slump in the AI hardware sectors of both the U.S. and Chinese stock markets (the Philadelphia Semiconductor Index in the U.S. fell by 6.27%, and the Sci-Tech 50 index in China fell by 7.7%). However, Meta's stock price actually rose by 8.81% – as the market believes it has found a way to hedge against the risks of the AI arms race. Behind this is a "prisoner's dilemma" faced by the four major tech giants (Meta, Microsoft, Google, and Amazon): despite the severe imbalance between the cost and output of their spending on building computing power, none of them dare to stop. Meta, lacking its own cloud business to offset these costs, was the first to "defect" by offering its computing power for rent, sounding the alarm for the AI bubble. However, a complete collapse is still too early to predict.
I. Meta Selling Computing Power: A Struggle for Survival in the AI Arms Race
Why did Meta suddenly decide to sell its idle computing power? The main reason is fear of losses.
Meta is different from Microsoft, Google, and Amazon, which have their own cloud services (Azure/GCP/AWS) and can recoup the costs of building computing power through these services. In contrast, Meta's previous investments in computing power were purely for internal use in its advertising and AI applications. If future demand for AI declines, this excess capacity could become a burden.
Mark Zuckerberg had previously hinted that companies were willing to pay a premium for Meta's computing power; now, by officially launching this service, Meta is creating a backup option for itself – should there be an overcapacity, it can still monetize it. This suggests that Meta's confidence in its all-in-on-AI strategy has wavered.
II. The Market Panics: AI Hardware Sectors Take a Dive
As soon as Meta's announcement was made, the market reacted immediately:
- U.S. Stock Market: The Philadelphia Semiconductor Index fell by 6.27%, with all 28/30 components declining. Stocks of Micron and Samsung fell by more than 10%, while Intel, AMD, and TSMC fell by 6%-9%. Pure computing power rental companies fared even worse, with Nebius dropping by 17% and CoreWeave by 13.9%.
- Chinese Stock Market: AI hardware sectors such as chips, optical modules, and PCBs all experienced declines, with the Sci-Tech 50 index falling by 7.7%, and the Sci-Tech Semiconductor ETF leading the drop by 11.25%.
Why is there such panic? The core rationale that has driven the AI sector's growth for two years – namely, the perpetual scarcity of computing power – has been shattered. If the giants start selling excess capacity, companies that previously speculated on the scarcity of computing power will lose their value.
III. The Tech Giants in a Prisoner's Dilemma: Burning Money but Unable to Stop
The four tech giants are caught in an awkward cycle:
Their combined capital expenditures for 2026 (on buying GPUs and building data centers) amount to $725 billion, a 77% increase from 2025, exceeding the GDP of many countries. However, the additional revenue generated by AI is only in the billions of dollars, creating a severe imbalance between input and output.
No one dares to stop first – it's like a prisoner's dilemma: if I stop, my competitor can buy more GPUs and develop stronger models, giving them an advantage. Meta was the first to "test the waters" by selling its computing power, possibly due to its lack of a cloud business to absorb the losses.
IV. Could the AI Bubble Burst? The Possible Order of Collapse
Meta's move is a warning sign, but a full collapse is still uncertain. If the bubble does burst, it might happen in the following order:
1. Computing Power Rental Companies: Those that rely solely on renting out computing power (like CoreWeave) will be the most vulnerable. Their profits depend on the scarcity of computing power, which would lead to a sharp drop in rent prices if the giants start selling excess capacity.
2. GPU Brokers: Companies that profit from reselling GPUs or trading computing power quotas will see their arbitrage opportunities disappear once supply and demand balance out.
3. Chip Manufacturers: Key players like NVIDIA and AMD will be affected, but long-term demand for AI remains; however, the growth rate will slow down from its frenzied pace.
4. Cloud Giants: Microsoft, Google, and Amazon have software services as buffers, giving them more resilience than pure hardware companies.
There will certainly be short-term adjustments, but a collapse depends on whether the giants actually cut back on their spending. For now, Meta is just exploring alternative strategies; it's not yet time for a collective reduction in investment.
V. Why Did Meta's Stock Price Rise? The Market Recognizes Its "Smart Move"
While other companies fell, Meta's stock price rose because the market saw this as a wise move:
- Unlike the other giants, Meta did not stubbornly continue to fund the arms race but instead turned its idle capacity into revenue.
- This reassured investors that even if AI demand falls short of expectations, Meta can still generate income by renting out its computing power, reducing its risk exposure.
Therefore, Meta's stock price rise reflects the market's approval of its strategy of finding a way out of the situation.
Conclusion
Meta's decision to sell computing power is essentially a sign of relaxation in the AI arms race. It has shattered the myth that computing power will always be scarce, causing short-term panic in the hardware sectors. However, it also serves as a reminder to other giants that their spending on AI cannot continue indefinitely. In the short term, AI-related stocks may experience adjustments, but long-term demand for AI remains strong – after all, AI is a future trend that has been overhyped and now needs to return to more rational levels. Whether the bubble will burst depends on whether the giants truly reduce their capital expenditures in the coming days.