#Meta卖算力引发存储股大跌 On July 2, 2026, Meta announced plans to sell idle AI computing power externally, sparking market concerns about oversupply of computing power and causing a sharp decline in global AI tech stocks. On the same day, U.S. stock memory concept stocks SanDisk and Micron Technology fell over 10%, South Korea's SK Hynix and Samsung Electronics fell 14.57% and 9.06% respectively, and A-share semiconductor and computing hardware sectors generally fell over 5%.
1. Event Background
Meta plans to launch a cloud infrastructure business, intending to rent or sell part of its AI computing power externally, involving two business lines: first, providing self-developed large model invocation services; second, directly selling computing power resources. The company's 2026 capital expenditure is expected to reach $125 billion to $145 billion, double that of 2025, mainly invested in AI infrastructure.
2. Market Reaction
U.S. stocks: The Philadelphia Semiconductor Index fell 6% in a single day, memory stocks like Micron and SanDisk fell over 10%, Meta's stock briefly rose over 10% that day, finally closing up 8.81%.
Asian markets: South Korea's KOSPI index triggered a circuit breaker, the Nikkei 225 fell 2.47%, and A-share semiconductor equipment ETFs saw net capital inflows of over 4.7 billion yuan in a single day, showing capital divergence.
3. Controversial Focus
The market worries that Meta, as a leading computing power buyer turning into a supplier, may signal overheating in AI infrastructure investment, shaking the valuation logic that "computing power will be in short supply for a long time."
Some institutions believe that Meta's move is actually to revitalize idle resources (current utilization rate about 65%) and convert capital expenditure into revenue, rather than a turning point for industry demand.
4. Industry Views
Pessimists believe that AI application monetization is below expectations, and the high valuations of the hardware sector are under pressure;
Optimists point out that North American cloud vendors have backlogs 14 times their quarterly revenue, the computing power gap still exists, and after short-term adjustments, leading stocks may present structural opportunities.
1. Event Background
Meta plans to launch a cloud infrastructure business, intending to rent or sell part of its AI computing power externally, involving two business lines: first, providing self-developed large model invocation services; second, directly selling computing power resources. The company's 2026 capital expenditure is expected to reach $125 billion to $145 billion, double that of 2025, mainly invested in AI infrastructure.
2. Market Reaction
U.S. stocks: The Philadelphia Semiconductor Index fell 6% in a single day, memory stocks like Micron and SanDisk fell over 10%, Meta's stock briefly rose over 10% that day, finally closing up 8.81%.
Asian markets: South Korea's KOSPI index triggered a circuit breaker, the Nikkei 225 fell 2.47%, and A-share semiconductor equipment ETFs saw net capital inflows of over 4.7 billion yuan in a single day, showing capital divergence.
3. Controversial Focus
The market worries that Meta, as a leading computing power buyer turning into a supplier, may signal overheating in AI infrastructure investment, shaking the valuation logic that "computing power will be in short supply for a long time."
Some institutions believe that Meta's move is actually to revitalize idle resources (current utilization rate about 65%) and convert capital expenditure into revenue, rather than a turning point for industry demand.
4. Industry Views
Pessimists believe that AI application monetization is below expectations, and the high valuations of the hardware sector are under pressure;
Optimists point out that North American cloud vendors have backlogs 14 times their quarterly revenue, the computing power gap still exists, and after short-term adjustments, leading stocks may present structural opportunities.
