Meta Plans to Sell Excess AI Computing Power, Triggering Global Chip Stock Selloff

Bloomberg revealed on July 1 that Meta is preparing to launch Meta Compute, a business that will sell external customers access to the surplus computing power of its AI data centers as well as rights to use its own models, including the closed Muse Spark model. The market reaction was immediate and severe: Meta's stock jumped more than 10 percent, while the Philadelphia Semiconductor Index plunged more than 10 percent over the following two sessions.

From Building to Renting

Meta had previously announced AI infrastructure spending of up to $182.9 billion, building among other things a massive computing campus in Ohio, compared in size to Manhattan, along with additional facilities in Louisiana and Texas. Meta Compute is set to be led by Santosh Janardhan, the company's current head of infrastructure, and Daniel Gross, head of Meta Superintelligence Labs, with Dina Powell McCormick serving as president of the new unit.

The model resembles SpaceX's strategy, which has long sold excess satellite bandwidth to outside customers rather than leaving it unused. For Meta, this marks a shift from being purely an investor in AI infrastructure to becoming a direct competitor to Amazon Web Services, Google Cloud and Microsoft Azure in the cloud computing market.

Panic in the Chip Market

News of Meta Compute hit hardest not at Meta's cloud rivals but at memory and processor makers. Micron Technology lost 10.57 percent in a single day, SanDisk fell 10.62 percent, and Intel and AMD dropped 9.03 and 6.89 percent respectively. The wave reached Asia too: on July 2, Samsung Electronics shares fell 9.06 percent and SK Hynix dropped 14.57 percent, enough to rattle the entire Korean stock index.

The reason for investors' panic is simple: if Meta, one of the world's largest buyers of AI chips, is admitting to having excess computing capacity, then the market narrative of a perpetual computing power shortage, on which memory and processor makers' valuations were built, starts to crumble. Analysts quoted by financial media are speaking openly of the end of the invest-at-any-cost phase and the beginning of an era focused on return on invested capital.

What This Means for the Industry

For companies planning investments in AI infrastructure, including in Poland, the signal is clear: the market is beginning to test the assumption that demand for computing power will grow without limit. If even Meta has surpluses, plans by smaller players to build additional data centers may need revisiting, and cloud computing rental prices could start falling in the coming months.

For Polish companies using cloud services, this could mean cheaper access to AI computing power in the medium term, though in the short term the market remains volatile, investors are still weighing whether Meta's move signals market maturity or the first sign of a bursting investment bubble around AI infrastructure.

Meta has not yet officially confirmed all details of the plan, relying so far on leaks and Bloomberg's reporting, but the sheer scale of the market reaction shows just how sensitive the semiconductor sector has become to any signal of a shift in chip-buying pace by major AI players.

Sources: Meta pops 9% as company makes cloud push to sell excess AI compute power capacity (cnbc.com), Meta, like SpaceX, looks to turn excess AI compute into cash (techcrunch.com), Meta Compute Launch Sends AI Compute Stocks Tumbling Globally (finance.yahoo.com)