Meta stock has fallen about 10% in 2026, with large investors quietly selling shares while possibly buying Google instead, according to latest market data. The shift comes as Meta plans to spend between $125 billion and $145 billion in 2026, most of it on AI data centers, creating significant investor concerns about the company's return on investment. Chaikin Money Flow reads -0.209 for Meta, a sign of steady selling, while Google reads +0.177, a sign of buying. Meta also trails in relative strength, scoring 95.8 and lagging, while Google scores 123.0 and leads. Unlike Google, Amazon, and Microsoft, Meta runs no cloud business to rent out its new computing power, creating uncertainty about how it will monetize its massive AI infrastructure investments.
Meta has officially launched Meta Compute, a new internal division code-named to sell idle AI computing power externally, according to multiple foreign media reports. The initiative mirrors SpaceX's model of renting spare capacity to firms including Anthropic. Meta is exploring two pricing models: one similar to Amazon's AWS Bedrock service, allowing developers to pay for access to AI models deployed on Meta's infrastructure; the other involves directly leasing 'raw compute'—pure GPU computing resources—which would put it in direct competition with emerging cloud service providers like CoreWeave and Nebius. This strategic pivot represents a significant shift from the company's traditional approach to AI infrastructure, potentially reshaping the entire AI computing landscape. According to Bloomberg, citing people familiar with the matter, Meta is developing a brand-new cloud business plan designed to sell computing resources to external customers, directly rivaling the three traditional cloud giants—Amazon's AWS, Microsoft Azure, and Google Cloud. The breaking news confirms that Meta is 'building a cloud business to sell excess AI capacity,' marking the first hyperscaler to admit to having excess compute capacity, with more likely to follow.
The announcement triggered a sharp selloff across AI chip and infrastructure stocks on Wall Street, confirming the AI bubble burst. Meta's stock surged 10% in early trading on Wednesday (July 1), marking its biggest single-day gains since January 29, 2026, while the broader AI computing sector faced significant pressure. As reported by multiple foreign media sources, the announcement flipped years of assumed AI compute scarcity into a supply warning, erasing billions in semiconductor and neocloud value in a single session. The Philadelphia Semiconductor Index fell 6.27% to close at 13,353.28 points, with 28 of its 30 constituents declining. Nebius crashed 14% to $229.18, becoming the S&P 500's biggest loser of the day, while CoreWeave also tumbled 11% to $85.69. The market interpreted Meta's move as a major signal: AI computing power may be shifting from 'perpetual shortage' to 'regional surplus' with implications of a reduced or stalled AI capex, confirming the broken AI business model and pointing to a deeper selloff as the AI bubble continues to burst.
Major semiconductor companies experienced significant losses following the announcement. Advanced Micro Devices Inc (NASDAQ:AMD) closed down 6.9%, Intel Corporation (NASDAQ:INTC) fell 9%, and NVIDIA Corporation (NASDAQ:NVDA) dropped 1.3% as investors work through the implications of reports that AI labs can squeeze dramatically more output from existing chips. TSMC ADR fell 7.01%, Arm Holdings declined 4.82%, Qualcomm down 1.55%, and Nvidia slid 1.25%. Storage stocks also came under pressure, with Western Digital dropping 6.32% and Seagate Technology declining 5.16%. Only Intel showed relative resilience, tumbling just 9.03%. The sell-off quickly spread from emerging cloud stocks to the entire AI hardware ecosystem, with the 'pick-and-shovel' trade that had led the market over the past year facing aggressive profit-taking. Alphabet, Amazon, Meta Platforms and others saw heavy selling in June, while the Philadelphia SE Semiconductor index climbed 11% last month. However, the chip index has fallen over 11% in the last two weeks, while the Roundhill Magnificent Seven ETF - a proxy to track the seven biggest Wall Street tech companies - has recovered some lost ground.
Companies specializing in GPU rental services faced particularly severe declines. CoreWeave and Nebius rent GPU capacity to AI developers and saw their stocks fall 11% and 14% respectively on fears that Meta will undercut their pricing. Meta's shift into the same business now puts it in direct competition with its own vendors, creating additional pressure on established players in the AI infrastructure market. Meta's relationship with CoreWeave is extremely close, with the two parties signing a $21 billion cooperation agreement last September, then expanding their partnership in April with an additional $21 billion contract. Meta also signed a deal worth up to $27 billion with Nebius in March for AI infrastructure services, with these combined $35.2 billion in orders accounting for over one-third of CoreWeave's order backlog.
Meta's AI cloud strategy comes weeks after SpaceX announced a similar move, intensifying competition in the fast-growing AI cloud market. Last month, Anthropic signed a deal with SpaceX to use the full computing capacity of its Colossus 1 AI data center in Memphis, gaining access to more than 300 megawatts of AI compute power. Under the agreement, Anthropic will pay about $1.25 billion per month through May 2029, with the two companies also exploring the development of multi-gigawatt AI computing infrastructure in space. This SpaceX model reinforces Meta's strategic pivot, as Meta CEO Mark Zuckerberg said in May that a Meta cloud computing business was 'definitely on the table'. The move could intensify competition in the AI cloud market, where companies such as CoreWeave have benefited from soaring demand for GPU-powered computing, with retail sentiment surrounding META on Stocktwits turning 'neutral' from 'bearish' a day earlier.
