Tech mega-caps lose $2.7T in June as AI capex concerns mount

The 'Magnificent Seven' plus Broadcom and Oracle shed roughly $2.7 trillion in market capitalization during June 2026, marking a significant repricing of companies funding and powering the AI buildout. The reset spans both sides of the AI complex: hardware vendors (NVIDIA, Broadcom) and spending platforms (Microsoft, Alphabet, Amazon, Meta, Oracle). The pressure reflects growing investor skepticism about the return on hyperscaler capex commitments estimated at $660–690 billion in 2026 alone.

The underlying driver is projections that hyperscaler free cash flow will decline sharply as AI infrastructure spending continues to escalate. Nomura strategists characterized the trend as hyperscalers becoming "the funding shorts" in AI infrastructure—companies simultaneously funding the buildout and providing revenue streams for suppliers, while squeezing the cash cushion that historically supported buybacks, dividends, and M&A. Morgan Stanley estimates AI-linked debt on track to nearly double to $570 billion in 2026, signaling the financing burden is shifting from equity to debt markets.

Market sentiment is further pressured by reports that OpenAI is considering delaying its IPO to next year due to SpaceX's volatility and overall softness in AI-adjacent equities. The Nasdaq Composite posted its fifth consecutive losing session on June 26, with the tech-heavy index down 4.6% for the week and the S&P 500 down nearly 2%, as investors rotate out of mega-cap tech into defensive sectors.

For practitioners: the capex repricing signals a market-wide reckoning on the cost structure of frontier AI. The build-out economics have shifted from "spend to scale" to "spend must justify returns." This tightens funding discipline across the entire infrastructure stack—from chip supply to neocloud capacity. Organizations should stress-test token cost assumptions against a scenario where infrastructure pricing does NOT continue to decline and where capex returns face higher hurdle rates.