Key Highlights

  • An industry research firm reportedly cut its forecast for global AI accelerator spending in 2028 by 28%, to about USD 758 billion, in a model shared with subscribers over the weekend of November 15, 2026.
  • The 2027 forecast was reportedly cut 19% to USD 627 billion, 2029 by 41% to USD 796 billion and 2030 by about 40% to USD 923 billion, a combined reduction of about USD 1.6 trillion.
  • The figures circulated through summaries of a subscriber-only model and have not been publicly confirmed by the firm.
  • VanEck Semiconductor ETF (NASDAQ:SMH) fell 5.11% on Monday, November 16, as the report and new US export rules weighed on chip stocks worldwide.

What the Revised Forecast Says

Accelerator spending is the money that cloud providers, AI labs and other buyers spend on the processors that train and run large AI models. The revised model reportedly leaves the 2026 estimate of about USD 392 billion largely unchanged but sharply lowers the path from 2027 onward. Growth in 2027 is still put at about 60%, but expected growth slows to about 21% in 2028, 5% in 2029 and 16% in 2030, against 37% and 28% previously for 2028 and 2029. Even after the cut, 2030 spending is projected at more than double the 2026 level, so the revision describes slower growth rather than a contraction. A separate estimate of total accelerator silicon area shipping between 2026 and 2030 was reportedly cut by 17%, with the Market Leader accounting for about two-thirds of that reduction. Because the model is sold only to institutional subscribers, the numbers reached the market secondhand, which helps explain why the reaction was abrupt rather than gradual.

Why the Firm Reportedly Changed Its View

Three reasons were cited in the summaries. The first and most important is slower Revenue growth at frontier AI labs, where net new annualised Recurring Revenue reportedly decelerated in the third quarter of 2026. The second is tighter financing for data centers, particularly projects without long-term offtake agreements from large cloud providers. The third is a set of delays to US Data Center capacity, put at about 5.8 gigawatts, linked to regulatory and political changes after the midterm elections, which was described as a secondary headwind. Together these point to a market where Demand is still rising but where funding and power, not chip supply, increasingly set the pace. The forecast also contrasts with other industry trackers that put 2028 Capital spending far higher, which shows how wide the range of views on AI infrastructure has become. For chip investors, the important change is not the absolute level but the shape of the curve.

What It Means for Chip Stocks

The market reaction was sharp because the most valuable chip companies are priced on years of strong growth. On Monday, Micron fell 7.6%, AMD 5.6%, Nvidia 4.1% and Broadcom 3.9%, while the iShares Semiconductor ETF lost 5.37% and a three times leveraged chip fund dropped more than 16%. Asian chipmakers had already fallen hard, with Taiwan's main index down 6.03%. The real test comes on Tuesday, when Nvidia reports. Its outlook for the January quarter and any comments on 2027 orders will either support or challenge the lower forecast. Strong commentary from the largest buyer group would likely ease concerns, while hesitation would give the revised numbers more weight. For SMH holders, concentration is a key factor, because the fund's largest positions are the companies most exposed to accelerator demand, so its performance will track that debate closely in the coming weeks.

Where Things Stand

A reported forecast cut, not a company warning, triggered Monday's chip sell-off. The figures remain unconfirmed, and Nvidia's results on November 17 are the first hard data point that will test whether AI accelerator spending is slowing as the revised model suggests.