Capital Economics expects the AI-driven stock rally to resume after the recent selloff and is holding its end-2026 S&P 500 forecast at 8,250. But the firm warns the earnings assumptions behind the rally look too optimistic, projecting the index to fall to 6,500 by end-2027.
Capital Economics has told clients the recent selloff in AI-linked tech giants reflects fear rather than hard evidence, and it expects the rally to resume, at least for now. Chief Economic Adviser John Higgins is holding the firm's end-2026 S&P 500 forecast at 8,250, despite the recent pullback in AI-trade names. Beyond that horizon, however, he expects the earnings assumptions driving the rally to prove too rosy.
S&P 500 seen falling to 6,500 by end-2027
According to Capital Economics, earnings expectations behind the rally "look exceptionally optimistic" and will be scaled back over time. As a result, the firm forecasts the S&P 500 falling to 6,500 by end-2027, even as it expects the near-term rally to continue.
Chip supply and China are pressuring the trade
Higgins pointed to several sources of pressure on AI stocks. He noted that Chinese memory maker CXMT listed this week, a step that could ease a shortage of mainstream memory chips even though it would leave high-bandwidth memory supply for AI largely unaffected. Separately, China's progress toward its own deep ultraviolet lithography machines has weighed on ASML shares, even though the Dutch firm's extreme ultraviolet machines remain far more advanced.
Reports that Nvidia may provide around $250 billion in financing guarantees for OpenAI have also revived concerns about circular financing in the AI buildout.
Memory chipmakers reinforce the AI trade
Even as some analysts flag risks, others are underscoring the AI infrastructure buildout supporting the rally. Bernstein reiterated its bullish stance on memory chipmakers after more than $700 billion in partnerships were announced at a South Korean government AI summit in San Francisco, where SK Hynix, Nvidia, Samsung and Broadcom unveiled new deals spanning memory, foundry and data centers.
SK Hynix and Nvidia signed letters of intent covering a partnership worth more than $500 billion, including memory supply and a 2GW AI data center project planned for 2027. Samsung and Broadcom, in turn, signed a memorandum covering $200 billion of memory and foundry services through 2030. Analysts led by Mark Li said the deals point to consensus expectations of roughly $1.3 trillion in annual memory revenue in both 2027 and 2028 — underscoring how central memory chips have become to the AI trade that Capital Economics still expects to climb, at least for now.
Source: Investing.com
Trading involves risk.
Risk Disclaimer
