Evercore strategist outlines bullish scenario with S&P 500 potentially reaching 9,000

Evercore ISI analyst Julian Emanuel said his central forecast places the S&P 500 at 7,750 by the end of 2026, while also assigning a 30% likelihood to a more optimistic scenario in which the benchmark index rallies to 9,000 on the back of AI-related momentum in technology, communications and consumer-focused companies.

Writing in a note to clients on Monday, Emanuel argued that investors are navigating an environment shaped by both a structural technology boom and major geopolitical change, creating a wider range of potential outcomes than conventional market models typically capture.

He compared the current backdrop to earlier transformative periods, writing: “The Pandemic changed everything. Warlike stimulus, surging M2, and a productivity shock collide with an ‘AI Revolution’ – reminiscent of the 1920s and 1990s.”

Emanuel believes these forces could ultimately lift productivity growth to 3% by the end of the decade.

Evercore ISI said investors may benefit from long-term call positions tied to what it called the “AI Class of 2026” companies as well as the QQQ ETF, citing the possibility of “seemingly unimaginable” upside potential.

At the same time, the firm recommended a collar strategy using the SPY ETF to help manage shorter-term risks related to rising oil prices and interest rate fluctuations.

Despite the bullish longer-term outlook, Emanuel stressed that artificial intelligence remains constrained and probabilistic in nature. According to Evercore ISI, large language models often demonstrate a “Narrow Consensus” bias by clustering around mainstream expectations and failing to adequately capture tail-risk scenarios.

Emanuel added that durable competitive advantages are more likely to come from specialized domain expertise and ownership of integrated workflows than from AI technology by itself.

He also warned that prediction markets tend to mirror prevailing investor sentiment rather than reliably forecasting future outcomes, limiting their usefulness for evaluating long-duration or highly skewed scenarios.