AI-led concentration is creating a market of stocks, not a stock market: Evercore

Evercore says S&P 500 concentration is at a record, with the top 10 stocks now near 40% of index weight and a handful of AI names driving more than 40% of the year-to-date revision in 2026 EPS. The firm kept its S&P 500 year-end target at 7,750 and bull case at 9,000, but warned renewed geopolitical escalation could drag the index back toward its 200-day moving average around 6,800. The backdrop remains inflationary, with core PCE at 3.3% year-on-year, the highest since 2023.

Analysis

The key second-order effect is that index-level strength is increasingly being monetized through a narrow AI capital cycle, while the rest of the market is becoming a valuation and margin-pressure story. That creates a fragile equilibrium: the megacaps can keep carrying the tape as long as hyperscaler capex and model demand stay intact, but the broader market is far more exposed to higher discount rates, weaker cyclicals, and any oil-driven hit to consumer spending. In practice, this means passive inflows can stay supportive even as breadth deteriorates further — a setup that tends to compress realized volatility until a catalyst forces correlation higher.

NVIDIA and the broader AI supply chain remain the cleanest relative winners, but the trade is becoming increasingly about positioning and execution quality rather than simple theme exposure. The more important nuance is that AI beneficiaries outside the absolute leaders may outperform on revised earnings durability, especially where expectations are still less crowded. Alphabet looks more interesting as a catch-up beneficiary than as a pure AI beta trade because its optionality is being underwritten by cash generation, while Evercore’s upside case implies the market is still willing to pay for companies that can convert AI into visible EPS.

The real risk is not just geopolitics or inflation in isolation, but the combination: higher oil prolongs sticky inflation, which keeps rates higher for longer just as equity concentration makes the index more vulnerable to any de-rating in megacaps. That makes the next few weeks a tactical window rather than a fundamental regime shift; absent a de-escalation, the market can easily rotate from complacent grind-up to a fast air-pocket toward technical support. If tensions cool and oil retraces, the market likely re-prices back toward broader cyclicals and small caps, but that requires a cleaner inflation impulse than the tape currently offers.

Consensus is probably underestimating how much of the AI trade is now a balance-sheet and supply-chain story, not just a software story. The winners are the firms that can fund capex, secure power, and preserve margins; the losers are the secondary participants without pricing power or differentiated access to compute. That argues for owning the strongest franchises and fading the index illusion rather than blindly buying broad tech beta at current concentration levels.

Key Decisions for Investors

• Long NVDA / short equal-weight S&P 500 (RSP) for 2-6 weeks: captures AI earnings leadership while hedging against breadth deterioration and non-AI index weakness; best risk/reward if megacap flows persist but internals stay fragile.
• Buy GOOGL calls or call spreads with 1-3 month tenor: lower crowding than NVDA, strong free-cash-flow support, and upside if AI monetization broadens from infrastructure spend into advertising/search productization.
• Initiate a tactical short in RSP or IWM against long NVDA for the next 2-8 weeks: if inflation/geopolitical noise keeps rates elevated, broad beta should underperform the AI complex by several hundred bps.
• Use NVDA downside puts as a portfolio hedge into geopolitical headlines: concentration makes the index vulnerable to a 3%-5% air-pocket in megacaps even if the broader market only grinds lower.
• If crude and inflation expectations roll over, rotate partially from mega-cap tech into cyclicals only after confirmation; until then, do not chase broad market strength because the upside is increasingly index-concentrated and technically brittle.