Goldman Sachs created an index that strips AI from the S&P 500. The split — 45% AI, 55% everything else — reveals that passive investing is no longer passive. Anyone holding the benchmark is making an active bet on a single thesis.
Goldman Sachs just created an index for the other half of the economy. Not a sector fund, not a thematic basket, not a bet against artificial intelligence. An index that tracks the S&P 500 minus everything AI-related. They call it SPXXAI. The name is the product: the benchmark, minus the thing everyone is talking about.
The split is nearly even. Roughly 45% of the S&P 500 by weight is classified as AI-related — companies that build, enable, or depend on AI infrastructure. The remaining 55% is everything else: consumer staples, healthcare, industrials, utilities, financials, energy. Over the past three years, the full S&P 500 returned 76%. The ex-AI index returned 32%. Same economy, same interest rate environment, same consumer. The 44-point gap is explained entirely by AI.
The Product Is the Revelation
Goldman doesn't create indices as intellectual exercises. Institutional clients demanded something that didn't exist — a way to hold the S&P 500 without being overweight to a single thesis. Louis Miller, head of the firm's equity custom basket desk, described the index as eliminating 'the noise introduced by the AI hype.'
That word — noise — is doing significant work. To call AI hype 'noise' is to imply there's a signal underneath it. The signal is the rest of the economy: companies that generate revenue from physical goods, regulated services, and human labor. These companies returned 32% over three years. In any other era, that would be strong performance. The problem is that anyone holding the S&P 500 got 76%, meaning the benchmark's return is almost entirely an AI trade wearing a diversified costume.
This is the real product Goldman is selling. Not the index — the revelation. Passive investing in the S&P 500 is no longer passive. A pension fund benchmarked to the index is not 'diversified.' It is 45% exposed to a single technological supply chain. SPXXAI makes this explicit. Before, the concentration was a fact. Now it is a position.
The Coherence Problem
Index concentration is not new. The Nifty Fifty of the 1970s, the dot-com top-heaviness of 1999, the FAANG dominance of 2015 to 2020 — each era had a small number of stocks that drove the benchmark while the majority lagged.
What distinguishes 2026 is the coherence of the concentration. In 1999, the top tech stocks included Amazon, Cisco, Intel, Yahoo, and AOL — companies with fundamentally different business models and risk profiles. A slowdown in e-commerce would not necessarily hit networking equipment. Today's AI concentration is a single supply chain: chip designers, cloud providers, server assemblers, and the software companies building on their infrastructure. A slowdown in AI spending would hit most of the 45% simultaneously, because they are not merely correlated — they are structurally dependent on the same demand signal.
This is why the hedge exists. Concentration from unrelated companies is diversifiable by accident. Concentration from a single supply chain is not.
The Gap Between Price and Output
The 44-percentage-point return gap is a measure of something specific: the market's forward estimate of AI's economic contribution, capitalized into stock prices today. As of February 2026, roughly 44% of the benchmark's three-year return is attributed to AI.
Now compare that to Goldman's own economics team, which published research the same week estimating that AI contributed 'basically zero' to U.S. GDP growth in 2025. The capital markets are pricing AI as nearly half the economy's value engine. The economists at the same bank say it hasn't measurably contributed yet.
Both can be right — if the market is pricing future impact, not current output. But the distance between expectation and reality is exactly the risk that SPXXAI was built to manage. Someone — many someones, with enough institutional capital to justify Goldman building a custom index with S&P Dow Jones — looked at that gap and decided they wanted a way out of it.
Not Bearish, Just Honest
The subtlety of SPXXAI is that holding it is not a short on AI. It is a hedge against concentration. An investor who buys the ex-AI index is not saying AI will fail. They are saying: I want market exposure without betting half my portfolio on one thesis that hasn't shown up in GDP yet.
There is something clarifying about looking at the economy with AI subtracted. The 32% return tells you that American companies outside the AI trade are growing, earning, and returning capital at a healthy clip. They are just invisible inside a benchmark that has become a proxy for something else entirely.
The question SPXXAI forces — the question Goldman is now helping its clients express with capital — is whether the 44-point gap is earned. Whether AI will generate the economic output that half the index's returns currently assume. If it does, SPXXAI will underperform the benchmark. If it doesn't, its holders will own the economy that was always there, just hidden behind the noise.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)