Goldman's overlooked Al basket just ripped 115%

The biggest money in markets is rarely made where the crowd is looking. By the time a trade has a nickname, a dedicated cable segment and a line of exchange-traded funds chasing it, the easy gains have usually already been handed out. The real edge tends to sit in the unglamorous corners that bore most people into looking away.

For three years, the artificial intelligence (AI) rally has had exactly one address. It runs through the chipmakers, the hyperscalers and the software platforms, the names everyone can recite from memory. Investor attention is still trained on the frontier AI labs and the companies feeding them silicon, according to Axios.

That is the trade your neighbor brags about at a barbecue. It is also the trade now priced for something close to perfection, which is why a quieter question has started circulating on trading desks. If the obvious AI winners are already this expensive, where does the next leg of the money actually go?

I went hunting for that answer in the research this month, and the most striking number I found came from Goldman Sachs (GS).

A basket of stocks the bank has been tracking gained 115% since late 2025, while a comparable group of more fashionable names returned just 7% over the same stretch, according to Goldman Sachs Research.

The catch is that almost nobody owns it on purpose.

J Studios &sol Getty Images

How the AI boom spills into the physical economy

To understand why this corner matters, start with the sheer physical weight of what AI now requires. A model does not live in a cloud. It lives in a building full of servers that draws power like a small city, sits on a grid that has to be upgraded, and runs on metal that has to be dug out of the ground somewhere.

That is why the spending tied to AI keeps leaking out of Silicon Valley and into factories, mines, utilities and oil rigs.

Goldman estimates roughly $7.6 trillion will be invested globally in AI infrastructure between 2026 and 2031, across compute, data centers and power, according to Axios.

More Artificial Intelligence:

• Mark Cuban reveals what people really hate about Al data centers
• The Al memory crunch just came for Xbox gamers, too
• How the White House is quietly bottlenecking Al

The demand shows up first as electricity. A single large data center campus can pull as much power as a mid-sized town, which pushes utilities to build, grid operators to upgrade and the makers of copper wire, transformers and turbines to run flat out.

None of that is glamorous. All of it has to happen before a single new chatbot answers a single new question.

Software, meanwhile, accounts for less than 0.5% of global gross domestic product, leaving what Goldman calls "the other 99.5%" of the economy as AI's next frontier, according to Axios.

The people running Goldman's industrials practice say the conversation with corporate clients has already flipped. The question is no longer whether factories will adopt automation. It is how fast it spreads, and which companies get left behind for moving too slowly.

Related: Goldman Sachs has blunt message for AI stock investors

What Goldman's 115% AI basket actually owns

The basket is not a pile of chip stocks. It is built from emerging-market companies in capital-intensive industries, the grid operators, utilities, manufacturers and heavy-asset businesses that supply the picks and shovels for the AI buildout, according to Goldman Sachs Research.

Goldman strategists Sunil Koul and Tarun Lalwani gave the idea a label that is now spreading across Wall Street: HALO, for heavy assets and low obsolescence. The thesis is that businesses anchored to physical infrastructure are hard to copy and hard for software to disrupt, so they should hold their value even if AI upends large parts of the white-collar economy.

The idea did not start at Goldman. The HALO label bubbled up among investors earlier in 2026 before the bank formalized it into a research framework and a tradable basket.

Two forces are doing the heavy lifting underneath it. Governments are pouring money into re-industrialization and defense to harden critical supply chains, and the AI buildout is stacking fresh demand for power and hardware on top, according to Goldman Sachs Research.

What jumped out at me when I lined the two baskets up was not the 115% gain. It was the price tag attached to it. Here are the numbers that frame the trade:

• The emerging-market, capital-intensive basket gained 115% since late 2025, against 7% for capital-light names, as of June 5, according to Goldman Sachs Research.
• Even after that run, the basket still trades at a 20% discount to the capital-light group, according to Goldman Sachs Research.
• An earlier, broader version of the trade had already beaten its capital-light counterpart by about 35% since the start of 2025, according to Bloomberg.
• Tech mergers and acquisitions hit $566 billion in 2026, up from $334 billion in all of 2025, according to Dealogic data cited by Goldman.

A stock that doubles and then keeps looking cheap is unusual. It usually means the market does not yet believe the earnings are real. Goldman thinks they are, and expects the gap to keep widening into 2026 and 2027.

What the asset-heavy AI trade means for your money

Here is the part that should make a regular investor sit up. You may already own a slice of this trade without knowing it.

The utilities, industrials and emerging-market holdings buried inside a target-date fund or a broad index fund are exactly the kinds of unglamorous, asset-heavy names this rotation is rewarding. The boring sleeve of your portfolio, the part you never check, may quietly be doing the heavy lifting.

For anyone who spent the past two years feeling priced out of the AI story, that is a strange kind of relief. The trade was never only the stock you were too nervous to buy at the top. Some of it was sitting in the safest, dullest fund you already owned, earning nothing in the way of bragging rights.

Goldman frames the appeal in plain terms. Capital-light businesses in software and IT services, the firm warns, remain "increasingly at the risk of disruption from artificial intelligence," according to Goldman Sachs Research. The hard-asset companies, by contrast, own things AI cannot conjure into existence.

There is a fair counterargument, and it deserves space. Goldman's own report concedes the trade closely tracks the old value versus growth rotation, which means some of this may be a familiar factor dressed in new clothing rather than a fresh discovery. 

The honest read sits in between. The 115% number is real, the discount is real, and the capital is genuinely flowing into power and steel and grid. Whether that makes the next two years or simply rhymes with the last value head-fake is the part no banker can promise.

What I keep coming back to is the setup itself. The most crowded trade in the world spent three years convincing everyone that AI lived in a chip. The data now says a large share of the payoff is being quietly collected by the companies that pour the concrete and keep the lights on. The next time the AI rally wobbles, that is the corner worth watching, because it is the one almost nobody is watching yet.

Related: Amazon quietly raises price tag on the AI boom