Frontier Tech Baskets Chase the Future. The Grid Is the 128-Week Box They Must Cross.

A data center can order every GPU it wants, sign up for a thousand software seats, and still be unable to turn its racks on. The missing part is not a chip. It is a high-voltage power transformer — a shed-sized box of copper, steel, and cooling oil that now takes roughly two and a half years to build. Capital arrived. Capacity did not.

That is where the "frontier tech" trade runs into the physical world. Over the past few years investors have been handed a heady new way to own the future: thematic baskets that bundle artificial intelligence, robotics, autonomous driving, and fintech into a single fund. The Innovator Deepwater Frontier Tech ETF, built on the Loup Frontier Tech Index, is one of the better-known examples, and it advertises itself by how little it overlaps with the Nasdaq 100. The pitch is access to the companies building the future, the ones not in the obvious index.

The baskets have real money behind them. Robots and AI is now the largest thematic ETF category, and robotics-and-AI funds pulled in $19 billion over the trailing twelve months. The market is so convinced this is a durable theme that exchange operator CME launched single-stock futures on more than fifty of the biggest U.S. technology and software names in late July — building the trading tools for a boom it expects to keep running.

Every one of those themes has the same quiet dependency. Chips need to be cooled, robots need to be built, data centers need electricity — and the grid is the one layer where money cannot buy its way to the front of the line. The hyperscalers — Alphabet, Microsoft, Amazon, Meta, Oracle — are pouring over $800 billion into AI data centers this year and plan to add more than $1.2 trillion next year. That is demand measured in hundreds of billions. The transformer that connects a new facility to the grid is a couple of million dollars. But there are not enough of them, and they cannot be rushed.

A box of copper and steel is now the constraint

The numbers on the ground are striking. Power transformer lead times now average about 128 weeks, or two and a half years, with generator step-up transformers running even longer at 144 weeks. Prices have risen roughly 77% since 2019, and demand for large power transformers is up about 116% over the same stretch. Analysts expect 40% of AI data centers to be power-constrained by 2027, and Bloomberg has reported that almost half of the U.S. data centers planned for 2026 face delays or cancellations — not for lack of capital or chips, but because the power cannot arrive in time.

This is the classic setup this column hunts for: theme demand colliding with a step that takes years, not quarters, to relieve. Being necessary gets you orders. Being scarce decides who keeps the money. The interesting question is who is standing at that gate.

The tollbooth, and how much of it actually reaches earnings

GE Vernova makes the equipment the whole boom has to wait on: gas turbines to generate power, and the transformers and switchgear that move it onto the grid. The rent-capture evidence is unusually clean. Its gas turbine backlog has hit 100 gigawatts as prices rise, and the company expects to be largely sold out through 2030, with customer down payments helping to fund its own capacity expansion. In the first quarter of 2026 it booked $18.3 billion of orders, up 71% from a year earlier, a book-to-bill ratio of roughly two times. By the first half of the year, equipment orders were priced more than 20% above the quarter-earlier level. That is not a company hoping for demand. That is a company printing pricing power because its order book is full.

The big honesty gap shows up when you ask how purely the theme reaches the P&L. GE Vernova is not a pure data-center play; it also owns wind turbines and nuclear, and wind has been the laggard dragging on results. The market has nevertheless noticed the power story — the stock is up about 46% year to date — and it trades at roughly 27 times trailing earnings. Compare that with the alternatives. Vertiv, the cleanest play on the power and cooling that keeps a data center alive, trades at about 57 times earnings. Quanta Services, the contractor that actually builds grid and power infrastructure, trades near 74 times. On a price-to-earnings basis GE Vernova is the cheap one of the group — but that discount is partly because its earnings are diluted by the wind and generation businesses the pure plays do not carry.

So the trade has a fork. Buy the purer names and you pay far more per dollar of the theme, accepting that a downturn hits a concentrated bet hardest. Buy Vernova and you get the scarce equipment maker at the lower multiple, but you are also buying the ugly wind business and hoping the grid-and-turbine engine is large enough to swamp it.

The clock on the scarcity

No bottleneck is permanent, and this one has a date stamped on it. Manufacturers are spending billions on new transformer capacity — Siemens Energy, for instance, is building a $421 million plant in North Carolina targeted for 2027. Analysts still expect U.S. transformer supply to run short for years even with the new factories, so the scarcity looks durable, not seasonal. But there is a minority view worth keeping in mind: some brokers argue the shortage is partly self-inflicted, an artifact of procurement practices rather than an absolute lack of supply. That dispute matters, because the whole thesis rests on the grid staying the constraint.

The confirmation metric to watch is the lead time. So long as transformers and turbines are quoted out years and priced up, the queue is real and the tollbooth collects. The signal that the hidden winner is becoming ordinary is the opposite — when transformer lead times compress, when GE Vernova stops being sold out through 2030, when new capacity turns the wait into girth. The good news for holders is that capacity of that kind does not appear overnight. The clock is running one way that is favorable to the equipment makers, and it is measured in months, not adjectives.

The frontier tech basket is a genuine thing with real flows behind it. But the bundle of poster children is not where the money must physically go — that money has to cross a transformer. If you want the part of the theme that scarcity, not narrative, has priced, that is where to look. And the day the lead time breaks, that is the day to stop looking.