Three AI Boom Stocks Have Fallen as Much as 36%. Which One Deserves the Drop?

Google and Amazon’s nuclear power supplier, Uber’s most profitable year ever, and a gaming engine with a record quarter. All three are reporting their best numbers ever, yet the market is selling them off. Each for a different reason, and only one of those reasons makes sense.

Record results, 12-month lows

A year ago, they were among the sure winners of AI and electrification. Constellation Energy $CEG was set to supply power to data centers, Uber $UBER was expected to become a robotaxi platform, and Unity Software $U was poised to sell AI-powered advertising in games on billions of phones. Today, all three are 24 to 36% below their 12-month highs.

Yet none of these declines is based on the latest earnings reports: all three companies delivered record quarters in August. The market is punishing them for things that have yet to happen: rules the regulator has not written, an app that will not launch until January 2028, and an advertising network at the center of a dispute with its biggest competitor.

Constellation: two 20-year contracts, $5 billion from the government, and still at the bottom

In the second quarter, Constellation raised operating earnings to 2,55 USD per share from 1,91 USD, lifted its full-year outlook to 11,50 to 12,50 USD, and signed another 920 MW in long-term nuclear power supply contracts. Then came the fall: on Sept. 30, a 20-year contract with Amazon from the Calvert Cliffs plant, and on Oct. 6, a 20-year contract with Google for 890 MW of new capacity across eleven reactors, which will require Constellation to invest more than 4,3 billion dollars. The stock jumped more than 15% above $300 during that day’s session, then gave back the entire gain over three weeks.

The reason lies a level above the company. On Sept. 29, FERC suspended for five months a backup capacity auction in the PJM grid, the largest energy market in the US, through which Constellation sells most of its nuclear generation. The auction was meant to secure 6 831 MW of missing capacity and determine who would pay for it. Without it, the market does not know how much data center power supply will actually earn. So far, the company has roughly 30% of its net generation covered by long-term contracts, free cash flow over the past 12 months of just 309 million USD, and debt of 24,7 billion USD after acquiring Calpine. Even a conditional loan of up to 5 billion dollars from the US Department of Energy on Oct. 27, intended to increase the capacity of eight plants, failed to stop the decline: the stock lost 4,7% in a week.

We have a lot of unused capacity in both generation and transmission for more than 99% of the hours in a year. We have a peak-capacity problem, not an energy problem.

Joe Dominguez, President and CEO of Constellation Energy, second-quarter 2026 earnings call

Still, Constellation has the most secured future of the three: its contracts with Google, Amazon, Meta, and Microsoft run for 15 to 20 years, and the restart of the Crane plant remains on schedule for the second half of 2027. But Bulios’s fair value of 129 USD is a reminder that even after the decline, the stock’s P/E of 26,3 values it like an AI company, not a power plant (the industry average is 18,6).

Uber: $10 billion in cash a year, and a 12-month low a week before earnings

In the second quarter, Uber increased gross bookings by 22%, trips by 18% to 3,9 billion, and non-GAAP operating income by 40% to 2,1 billion dollars. Free cash flow over 12 months exceeded 10 billion USD for the first time. Yet on Oct. 28, the stock closed at 65,27 USD, just above a new 12-month low of 64,69 USD.

Gross bookings grew 22%, non-GAAP earnings per share rose 35%, and free cash flow over the last twelve months exceeded 10 billion dollars for the first time in Uber’s history.

Balaji Krishnamurthy, Uber CFO, press release on second-quarter 2026 results

The market is not selling the numbers; it is selling the scenario. In July, Waymo announced that it would launch its own app in Austin and Atlanta in January 2028, ending its exclusivity through Uber. In October, Tesla reported that miles driven by its unsupervised robotaxis had increased roughly eightfold over the quarter. Uber is now like a used-car dealership selling vehicles from every brand. But the biggest automaker is opening its own showroom right next door.

That is exactly where Uber differs from the other two. It is the only one of the three whose price is supported by fundamentals: a P/E of 13,9 versus a five-year average of 32,9, a free cash flow yield of 7,6%, and a Bulios fair value of 65,94 USD, virtually identical to its market price. The market is pricing in a threat that has not yet shown up in the numbers.

Unity: the smallest decline, the biggest question mark

In the second quarter, Unity increased revenue to 546 million USD from 441 million a year earlier. Its strategic advertising division, Grow, powered by the AI platform Vector, grew 63% to 329 million dollars, and Vector reached a billion-dollar annual revenue run rate two quarters earlier than the company had promised. The stock jumped 17% after the results.

This was probably the best quarter in Unity’s history as a publicly traded company.

Matt Bromberg, President and CEO of Unity, press release on second-quarter 2026 results

The company is deliberately shrinking its business: in April, it shut down its old ironSource Ads ad network, and in August it sold its Supersonic publishing business. It is sacrificing volume in favor of margins. The risk lies elsewhere: AppLovin, its biggest competitor in mobile advertising, is suing Unity over data collection through its Ad Quality tool. On Oct. 2, a court in San Francisco refused to block Unity from using the tool pending trial, but the dispute is continuing in arbitration. The entire Vector growth story thus rests on data that Unity is being sued over.

And here is the trio’s paradox. Unity has fallen the least, but it is the only one still unprofitable: a net margin of -28,9% over 12 months, a loss of 587 million dollars, and a P/S of 8,6. Bulios’s fair value of 28 USD is 29% below the market price. Of the three declining stocks, this is the only one where the hype remains in the price.

Three dates in November: how to tell who deserves the drop

All three companies report within one week, and each has to deliver something different:

• Uber, Nov. 4 (estimated EPS 0,86 USD, revenue 14,7 billion USD): the key question is whether trips are growing in Austin and Atlanta, where Waymo operates alongside human drivers, and what management says about January 2028. Trip growth of around 18% would mean robotaxis are still adding volume to the platform, not taking it away.

• Unity, Nov. 5 (estimated revenue 567 million USD, EPS 0,03 USD): watch whether Grow’s strategic revenue reaches the promised 380 to 385 million and whether the company reports a net profit under GAAP for the first time.

• Constellation, Nov. 9 (estimated EPS 3,85 USD, revenue 9,2 billion USD): the key figure is not earnings, but new megawatts in long-term contracts and the share of generation under contract rising above 30%. Then, by year-end, PJM must redesign its backup auction; if FERC allows it to start before Feb. 28, 2027, the stock’s main headwind will disappear.

For now, the picture is this: Uber has cash and profits today, Constellation has contracts for the next 20 years, and Unity has a record quarter but no profits. The market has punished the company with the most signed contracts the most, and the one with the most promised the least. November will show whether things even out.