Strong AI Earnings No Longer Move the Needle for Investors

Good numbers aren't good enough anymore. That's the uncomfortable lesson from this week's earnings parade, where a string of AI-related companies posted results that would have sparked rallies a year ago. Instead, investors sold.

The disconnect matters because it suggests something has shifted beneath the surface of the AI trade. Strong revenue, fat backlogs, confident guidance: none of it worked. That points to growing doubts about whether the AI boom can keep delivering returns commensurate with its valuations, not just whether the underlying businesses are healthy.

The casualty list is long and recognizable. Meta dropped 9.4% this week. Comfort Systems, the HVAC provider riding the data center cooling wave, fell 19.3%. Quanta, the engineering and construction giant building out AI infrastructure, slid 13.5%. Power-trade names took hits too: GE Vernova fell 11.5%, Bloom Energy dropped 12.7%. None of these are speculative story stocks. These are companies with real revenue tied to real AI buildout.

The Earnings Paradox

Meta's third-quarter results, reported Wednesday, were genuinely strong. It was a clean beat-and-raise, the kind of print that typically earns a company a victory lap on CNBC. The stock fell 7% the next day anyway.

Comfort Systems followed a similar script Thursday. Its order backlog nearly doubled year over year, driven by data center cooling demand. Results topped expectations across the board. The stock still dove 9.3% on Friday.

Western Digital delivered the most startling number of the week. The hard disk drive maker, once a commodity hardware player, has become essential to data center buildouts. Its earnings per share rocketed up 161% year over year. Investors shrugged. Shares slipped 3%.

Three different companies, three different businesses, one consistent pattern: spectacular results, negative stock reaction. That's not noise. That's a market recalibrating what it's willing to pay for AI exposure, regardless of how the underlying numbers look.

The pain wasn't limited to companies reporting this week. Scan the S&P 500's worst performers over the period and the list reads like a greatest-hits reel of last year's most beloved AI trades. Quanta, EMCOR and Vertiv in engineering and construction. KLA and Lam Research in chip-making equipment. Coherent in the fiberoptic cables that wire together server farms. These stocks didn't need bad news to fall. They just needed the market to stop rewarding good news the way it used to.

Why It Matters

This isn't the first time the AI trade has wobbled. The DeepSeek scare in early 2025 rattled the same cohort of stocks and proved temporary. Markets recovered, AI names resumed their climb, and the episode became a footnote.

This feels different in one important respect: the quality of the news flow. DeepSeek was a genuine shock, a competitive threat that forced investors to reassess assumptions about who would win the AI race. This week's sell-offs came alongside objectively excellent earnings. There was no bad news to explain the drop. That's a tougher signal to dismiss.

The stakes here go well beyond a handful of tech names having a rough week. AI-related shares, using the broader tech sector plus Meta, Amazon and Alphabet as a rough proxy, account for more than 50% of the S&P 500's total market capitalization. That's not a sector rotation story. That's a referendum on the index itself.

It's also not confined to companies that make chips or software. AI-related sales have become a meaningful swing factor for plenty of non-tech names across the index, from industrial firms selling into data center construction to utilities powering server farms. When the AI trade sneezes, the contagion doesn't stop at the edge of the tech sector.

The mechanism at work looks like simple multiple compression. Investors aren't questioning whether these companies can grow. Western Digital's 161% EPS growth settles that question. What they're questioning is whether current valuations already assume a flawless execution path stretching years into the future, leaving no room for upside surprises to actually move the stock.

When a stock is priced for perfection, perfect results become the floor, not the ceiling. Anything short of a miracle gets treated as disappointment, even when the headline numbers are objectively strong. That's the dynamic now playing out across the AI complex, and it's a far less forgiving setup than the one that prevailed for most of the past two years.

The Bottom Line

The market has spent the better part of two years treating AI earnings beats as a one-way bet. This week broke that pattern, and broke it with some of the strongest results the sector has produced all year.

If this is a temporary gut check, akin to the DeepSeek scare, the AI trade will likely resume its climb once sentiment resets. If it's the start of a durable repricing, the implications run deep. Half the S&P 500's market cap sits on the AI thesis holding up. A market that can't find upside in beat-and-raise quarters from Meta, Comfort Systems and Western Digital is a market signaling that expectations have outrun even very good reality.

Investors should treat this week as a stress test, not a verdict. But the results of that test, strong earnings met with a cold shoulder across multiple sectors and company types, deserve serious attention heading into the next round of reports. The AI trade has survived scares before. Surviving disappointment is one thing. Surviving indifference to good news is a different and harder problem.