AI Stocks Split as Investors Reward Sellers and Punish Big Spenders

The cleanest divide is between spenders and sellers.

Different Ways to Skin Capex

Via Jim Reid, Deutsche Bank

For most of the AI boom to date, if a company said “AI” with enough conviction, its shares went up. That changed on 30 October 2025. This was the day after Alphabet, Meta and Microsoft reported, when the market made its first serious attempt to distinguish between AI spenders and AI earners. Meta fell -11.3%, its worst day in three years, after guiding to capex “notably larger” in 2026. Microsoft slipped -2.9% despite strong cloud numbers. Alphabet, which had just posted its first $100bn quarter, rose +2.5%. A year earlier the consensus had Meta as an AI outperformer and Google as a likely casualty of generative search. Since then, the AI basket has continued to rally in aggregate, but the question is no longer whether AI will pay off, but for whom. These concerns have resurfaced again in the last few weeks.

Today’s chart plots over a hundred global AI-related stocks by market cap against their performance since 29 October 2025, the eve of the day the AI trade stopped being one trade. The basket is up around +18% since, but that headline hides a market that has spent the year sorting AI exposure into those who pay and those who get paid.

The cleanest divide is between spenders and sellers. The companies writing the AI cheques have generally gone nowhere: Microsoft (NASDAQ:MSFT) is c.-2%, Meta (NASDAQ:META) c.-6%, Broadcom (NASDAQ:AVGO) c.-15%, and Amazon (NASDAQ:AMZN) is up only c.+9%. Nvidia (NASDAQ:NVDA), the poster child of the boom, is only up c.+3%. The companies cashing those cheques have soared. SK hynix and Samsung are up over +100%, and Marvell, Seagate and Western Digital are even further ahead, while TSMC (NYSE:TSM) is up c.+60% and is the best performer of the trillion-dollar market-cap club.

There are two important mega-cap exceptions to that broad spender story. Apple (NASDAQ:AAPL) is up around +27% and Alphabet (NASDAQ:GOOGL) around +24% over the period. Apple is arguably the cleaner exception because, unlike its hyperscaler peers, it has not pursued the same data-centre capex arms race. The market has rewarded a mega-cap technology company generating strong earnings growth without needing to write enormous AI infrastructure cheques. Alphabet is different: it is both spender and increasingly demonstrable earner. Its cloud business has been the standout growth engine, while Search, YouTube and Cloud have all benefited from AI products such as Gemini and AI Mode. So, its share-price strength fits the wider theme rather than contradicting it: investors appear willing to fund very large AI budgets when there is visible evidence that those budgets are already translating into revenue growth. While Microsoft and Meta are also seeing clear AI revenue growth, Alphabet is coming from a base of lower expectations and has been particularly successful in recent months at demonstrating multiple profitable uses of its AI capabilities simultaneously.

The second divide is atoms versus bits. The outperformers are almost all physical bottlenecks: memory, storage, chip packaging, optical networking and servers. The underperformers are dominated by software, with Adobe, Intuit, ServiceNow, Salesforce (NYSE:CRM) and Wix all in the red, as investors decide AI is more likely to eat software than be sold by it. Fifteen years after software was supposedly eating the world, the menu appears to have been reversed.

That split also runs through Asia. Taiwanese and Korean hardware names such as MediaTek, ASE and SK hynix are among the biggest outperformers. India’s IT services giants — Infosys, Wipro, HCL and Tata — sit almost entirely below the line. The same technology is shaping up as the best and worst thing to happen to two Asian export models.

Then there are the leverage casualties. Oracle (NYSE:ORCL), CoreWeave (NASDAQ:CRWV) and Super Micro Computer (NASDAQ:SMCI), the names most associated with debt- or contract-funded build-outs, are each down in the 40-60% range since 29 October 2025.

What holds this together is that most of the mega caps have acted as ballast rather than the engine, with Apple and Alphabet the notable exceptions. Much of the basket’s roughly +18% gain in just under a year has mainly come from the $100bn–$1trn tier rather than simply from the biggest companies as was previously the case. That said a look at the second chart at the end, that removes the log scaling, shows just how big, powerful and important those mega-cap names are.

More importantly, mega-cap capex is the revenue of almost every infrastructure winner on this chart. So, the AI trade probably cracks if mega-cap spending unexpectedly slows — or if investors stop believing that spending is generating an adequate return. So although investors are no longer simply just investing in the mega-caps, they are relying on their budgets, and increasingly discriminating between the companies spending those budgets and those proving they can monetise them.