Record- Low Valuation Gap Suggests Hyperscalers Are Less Expensive Than the Hype Says

Hyperscaler earnings have narrowed the valuation gap

After years of hyperscalers carrying more market value than their profits justified, the group now sits at a record-low negative 5.22% gap as of July 24. In plain English, a negative reading means these companies' share of aggregate earnings now exceeds their market-cap weight. That does not make them cheap in absolute terms. It does mean investors can no longer assume the market is giving them AI exposure for free.

Why the shift matters

The valuation gap narrowed because earnings finally chased the story. Recent results support that view: Microsoft reported fiscal-year revenue increased 18% to more than $331 billion, while operating income rose 21%. Azure growth accelerated to 43%, and Microsoft's commercial backlog reached $678 billion. That is what it looks like when AI demand begins to show up in revenue, margins, and deferred contracts.

Bulls see a group that is still benefiting from AI enthusiasm but is now also being validated by profits. Bears see a crowded trade whose relative valuation has finally improved after a long catch-up. The real question is no longer whether hyperscalers matter to AI. It is whether earnings can keep closing the gap between sentiment and fundamentals.

Micron and LPL show why ROI, not demand, is now the debate

The valuation gap has improved, but the market is still fixated on whether the AI story is real instead of focusing as closely on how well that story earns capital.

Strong results no longer guarantee a higher multiple

Micron is a useful read-through. The company delivered record revenue, along with record adjusted margins, rising memory prices, and AI demand that continued to keep supply tight. Even so, the stock wobbled, and by mid-July the higher-beta AI complex had already slipped into correction mode.

That reaction suggests the market is moving past a simple demand narrative. The issue is no longer just whether AI demand exists. It is whether current valuations already assume too much execution and too little friction.

LPL frames the next test around returns on capital

LPL Research frames the debate cleanly: the question is no longer how much hyperscalers are spending, but whether that capex can earn attractive returns. Its approach evaluates the return profile of an average hyperscaler through the lens of ROIC, growth, and capex. That is the real hurdle for a broader rerating.

Bears still worry AI spending could turn into a money pit. Bulls still worry demand alone may be enough to justify the buildout. The more useful question sits between those views: can hyperscalers convert capacity into durable returns rather than just larger top-line numbers?

Return quality matters more than headline demand

Demand looks stronger than it did a few years ago, but the key qualifier is still return quality. The same report highlighting Microsoft's growth also underscores the importance of whether cloud growth and margins can absorb rising infrastructure costs. That is the more important filter now.

The investment question, then, is not whether AI demand is real. It is whether hyperscalers can translate that demand into durable margins and economic profit.

Slower capex growth could help the valuation case

The valuation-gap setup only works if the next phase of the story is less about bigger spending and more about better economics. UBS estimates hyperscaler capex will rise 76% this year, then slow to 25% next year and just 6% in 2028. That path matters because the market does not need AI demand to cool. It needs spending growth to cool while monetization holds.

If that happens, hyperscalers with clear proof of monetization and backlog conversion could continue to make the case. If not, the market may decide the current premium still depends too heavily on perfect execution.