Is the AI Bet Starting to Crack? Big Tech's AI Gamble Shows Negative Returns Despite Surge in Spending

The expensive bet on AI is starting to crack after fresh financial modelling cited by the Financial Times suggests that most major US technology firms are still expected to generate negative returns on artificial intelligence investments between 2025 and 2030, even under highly favourable assumptions.

The analysis, produced by Panmure Liberum, indicates that Microsoft, Alphabet, Meta, and Oracle all remain in negative territory on implied AI return on investment, while only Amazon shows a positive outcome. The findings land as Silicon Valley continues to pour unprecedented sums into AI infrastructure, data centres, and computing systems in anticipation of future demand.

Big Tech's AI Bet Starting to Crack Under Cost Pressure

The modelling from Panmure Liberum rests on an unusually generous assumption: that the cost of building and running AI systems is effectively zero. Even with that stripped-back scenario, projected returns across most hyperscalers remain negative over the 2025 to 2030 period.

In practical terms, it suggests that the current scale of investment in artificial intelligence is still not generating sufficient projected financial upside within the medium term horizon being assessed.

What the figures expose is a widening disconnect between capital spending and monetisation. Tech groups are committing vast sums to chips, servers, and large-scale data centres in the expectation that demand for AI services will eventually expand into meaningful revenue streams. Those streams are expected to emerge across cloud computing, enterprise software, and digital advertising, but much of that remains forward-looking rather than firmly established in today's earnings.

Analysts from X describe this phase as a classic infrastructure build-out, where costs arrive first and profits later. The difference this time is scale. The current spending cycle is unusually large, leaving companies exposed if adoption grows more slowly than the most optimistic projections assume.

AI Bet Across Microsoft, Google, Meta, and Oracle

Across the major players, the projected returns vary significantly.

Microsoft is estimated at -9.2%, placing it closest to breaking even. Alphabet follows at -15.7%, while Meta shows a deeper projected shortfall at -28.8%. Oracle sits further down the curve at -35.6%.

Amazon stands apart as the only firm to register a positive return, at roughly 7.2%, under the same assumptions.

The results show that AI is not affecting all big tech companies in the same way. Firms like Microsoft and Alphabet, which already have strong cloud businesses and large corporate customers, are closer to breaking even. They can plug AI into systems they already sell, which helps them earn money faster.

But other companies, like Meta and Oracle, are spending heavily upfront on AI infrastructure without seeing clear or immediate ways to make that money back. That creates a bigger gap between what they are spending and what they are earning.

Because of this, investors are no longer treating AI as one single 'winning bet' across the whole tech sector. Instead, each company is being judged separately based on how well it can turn AI into real income.

The focus now is on whether customer demand for AI will grow quickly enough to close these gaps, or whether companies are spending faster than the market is ready to pay for.

Is It Only Early-Stage Economics?

The main issue is not whether AI will become important, but when it will start making real money. Supporters of the big spending say this is normal for new technologies. They argue that companies usually spend heavily at the beginning, and profits only come later once the technology is widely used across industries and everyday products.

But critics are more cautious. They worry that companies are spending too much too quickly, betting on future demand that may not grow fast enough to justify the massive investment.

The analysis from Panmure Liberum suggests that the current AI boom is still mostly based on expectations about the future, rather than profits being made today.

Whether this turns out to be a successful long-term shift or an expensive overreach will depend on whether demand for AI grows fast enough to match all the money being poured into it right now.