Hyperscalers Are Raising AI Capex Again-Amazon's 37% Cloud Beat Says the Spending Isn't Dead Yet

The AI capex debate has moved from scale to monetization

The real question is no longer whether AI capex is huge. It is. The real question is whether that spending is converting into revenue fast enough to justify the bet.

Why investors should focus on returns, not just spend

The scale is now hard to ignore. The five biggest US cloud and AI infrastructure spenders are collectively planning roughly $660 billion to $690 billion in 2026 capex. As the buildout expands, the debate has shifted from who is building fastest to who is monetizing first. That matters because bigger capex does not automatically translate into better shareholder returns.

Amazon's latest result matters because it gives investors a fresh data point. AWS topped market expectations for quarterly cloud revenue growth, and management said AWS' annual AI revenue run rate had surpassed $15 billion. That does not fully validate the entire AI buildout, but it does show that at least some of the spending is meeting demand. That is the signal investors should keep watching: revenue traction, not just headline capex.

Amazon shows the demand story is still alive

Amazon matters because it turned an abstract capex debate into a revenue story. AWS delivered $42.2 billion in quarterly revenue and grew 37% year over year, well above expectations, while the unit now sits at a $169 billion annualized revenue run rate. That is not what stagnant demand looks like.

The monetization signal is broadening

Amazon also said both AI revenue and chips revenue are now running above $25 billion annually. That matters because it suggests AI demand is not limited to a narrow GPU layer. It is showing up across services, inference, and custom silicon.

Amazon's capacity constraint suggests demand is still ahead of supply

Amazon's own message was straightforward: even after raising guidance to $220 billion in capital spending this year, management said supply still would not meet demand into 2026 and 2027, while 2028 demand was already striking. That does not prove long-term returns will be strong, but it does support the case that capex is still being driven by demand rather than built in isolation.

Alphabet reinforces the cloud signal

Amazon is useful, but not sufficient. Alphabet said Cloud revenue grew 82% and Cloud backlog grew to $514 billion. Two major cloud platforms are pointing in the same direction: enterprise AI demand is showing up in both current revenue and contracted backlog.

What to watch next: - AWS AI and chips run rates holding above $25 billion - Alphabet backlog staying healthy after $514 billion - Evidence that capex hikes are chasing demand rather than outrunning it

If those signals persist, Amazon stops looking like a one-quarter exception and starts looking more like a broader benchmark.

The real risk is the funding stack, not demand itself

Demand can be real and the stock can still disappoint. The fragility is in how this build is being paid for.

Higher capex raises the return hurdle

Estimates have moved higher again. Analyst work now points to roughly $750 billion in 2026 capex for the top five hyperscalers, above the earlier $660 billion to $690 billion base. That raises the return hurdle for shareholders. The issue is no longer just whether AI demand exists, but whether returns can clear a much more expensive funding bar.

Project 2026 capex as a share of sales is projected as high as 86% for Oracle, 54% for Meta, 47% for Microsoft, 46% for Alphabet, and 25% for Amazon. Bulls can argue these are front-loaded platform investments. That is plausible. But at these levels, even a modest delay in monetization could weigh heavily on earnings quality.

External financing is becoming more important

The more significant shift is funding. Five hyperscalers moved from 9% of capex financed externally in FY24 to 32% by mid-2026, while Alphabet priced an $84.75 billion equity raise. That matters more than another capex beat. Self-funded buildouts leave more flexibility for shareholders. Debt adds interest pressure. Equity directly dilutes ownership.

What bears are likely to focus on next: - capex-to-sales staying extreme, especially at Oracle and Meta - external funding still running around 32% of capex - more bond or equity issuance after Alphabet's equity raise - free cash flow staying under pressure enough that leasing becomes central to the strategy

That is the real trap: solid demand, but a funding model that complicates shareholder returns.

Prefer monetization first, infrastructure second

The positioning call after the latest cloud prints is simple: favor monetization first, infrastructure second. Amazon's cloud growth topped expectations and Alphabet's Cloud backlog grew to $514 billion. That keeps the demand signal intact.

Where the cleaner exposure looks today

Platform names still have the cleaner repricing path because they are closer to billed demand. AWS is the clearest proof that AI spending is turning into real usage, not just construction headlines. Supplier exposure still has a case, but it is more indirect: the build matters to the recipients today even if final hyperscaler ROI is still being proven.

Earnings checklist

At the next earnings run, the most useful checks are: - whether capex is staying ahead of monetization - whether financing is becoming more external - whether cash receipts are improving alongside revenue

If capex keeps outrunning monetization and funding gets more external before shareholders see better cash returns, the thesis weakens. That is the early warning sign to watch.