Published on: 2026-07-31
Updated on: 2026-07-31
AI spending is producing opposite stock reactions across Big Tech because the market is judging the return, not simply the size of the bill. Amazon, Microsoft and Alphabet have gained support when cloud growth and operating profit accelerated, while Meta faced more pressure as costs rose faster than earnings.
Amazon’s latest quarter shows the dividing line, with AWS revenue up 37% and operating income up 64% even as expected 2026 capital spending rose to about $220 billion.
AI spending gains support when paying demand, revenue growth and operating profit rise together.
Amazon’s $220 billion spending plan drew less concern after AWS revenue grew 37% and operating income climbed 64%.
Infrastructure investment can reduce free cash flow long before the completed assets produce their full revenue.
Microsoft and Alphabet paired heavy spending with stronger cloud profit, while Meta’s operating income fell 8% as costs surged.
Cloud margins and free cash flow will show whether the current spending cycle is creating durable returns or only larger cost bases.
Four signals usually determine whether a large AI investment is received positively.
Paying demand is visible. Customers are already using the services, signing long-term commitments or waiting for additional capacity.
Revenue produces profit. Growth reaches operating income instead of being absorbed by depreciation, energy costs and price competition.
Cash pressure remains manageable. The existing business generates enough cash to fund expansion without steadily weakening financial flexibility.
Results exceed expectations. Strong growth can still disappoint when the market had already priced in an even better result.
A company does not need perfect results across every measure in one quarter. The strongest spending case appears when demand, revenue and profit improve before cash pressure becomes permanent.
The opposite pattern creates concern. Slower growth, weaker margins and rising capital expenditure suggest new infrastructure is arriving faster than customers can use it profitably.
Amazon reported $62.6 billion of net income, although $53.4 billion came from non-operating pre-tax gains, mainly related to its Anthropic investments. AWS provided a clearer picture of the quarter’s operating performance.
AWS revenue rose 37% year over year to $42.2 billion, its fastest growth in 18 quarters. Operating income increased from $10.2 billion to $16.6 billion, expanding the segment’s operating margin from 32.9% to 39.4%.
AWS generated about 60% of Amazon’s operating income while accounting for roughly one-fifth of revenue, making the cloud business the main reason investors accepted higher spending.
AWS’s AI business and custom-chip business each exceeded annualised revenue run rates of $25 billion.
Amazon said available infrastructure remained below expected customer demand.
Anthropic and OpenAI had made multi-year commitments linked to Amazon’s infrastructure and Trainium chips.
These figures gave Amazon room to increase spending because paying usage and contracted demand were already visible. The $220 billion plan also covers custom chips, fulfilment assets, robotics and satellite technology, so it should not be described as an entirely AI-focused budget.
AI spending creates a timing gap: companies pay for infrastructure before they earn revenue from it.
When companies build AI capacity:
Cash leaves immediately: Companies spend billions on chips, servers, data centres and power infrastructure before customers use the capacity.
Revenue arrives later: Customers only generate revenue after workloads move onto the new infrastructure.
Costs continue after launch: Depreciation, electricity, maintenance and staffing expenses increase as the infrastructure comes online.
Accounting makes the gap larger. Infrastructure purchases reduce cash flow upfront, while their costs are spread across years through depreciation. This allows operating profit to remain strong even when free cash flow declines.
Amazon generated $161.4 billion of operating cash flow over the 12 months through June, up 33%. However, property and equipment purchases increased by $66.1 billion from the previous year, pushing trailing free cash flow from a positive $18.2 billion to a $7.6 billion outflow.
One negative free-cash-flow period does not prove AI spending is failing. The concern would be slower AWS growth, weaker margins and continued cash pressure without enough revenue growth to justify the investment.
The table shows why similar AI spending plans can produce different market reactions. The figures to watch are the growth and profit created by the investment, followed by the cash pressure required to support it.
Company |
Evidence of payback |
Financial pressure |
Amazon |
AWS revenue +37%; AWS profit +64% |
TTM free cash flow −$7.6B |
Microsoft |
Azure +43%; cloud profit +31% |
Property additions $35.8B |
Alphabet |
Cloud +82%; margin 35.6% |
Q2 free cash flow −$5.9B |
Meta |
Revenue +28% |
Profit −8%; free cash flow $784M |
Microsoft’s Azure and other cloud-services revenue grew 43%, while Intelligent Cloud operating income increased 31%. Alphabet’s Google Cloud revenue rose 82%, with operating income more than tripling to $8.8 billion.
Meta showed the more difficult combination. Revenue grew 28%, though costs and expenses increased 55%, operating income fell 8%, and free cash flow dropped to $784 million. Legal charges and severance also contributed to the cost increase, so AI investment was only part of the pressure.
Amazon, Microsoft and Alphabet paired infrastructure expansion with stronger cloud profitability. Meta’s results showed the risk when higher costs reach the financial statements before an equivalent increase in earnings appears.
The companies report their cloud operations differently, so the percentages do not create a direct ranking. They still reveal whether commercial progress is keeping pace with the spending programme.
The next stage of the AI buildout will be judged through margins and cash generation rather than spending announcements.
Three signals will show whether the buildout remains productive:
Cloud growth remains firm as more capacity opens.
Segment margins withstand higher depreciation and operating costs.
Free cash flow begins recovering as completed infrastructure produces revenue.
Backlogs and long-term contracts support future demand, though they only become revenue when customers begin using the capacity. Delays in construction, power shortages or slower usage could push returns further into the future.
Amazon’s AWS margin and free-cash-flow outflow provide one version of this test. Microsoft, Alphabet and Meta face the same pressure through different business models. Each must show that the previous round of investment can help finance the next one.
The market responds more positively when spending is accompanied by faster revenue growth, stronger operating profit and visible customer demand. A large budget becomes easier to support when the assets already in service are producing commercial returns.
Not necessarily. Infrastructure purchases reduce free cash flow before the completed assets produce their full revenue. The warning becomes stronger when negative free cash flow continues alongside slower growth and weaker margins.
The most useful measures are cloud or AI-linked revenue growth, segment operating income, operating margin, capital expenditure and free cash flow. Customer commitments and capacity utilisation provide additional evidence of future demand.
No. Amazon reports AWS as a separate segment, Microsoft reports Azure and other cloud-services growth, and Alphabet’s Google Cloud segment covers a broader collection of products. The figures are better suited to tracking each company’s direction than ranking them against one another.
Several quarters of slower cloud growth, falling margins, rising depreciation and persistent free-cash-flow pressure would provide the clearest warning. That combination would indicate that capacity was expanding faster than profitable customer demand.
AI is pushing Big Tech away from its traditionally asset-light economics and towards an infrastructure model that requires constant reinvestment. That shift raises the cost of being wrong because underused data centres still carry depreciation, electricity and maintenance expenses. The advantage will belong to companies that can finance the next round of capacity with cash generated by the previous one.
