AI Capex Splurge Burns Billions in Big Tech Cash — and Investors Are Buying In

How AI Infrastructure Spending Dragged Four Big Tech Cash Piles

The costs of the artificial-intelligence buildout are now visible in the cash flows of the world's biggest technology companies. Alphabet, owner of Google, ended the second quarter with negative free cash flow of $5.9 billion, its first cash burn since going public in 2004. Amazon reported negative free cash flow of $7.6 billion, its first in three years. Meta Platforms saw its cash reserves fall 91% year on year to $784 million by the end of June, and Microsoft closed its fiscal fourth quarter with $19.6 billion in cash, down 23% from a year earlier.

The spending is not going unnoticed, but the market's response has been more nuanced than simple alarm. Cloud computing revenue is growing quickly enough that analysts interviewed by Valor consider the outlays justified. Amazon Web Services, the market leader, generated $42.2 billion in the second quarter, up 36.7% year on year. Microsoft's Azure grew 43% in its fiscal fourth quarter, pushing its cloud division past $100 billion in annual revenue. Google Cloud posted the fastest jump, with $24.8 billion in second-quarter sales, up 82%.

Amazon's chief executive, Andy Jassy, laid out the reasoning in an analyst call on July 30. The company now expects to spend $220 billion on infrastructure this year, up from a prior $200 billion forecast, and still expects capacity to fall short of 2026 demand. Jassy said data centers can generate more than 30 years of monetization without additional upfront capital, that servers and network equipment break even in less than three years, and that cloud clients typically commit to AI capacity for at least five years.

Investors rewarded the argument selectively. Amazon shares rose 15.32% on July 31, Alphabet gained 6.88%, and Microsoft added 3.02% after jumping more than 15% following its earnings. Meta, by contrast, saw its stock fall more than 10% after reporting earnings per share below estimates and raising its capital-expenditure outlook. Apple, which lacks a large-scale cloud business, lost 7.35% in the same week after missing growth expectations in China and services, even after briefly surpassing a $5 trillion market value.

Why the Market Is Backing Amazon, Microsoft and Google's AI Bets — and Punishing Meta and Apple

The Hyperscaler Math Behind the Billion-Dollar Burn

The central question for investors is whether the cash burn is an investment or a mistake. The argument in favor rests on three mechanics laid out by Jassy and echoed by analysts: equipment pays for itself in under three years, GPUs can stay productive for roughly a decade, and customers sign cloud and AI contracts lasting five years or more. That means the cash outflow now is expected to produce revenue for years after the infrastructure is built. José Medeiros, a partner at São Pedro Capital, argues the logic applies across hyperscalers, not just Amazon.

Why AWS, Azure and Google Cloud Are Backing Up the Optimism

The revenue data gives the thesis some support. AWS is still the largest cloud provider, but its 36.7% second-quarter growth is slower than Microsoft's 43% Azure growth and far behind Google Cloud's 82% jump. Itaú BBA analyst Stephano Gabriel notes Google Cloud accelerated from 40% growth in late 2025 to more than 80% now, and says the segment's backlog could push growth above 100% by year-end. Legacy Capital's Letícia Tonholo expects AWS's artificial-intelligence revenue, about $25 billion in the second quarter, to double by the end of the year. These are projections, not company forecasts, but they explain why analysts believe demand is real.

Meta Remains the Investor Skeptic's Target

Meta's response to its results was the toughest among the five companies. Its free cash flow fell 91% to $784 million, and it raised its annual capex guidance to between $130 billion and $145 billion. The quarter included $2.4 billion in litigation costs and $1.18 billion in severance charges, and the company cut about 8,000 jobs in May. Analysts see a different story than Meta's 2022 metaverse spending, pointing to plans to sell excess cloud capacity and to AI tools that support ad generation. The market, however, is not yet giving Meta the same credit it gives Amazon and Microsoft.

Apple Is the Surprising Loser of the AI Spending Cycle

The capex boom has also shifted the competitive backdrop for Apple. The iPhone maker has not built a hyperscale cloud business, so it kept positive cash flow, but its share price fell 7.35% in the same week after services and China revenue missed analyst estimates. Memory-chip inflation adds pressure to margins, and Itaú BBA's Gabriel notes that chip suppliers are prioritizing data-center customers over device makers. In a market now focused on AI capacity, Apple's capital-light position has gone from a financial strength to a strategic question.

Where the AI Capex Cycle Leaves Investors, Suppliers and Enterprises

For investors and technology buyers, the earnings season sets concrete markers to track rather than a single verdict on AI spending.

• Watch the growth gap among cloud platforms. AWS grew 36.7% in the second quarter, Azure 43% and Google Cloud 82%; if Google Cloud's backlog pushes growth past 100%, its share of the AI infrastructure market will be difficult for rivals to ignore.
• Treat Amazon's capex guidance as a demand signal. Jassy raised 2026 spending from $200 billion to $220 billion and said capacity will still not cover 2026 demand; a cut in that figure would be a more meaningful warning than any management commentary.
• Judge Meta on execution, not promises. Its cash reserves fell 91% to $784 million against a $130-145 billion capex plan, with $2.4 billion in litigation costs and $1.18 billion in severance; revenue from cloud and AI-driven advertising needs to appear in upcoming quarters to justify the outlay.
• Track Apple's services and China growth next quarter, along with gross margin, because memory-chip inflation and supplier priority for data centers are the two named pressures on its profitability.
• Enterprises planning large AI workloads should expect tight capacity and multi-year commitments, since AWS clients are already signing at least five-year AI contracts and providers say they cannot meet demand.