This earnings season has crystallised a stark dilemma for Silicon Valley: spend big on AI infrastructure now and hope revenue follows years later, or conserve capital and risk falling behind. Amazon, Alphabet, Microsoft and Meta together outlined plans to spend roughly $660 billion in 2026 on data centres, custom chips and other AI infrastructure — a jump of about 60% from 2025 and nearly three times the outlays of 2024. The scale of those commitments has unnerved investors even as the companies report robust cloud revenue growth.
Markets reacted sharply. The three cloud-heavy names that released quarterly results last week — Amazon, Google and Microsoft — lost nearly $900 billion of combined market value in the immediate aftermath. Amazon’s announcement that it will spend about $200 billion this year, $50 billion above expectations, sent its share price tumbling more than 10% as management defended the move as necessary to position the company across chips, robotics and satellites as well as AI. Microsoft’s shares have fallen the most, down roughly 18% since its results, after data-centre spending surged 66% and some cloud metrics disappointed versus optimistic forecasts.
Investors’ anxiety has two clear veins. First is timing: the greater the capital intensity, the longer the lag before that spending can generate acceptable returns. Second is concentration risk: Microsoft disclosed that roughly 45% of a roughly $625 billion cloud contract backlog is tied to a single partner, intensifying concerns that a lot of projected revenue is dependent on the fortunes of a few AI startups. Even Alphabet’s record revenues and profits could not offset investor unease when it said it planned to nearly double capital spending to about $185 billion.
Apple stands out as the contrarian case. By choosing to avoid large incremental infrastructure builds and relying instead on partnerships — including an arrangement to use Google’s Gemini models to bolster Siri and device AI — Apple reported a blow-out quarter with record revenue and a share-price gain of about 7.5%. Its modest capital expenditure (about $12 billion for the year) contrasts sharply with the peer group’s headline-grabbing budgets and highlights an alternative, leaner approach to accessing advanced AI capabilities.
The market rout is reshaping supplier and competitor expectations. Chipmakers and cloud infrastructure vendors are under scrutiny: rumours that a headline investment-and-infrastructure deal between OpenAI and Nvidia worth some $100 billion failed to materialise fed the sell-off, and companies perceived as heavily reliant on AI startups — Oracle among them — saw pronounced share weakness despite debt raises intended to shore up capacity. Software companies are nervous too, facing potential disruption from new AI coding tools from players such as Anthropic and OpenAI.
The broader question is whether investors’ patience for a protracted AI payback window has run out. Combined, these firms still delivered solid top-line growth — aggregate revenue rose about 14% to roughly $1.6 trillion — but rising capex and stretched free cash flow suggest a more capital-constrained era ahead. If AI requires continuing rounds of infrastructure spending before profitability catches up, markets may demand clearer evidence of near-term monetisation or else penalise valuations further.
For policymakers and industry watchers the capex arms race matters beyond quarterly returns. It determines where compute capacity is concentrated, who captures the economic upside of advanced AI, and how quickly the technology diffuses across sectors and geographies. A sustained build-out will favour large cloud incumbents and chip suppliers, but it also raises systemic risks: overinvestment that fails to deliver commensurate revenue, tighter competition for skilled labour, and geopolitical frictions around critical components and data flows.
Ultimately, this episode signals a shift: the era of lavish, open-ended AI spending is colliding with more discerning capital markets. Companies will need to show either clearer pathways to revenue from AI investments or demonstrate alternative, lower-capex routes to innovation if they want to restore investor confidence without ceding technological leadership.