NextFin News - A fresh wave of selling hit the world’s largest technology companies on Monday as investors grew increasingly wary of the staggering capital requirements needed to sustain the artificial intelligence race. Shares of Microsoft, Alphabet, and Meta Platforms all retreated in late March trading following reports of massive new data center commitments that suggest the "hyperscale" giants are now effectively subsidizing the infrastructure for AI startups like OpenAI and Anthropic.
The market reaction follows a series of disclosures indicating that the collective capital expenditure (capex) for the "Big Four"—including Amazon—is projected to reach approximately $650 billion in 2026. This represents a 60% increase from the $410 billion spent in 2025 and a nearly threefold jump from 2024 levels. According to CNBC’s MacKenzie Sigalos, the deepening unease on Wall Street stems from the realization that even the most well-funded AI labs cannot support these buildouts on their own, forcing the cloud providers to absorb the financial risk of multi-billion dollar data center deals.
Alphabet has signaled a particularly aggressive trajectory, with 2026 capex projected between $175 billion and $185 billion, nearly doubling its 2025 spend. While Microsoft reported a 60% jump in quarterly profit earlier this year, Chief Financial Officer Amy Hood has maintained that demand for AI computing power continues to outstrip supply, necessitating relentless investment. However, the sheer scale of these outlays is beginning to test the patience of institutional investors who are looking for more immediate evidence of AI-driven revenue growth to justify the shrinking margins.
The current skepticism is not yet a consensus view, but it marks a significant shift in tone from the unbridled optimism of 2025. Analysts at S&P Global Market Intelligence noted that while the infrastructure being built today could generate returns for decades, the immediate impact is a massive drain on free cash flow. For Meta Platforms, which has pivoted its entire business model around "sovereign AI" and integrated Llama models, the risk is particularly acute if advertising revenue fails to scale at the same pace as server depreciation costs.
The tension in the market reflects a fundamental disagreement over the "payback period" for AI. Bulls argue that the current spending is a prerequisite for capturing a multi-trillion dollar market, while skeptics point to the historical precedent of the fiber-optic overbuild in the late 1990s. For now, the "hyperscalers" are trapped in a prisoner’s dilemma: they cannot afford to stop spending and risk losing the AI arms race, but they cannot continue spending at this velocity without eventually triggering a more severe re-rating of their stock prices.
Explore more exclusive insights at nextfin.ai.
