Sequoia analyst warns of $3 trillion revenue gap in AI infrastructure

The massive capital expenditure by hyperscalers underpins the entire streaming technology stack, from encoding pipelines to recommendation engines. If the revenue generated by AI applications fails to meet the $3 trillion threshold, these essential technology partners may face significant margin compression or a market correction, leading to higher cloud costs for streaming platforms. Furthermore, the trend toward token efficiency reduces immediate revenue for infrastructure owners, even as they scale capacity. Industry leaders must watch for hyperscalers to potentially raise service fees or pivot their pricing models to protect their 2028 free-cash-flow targets.

The financial tension identified by Sequoia is reflected in the most recent quarterly reports from the 'Big Four' hyperscalers. Per the Financial Times in July 2026, combined capital expenditures for Amazon, Google, Microsoft, and Meta are projected to reach $725 billion for the current year alone, a 77% increase over 2025. Amazon has emerged as the most aggressive spender, guiding for approximately $200 billion in total outlay as it attempts to secure 5 gigawatts of compute capacity to support its internal workloads and Its partnership with Anthropic. This scale of investment has begun to strain balance sheets; per Forbes in June 2026, Amazon's trailing-twelve-month free cash flow plummeted to $1.2 billion in Q1 2026, a 95% year-over-year decline.

While infrastructure costs soar, the software layer is showing signs of extreme volatility. Anthropic reported an unprecedented revenue climb from $1 billion in late 2024 to an annualized $47 billion by May 2026, according to internal documents reviewed by MLQ.ai. However, this growth is being met with skepticism regarding long-term profitability. Auditor documents for OpenAI's anticipated IPO revealed that the company incurred $38.5 billion in losses on $13.07 billion in revenue for the 2025 fiscal year, as reported by Medium in June 2026. This disconnect between record-breaking top-line growth and widening operating losses suggests that even the most successful AI startups are currently struggling to outrun the costs of the hardware they require.