Bank of America Projects Significant Rise in AI Capital Expenditure by 2027

Bank of America has updated its forecasts for capital expenditure in the AI sector, following the Q2 earnings reports from major U.S. hyperscalers, including Alphabet, Microsoft, Amazon, and Meta. The bank now projects that hyperscaler capex will surpass $860 billion in 2026, representing an 80% year-over-year increase, with a trajectory towards $1.2 trillion in 2027, marking a further 38% rise.

The demand for compute resources remains high, with supply constraints persisting. Management teams across the sector have indicated a commitment to multi-year spending plans. Bank of America notes, "We see increased hyperscaler appetite to continue investing in capacity."

A critical figure in this landscape is the $2.3 trillion backlog of customer cloud commitments from the top four cloud providers, which has risen 16% quarter-over-quarter from $2.0 trillion in early July. Microsoft has reported a commercial backlog of $678 billion, while Oracle's backlog stands at $638 billion, supported by approximately $75 billion in prepaid hardware contracts, which mitigates funding risks.

Despite concerns regarding free cash flow, which Bank of America acknowledges may turn negative, with aggregate margins projected to decline to around -1% in 2026 and potentially bottoming out at -5% to -6% in 2027-28, AI sales are experiencing triple-digit growth. AWS has reported that AI and chip revenues have each surpassed $25 billion annually, with cloud and AI margins expanding ahead of expectations.

As capital expenditure rises, the market faces the critical question of whether this spending will yield substantial returns. The top five hyperscalers have raised approximately $270 billion since the beginning of 2026, primarily through long-dated debt instruments. Bank of America asserts that "capital market access remains favorable," suggesting that free cash flow could significantly improve post-buildout.

In the current market environment, sectors such as compute, memory, semiconductors, power semiconductors, and optics are experiencing increased demand. Notably, memory stocks have seen substantial gains in premarket trading, with the DRAM ETF up 4.4%, Micron up 5.2%, and Nvidia up 1.7%.