Hyperscalers, led by Google, Meta, Microsoft, and Amazon, are aiming for strong free-cash-flow growth by 2028, but any slower payback could trigger broader economic risks, including a possible S&P 500 correction or recession if cash-flow targets aren’t met.

Current major players in AI, such as Anthropic and OpenAI, report substantial ARR estimates—roughly $60 billion for Anthropic and between $13 billion and $20 billion for OpenAI—yet the gap to cover infrastructure outlays remains wide.

Risks include a shift toward cheaper open-weight models and lower token prices, which could undermine the profitability of large-scale token factories built by frontier labs, even as token efficiency improves (for example, the latest OpenAI model is about 54% more token-efficient for coding tasks).

If payoff timelines slip, market reactions could be severe due to concentrated investment and expectations among a few key players, with potential ripple effects for the broader economy and equities.

Sequoia partner David Cahn points to the payback gap—from Nvidia’s roughly $50 billion annual GPU revenue in 2023 to current higher required spend—highlighting the widening gap between infrastructure costs and potential AI product revenue.