PwC's Mid-Year CEO Check-In: AI's Payoff Is Steady, Its Use Case Is Shifting

PwC’s mid-year “Snapshot” follow-up to its 29th Global CEO Survey re-surveyed 351 of the roughly 4,500 CEOs who answered the original late-2025 fieldwork, checking in between May 15 and June 22, 2026 on how a volatile first half shifted their outlook. Revenue confidence held up: the share very or extremely confident in 12-month growth rose from 39% to 42%, and three-year confidence rose from 46% to 51% — though the aggregate hides real churn, since nearly 60% of returning CEOs reported a changed confidence level, split 33% up and 26% down.

On AI specifically, the picture is stable rather than improving: 39% of CEOs say AI’s impact on revenue or costs stayed positive or got better since late 2025, against 16% who say it stayed negative or got worse. What’s shifting is how CEOs are pointing AI: 38% say their company has used it since January 2026 to identify new business opportunities created by disruption, versus just 23% who used it to anticipate the effects of shocks before they hit — CEOs are reaching for AI as an offense tool more than a defense one, even as roughly 70% report rising energy and non-energy costs this year.

That offense lean sits next to a cost discipline problem BCG flagged back in July: its “Return on AI” research warned that as metered token pricing replaces flat AI subscriptions, the per-token spread between a simple model and a frontier model can run 5 to 25 times, and that costs scale roughly with the square of a session’s length — meaning a session twice as long can cost four times as much. BCG also cited IDC’s estimate that the top 1,000 global companies will underestimate their AI infrastructure costs by as much as 30% through 2027. Read together, the two findings frame the same tension: CEOs are increasingly comfortable pointing AI at growth, but the cost side of that bet is still catching most of them off guard.