AINEWS
Netflix co founder and Anthropic board member Reed Hastings tells CEOs to halve their workforce or use AI to attack new markets. The planning unit, he says, is AI capability, not headcount.
Hastings sat down on Semafor's "The CEO Signal" podcast this month and gave CEOs a binary: cut your workforce in half and stay in your current business, or use the new AI capacity to expand against competitors. Business Insider's re-report of the interview carried the quotes out to a wider audience. Wire copy will lead with the layoff line. The interview is about something else.
Hastings is not a neutral observer. He co-founded Netflix, left its board in June 2026, and sits on Anthropic's board of directors. When he talks about AI capability expanding rapidly, he is talking about a category his board seat has a commercial interest in. That is worth knowing. The signal is still worth taking seriously; the discount rate is the reader's call.
Hastings' Semafor interview makes a different point: the unit a CEO plans against is no longer headcount. It is AI capability. Hastings forecasts frontier models will get 32 times better in five years. If the pace continues, about 1,000 times better in ten years. Both figures are his prediction, not an industry consensus and not an Anthropic commitment. The relevant policy context, Anthropic's "Policy on the AI Exponential", is the company's own articulation of that same curve, not an independent validation of his numbers.
"Capability doubling every year" is the unit of account. "Doubling your workforce" is Hastings' shorthand for what that means in practice: each year, an effective team gets bigger without the headcount growing. The half-cut-or-expand binary follows. If your current team can already ship a year's worth of product, you can shrink to the people you need and stay in the same business. Or you can take the new capacity and attack adjacent markets that used to be too expensive to enter. The fork is real, and Hastings is not arguing that one branch is universally better.
He pushes back on the assumption that only large companies benefit. The cost gap, he says, is "more attitudinal than the size of the firm." A small, aggressive team can adopt AI faster than a slow incumbent, even if the bigger firm can write a larger initial check. He concedes the upfront compute bill still favors larger players today, and expects that gap to narrow as model usage costs fall. The competitive thesis he lands on is "the great act of creative destruction": companies attacking each other because the cost of making software, and soon most things, is collapsing.
Hastings' own habits underline the point. He told the podcast he uses AI "all the time" to learn domains he does not know, from ski lift mechanics to sewer systems for a Utah development he is involved with. That is what capability doubling looks like for a single executive: a year of subject-matter learning that used to require consultants, now absorbed in an afternoon.
For an operator, the practical move is this: stop budgeting against last year's headcount and start budgeting against next year's capability. The two numbers will diverge. The layoff question is downstream of that gap, not upstream of it. Hastings does not pretend the choice is painless, but he is clear that doing nothing is the losing branch. A planning frame that ignores capability doubling will under-resource the offensive branch, which is the branch where new market share actually comes from.
The next test of his thesis will come with the next round of Anthropic capability releases, which the company's responsible scaling policy treats as the gating event for new deployment tiers. Watch whether the "32 times" curve holds for the current model generation, and whether CEOs cite the interview when they cut or expand over the next four quarters.