All news

Cloudflare's 20% cut makes headcount an AI design choice

A long white desk with a row of graphite chairs and clear gaps where chairs were removed

Cloudflare cut more than 1,100 jobs in a record quarter. Software companies now plan staff around agents, and that changes what investors pay for.

More than 1,100 people, about 20% of Cloudflare's staff, will leave the company as it moves to what it calls an agentic AI-first operating model. The cut, announced on 7 May, came with a record quarter: first-quarter revenue rose 34% year over year to $639.8 million.

A software company growing at 34% cut a fifth of its staff in a strong quarter and tied the decision to agents, not to a downturn. That breaks the old link between revenue growth and hiring, and it changes the number capital pays for. Headcount growth stops working as a proxy for momentum. Revenue per employee takes its place, along with the share of the cost base that shifts from salaries to models and compute. Boards across public and late-stage private software will ask the same question within quarters.

The internal numbers explain the confidence. The founders said internal AI usage grew more than 600% in three months, with employees from engineering to finance running thousands of agent sessions a day, the same tools now cutting into software seats. The company kept the people closest to revenue, sparing salespeople who carry quotas, and it expects $140 million to $150 million in charges, most of them this quarter. That is a one-off price for a smaller payroll.

The savings do not vanish. Part of them turns into spending on inference, already straining hyperscaler capacity, and part into the tooling that lets agents do real work inside a company. Incumbents in the middle of this redesign need orchestration, permissions, audit and cost control for thousands of internal agent sessions a day, built for finance, HR and legal teams as much as for engineers. Founders who sell that layer meet a buyer with a budget already freed up.

The same move raises the bar for young companies. An AI-native startup was supposed to win on a lean cost structure. When an incumbent with more than 5,000 people restructures around agents, that edge narrows, and the moat has to come from product and data rather than from a small payroll. Plan hiring around what agents cannot yet do, and put revenue per employee in every board deck.

Cutting staff is easy to announce and harder to turn into profit. Cloudflare's operating margin over the next two quarters will show how much of the saving reaches profit and how much moves into compute. Revenue growth is the sharper test: if it holds near 34% with a fifth fewer people, the agentic model works. The next earnings cycle will show how many other software companies announce similar plans, and whether they also protect quota-carrying sales staff.

Fanaura looks for companies that make this shift safe and auditable, and for founders building from day one to grow revenue far faster than headcount. The first large, measurable payoff from AI inside companies is showing up in the cost base, not only in new products.

For founders

Building the controls that let a company run thousands of internal agents safely? Send us the project.

Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.

Apply in five minutes
Subscribe

Get the news by email

One short note each weekday. No marketing, and you can unsubscribe at any time.