Daniel Nadler runs a company worth $12 billion. It has fewer than 100 employees. This month, at Nvidia’s GTC 2026 summit, he said the quiet part out loud: “I think the world’s not prepared for that.”

Nadler’s company, OpenEvidence, will be used by a doctor treating 300 million Americans this year. His claim isn’t that companies should be small. It’s that the structural ceiling — the point where a company must hire to grow — has shifted in a way that most organisations haven’t absorbed yet.

The response from the founder community has been predictable: people are treating this as a headcount argument. Cut staff. Run lean. Race to the smallest org chart. That’s the wrong reading, and acting on it will hurt you.

What the thesis actually says

The sub-100 thesis isn’t about cost reduction. It’s about org design.

Traditional companies scale by adding people to add capacity. More customers → more account managers. More product surface → more engineers. More markets → more local hires. The org chart grows because the work grows, and humans are the unit of work.

AI-native companies are breaking that relationship. Cursor has $2 billion in annualised revenue with around 300 people. Midjourney did $500 million in revenue with roughly 100. These aren’t lean startups grinding through adversity — they’re companies where the ratio of output to headcount is structurally different from anything that existed five years ago.

The lesson isn’t “hire fewer people.” The lesson is: if you’re designing your org the same way you would have in 2019, you’re building a company that’s already obsolete.

The headcount framing is a trap

Jack Dorsey announced 40% cuts at Block in February 2026 and framed it as an AI transformation. Darden Business School and Bloomberg both noted the suspicion of AI-washing — routine cost restructuring given a technological veneer.

Oxford Economics analysis found that of 1.2 million U.S. job cuts announced in 2025, AI was cited as the reason for only 55,000 — about 4.5%. Sixty percent of executives who made cuts did so in anticipation of AI efficiencies that hadn’t materialised yet.

Cutting headcount because you heard a compelling GTC talk is a one-time cost reduction dressed up as strategy. The market will see through it, your remaining team will feel it, and you’ll have made your organisation weaker without making it more capable.

The actual structural advantage

The reason small AI-native companies can outperform large traditional ones isn’t the size of the payroll. It’s the communication overhead.

Tomasz Tunguz ran the org chart math: a 150-person organisation has 11,175 potential communication channels. A 30-person team producing equivalent output has 435. That’s a 96% reduction in coordination drag — not from cutting costs, but from designing for leverage from the start.

PwC’s Global AI Jobs Barometer found that industries with high AI exposure show labour productivity growing 4.8x faster than the global average. That’s not about having fewer people. It’s about each person doing fundamentally different work.

The advantage comes from designing for this from day one — not from retrofitting it onto a bloated org structure after the fact.

What Vinod Khosla said that nobody quoted

While Nadler’s GTC comments were circulating, Vinod Khosla posted on X something more precise: “A few years ago I asked when will a hundred-employee company reach a billion dollars in revenue. We are getting close to 1,000 employees with a billion dollars of revenue! A 10x or even a 100x lower employee count.”

He’s not predicting mass layoffs. He’s describing a new benchmark for what a well-run company looks like. The target isn’t zero employees — it’s maximum output per person. Those are different objectives, and confusing them produces very different companies.

If you’re building now

If you’re pre-Series A: don’t build an org structure that assumes you’ll need 200 people to hit $10M ARR. Design for 20. If you get there faster and need to grow, you can — but most founders hire ahead of leverage and then spend three years managing complexity they created.

If you’re post-Series A: audit your org for where headcount adds capacity versus where it adds coordination. Not every role is replaceable by AI. Some roles exist because humans are still better at the work. The question worth asking is whether your headcount growth curve is proportional to your output curve — and if it isn’t, understanding why.

The sub-100 thesis isn’t a cost story. It’s a design challenge: can you build a company where the structure itself multiplies what each person can do, rather than just adding more people doing more things?

If you’re working through that question, we’ve helped a lot of early-stage teams get the architecture right — jfsi.io.

Sources: Fortune — $12B AI Startup Founder: Future Tech Giants Will Have Sub-100 Employees · TechCrunch — Cursor Has Reportedly Surpassed $2B in Annualised Revenue · Fortune — Block/Dorsey 40% Layoffs · Darden/UVA — Is AI the Strategy or the Scapegoat Behind Block’s 40% Layoff? · Bloomberg — Jack Dorsey’s 4,000 Job Cuts Arouse Suspicions of AI-Washing · Fortune — AI Layoffs: Convenient Corporate Fiction? · Vinod Khosla on X — 10 employees, billion in revenue · Tomasz Tunguz — Communication Tax in Small Orgs · PwC Global AI Jobs Barometer 2025