AI Is Changing the Economics of Scale
AI-native companies are generating ten to twenty times the revenue per employee of traditional SaaS. Cloud reduced the capital needed to start a company. AI is reducing the capital needed to scale one.
For decades, starting a technology company required significant capital — for servers, office space, and to build large engineering and sales teams. The cloud era upended the first part of that equation. By using infrastructure from Amazon, Microsoft, or Google, founders no longer needed to buy and manage their own infrastructure. The cost to launch a tech company fell from tens of millions to hundreds of thousands of dollars.
Cloud reduced the capital needed to start a company.
AI is now reducing the capital needed to scale one.
The evidence
Cursor, an AI coding tool, reached $100 million in annual recurring revenue with approximately 20 people — averaging $5 million in revenue per employee.
Midjourney has generated over $200 million in revenue with roughly 40 people.
Replit grew from $2.8 million to $150 million in annualized revenue in under a year.
Vercel will do nearly $300 million in revenue with an inside sales team of two.
For context, in traditional SaaS companies, $300,000 in revenue per employee was considered a good benchmark. $600,000 was considered elite. These AI-native companies are generating ten to twenty times those benchmarks. We are seeing structural changes in the relationship between headcount and output.
Beyond software
The phenomenon extends well beyond software companies.
Klarna, the $80 billion fintech company, deployed an AI customer service assistant in early 2024 that handled 2.3 million conversations in its first month — doing the equivalent work of 700 full-time agents. Resolution times dropped from 11 minutes to under 2 minutes with no decline in customer satisfaction. Klarna subsequently reduced its workforce from 5,000 to 3,800 through attrition, with CEO Sebastian Siemiatkowski projecting the company could operate at equivalent output with 2,000 people.
The investment implication
If AI collapses the cost of building product and scaling operations, the barriers to entry in software are falling. More companies can reach $100 million in revenue with small teams and modest funding. This is unambiguously good for innovation and for entrepreneurs.
But these same forces are also putting pressure on the venture capital model. If AI is dramatically reducing the capital required to build and scale a company, then fewer venture dollars are needed — yet most venture firms remain organized around deploying large amounts of capital into staged rounds.
The companies that will define this cycle are the ones that have already crossed the threshold — companies with hundreds of millions in revenue, evangelical customer bases, and sustainable competitive moats. The paradox of AI-driven efficiency: it is easier than ever to build, but harder than ever to become the category-defining company that accrues the majority of value in a market.
The technology cycle framework
Every major platform cycle follows a predictable build sequence:
- Operating system — in this cycle: foundation models (Claude, ChatGPT, Gemini)
- Compute infrastructure — servers, networking, semiconductors, data centers
- Developer tools — coding assistants, deployment platforms, testing frameworks
- Applications — the products built on top of the new platform
We are looking across all of these layers, but with varying degrees of conviction. The compute infrastructure presents compelling economics but extreme valuations. Developer tools face the risk of platform absorption — the model providers themselves are shipping competing products. Enterprise applications remain the most durable opportunity, because they are so difficult to install, integrate, and replace.
The best risk-adjusted returns in this cycle will come from investing in scaled, durable companies that are already winning — not from betting on which early-stage startup will break through.