May 11, 2026

Agent training infrastructure scales across AI-native companies

Polsia hit $1 million ARR in 30 days, ElevenLabs reached $330 million ARR with 200 employees, and Anthropic commanded a $900 billion valuation discussion with 48-hour investor deadlines. The velocity asymmetry between AI-native companies and traditional scaling models has become unmeasurable.

This isn't incremental improvement. AI-native companies are achieving scale milestones in weeks that previously required years, while maintaining revenue-per-employee ratios 5-10x higher than the best traditional software companies. The efficiency gap has widened from a competitive advantage to a structural market bifurcation.

Polsia: $1M ARR in 30 days, $4.5M ARR in 60 days ElevenLabs: $330M ARR, ~$1.6M RPE, zero-to-$200M ARR in 3 years AI-native baseline: $500K ARR per employee (up from $200K) Top quartile AI-native: $3.5M RPE vs $610K traditional SaaS Anthropic valuation: $900B potential, 48-hour allocation window OpenAI scale: $25B ARR, 910M WAU, 9M business users

Secondary markets are adapting to this velocity by treating liquidity as infrastructure rather than exception. Carta's 396 tender offers in 2025 — up 62% with 99.9% subscription rates — signal that late-stage companies expect regular liquidity windows, not binary exit events.

The data reveals two parallel transformations: AI-native companies compressing development and scaling timelines, while secondary markets decompress the traditional binary exit model into continuous liquidity. Both trends serve the same economic reality — capital velocity must match operational velocity.

Source · Carta · SaaStr · Context Studios · Sacra · TechCrunch