Key points
- AI investment hit $377B in H1 2025, more than 2024's $363B, on track to double.
- Rule of 40 now predicts valuation best, with 34% R-squared versus 20% for growth alone.
- 94% of earnings calls mention AI, but only 59% of sampled SaaS companies have AI products, a gap Ray Rike flags as opportunity or issue.
Rule of 40 now beats growth at predicting SaaS valuations
Iconiq's State of Software report, built on 127 companies, says the SaaS market has stabilised rather than broken. Growth and profit now trade off in a predictable way, which is why the rule of 40 has become the best predictor of valuation and revenue growth alone has lost explanatory power. AI has not yet appeared in most of these companies' products, even as capital and hiring pour into it.
Takeaways
- Rule of 40 now predicts valuation best. Its R-squared against EV/revenue rose from 12% in early 2021 to 34%, while revenue growth alone fell from 49% to 20%. Growth still carries about 1.9 times the weight of free cash flow margin.
- AI talk far outruns actual AI products. 94% of earnings call transcripts mentioned AI, yet 59% of Iconiq's 127 companies are classed as non-AI, meaning no AI product or features. Ray Rike calls that gap an opportunity or a problem.
- SaaS multiples hold and growth still pays. Median forward EV/revenue for public SaaS companies is stable near 5, and companies growing above 20% sit at 12. Valuations remain below historical peaks, though earnings beat rates are recovering.
- AI capital is doubling, not flat. 377bn dollars went into AI in the first half of 2025 against 363bn for all of 2024. Dave Kellogg says the chart looks flat only because of that comparison, and the pace is on track to double.
- Efficiency benchmarks settle near familiar norms. Top-quartile CAC payback is 15.9 months, median burn multiple below 100m dollars of ARR is 1.7 times, the net magic number is about 1, and ARR per FTE runs from 200,000 to 330,000 dollars.
- AI startups scale ARR with tiny teams. Glean took three years and 1,000 people to reach 100m dollars of ARR, while Lovable took eight months with 45 people. Forward deployed engineer vacancies rose from 30 in February 2024 to 375.
- Benchmarks rest on shrinking samples. Top-quartile net dollar retention settled at 121% below 100m dollars of ARR and roughly 110% above it, but that chart now draws on 41 companies, down from 233 in 2020. The CAC payback chart fell from 142 to 21.
Reading Iconiq's benchmarks
- Iconiq flips between medians and top quartiles across its charts, which Dave Kellogg calls out. The two figures tell different stories about the same market.
- Ray Rike advises checking the footnotes on every chart, because the sample size and the segment change from page to page.
- Burn multiple segmentation is too coarse. Ray Rike says a 20m dollar company and an 80m dollar one burn at very different rates, so one median hides both.