The median public SaaS company traded at 3.1x ARR in June, the lowest reading since August 2011, and was back to 4.6x by the end of August. That is the backdrop for what we see on the desk: either you’re a great company or you’re a terrible company and that’s completely due to AI risk.

Of the public SaaS companies Aventis tracks, only 15% finished August above the Rule of 40 (revenue growth plus profit margin of 40% or more), and the median one came up well short with 11.8% growth and a 14.6% margin.

Veeva returned -2.8% from 12 months ago to a month ago while the median US stock did +15.8%, putting it 18.6 points behind. Even Moody's, the closest of these, is 8.1 points back.

80% gross retention used to be good enough for 5 or 6x ARR, and now it gets 1x to 3x if you can find a buyer at all. It doesn't measure up in SaaS Capital's 2025 survey either, where the median private B2B SaaS company holds 91% and it takes at least 90% just to keep up with peers.

If the team is good, you can get away with a lot fewer people. Cursor went past $1 billion of annualized revenue with over 300 people, and Lovable is packing $400 million of ARR with 146.

One realistic move for a big bloated company is to pull out of hiring. In Gartner's 2026 CEO survey, 88% plan to put more into AI and only 32% plan to hire more.

The Trade Desk left the S&P 500 on Sep 21 after a 73.8% drop over 12 months, and it's now a $6B company that still produced $1.1B of operating cash in the last year. The index goes by size.

There's more action in SaaS M&A than Software Equity Group has ever tracked, 2,784 deals in the twelve months through June, and the public SaaS median came down every quarter, from 5.7x revenue to 3.2x.

You don't need to pay for a class on AI, because the best way to learn it is by doing it and the AI walks you through the rest. Microsoft's 2024 survey of 31,000 knowledge workers had 75% using generative AI at work and only 39% with any AI training from their employer.

AI companies made up 42% of the value on Bessemer's 2025 Cloud 100, up from 21% the year before, and they average a stiffer multiple, 24x against 19x for everyone else.

Sezzle needs a 61.0% rise just to get back to its close before the August drop, and even Edison International, the shortest trip on the chart, needs 43.7%.

I can already replicate a lot of the junior-level job with AI, so there's going to be less and less of that work to go around, and that scares me. Stanford's payroll data shows employment down about 11% since late 2022 for 22 to 25 year olds in the jobs most exposed to AI, and up about 10% for the same ages everywhere else, with employers pulling out of hiring rather than firing.

Multiples at the high end of software have made a full comeback. Meritech has the top ten public names at 17.7x ARR, 6% above their pre-ZIRP median, while the overall median hangs at 4.2x, 52% below its own.

Nvidia's data center revenue went from $10.3 billion in its fiscal Q2 2024 to $89 billion three years later. I think the comedown starts once companies route each task to the cheapest model that can handle it, model efficiency gains land, and everyone gets bored of making so many dumb apps.

Use AI as much as you possibly can, at work and in your personal life, until you have a real understanding of what it can and can't do. Only 40% of companies in MIT's NANDA survey had bought an official LLM subscription, while workers at more than 90% of them were regularly using personal AI tools for work anyway.

The median private SaaS deal went from 5.4x revenue or more five years running through 2022 to 3.1x in the first quarter of this year. A company with 80% gross retention, 10% growth and 10% EBITDA margins used to trade at 5 or 6x ARR, and today it goes for 1x to 3x, if you can find a buyer at all.

Simple software by itself isn't a moat anymore, because if it can be vibe-coded the threshold for somebody else to come in is a lot lower. Sundar Pichai has gone from saying more than a quarter of Google's new code was AI-generated in late 2024 to 75%, approved by engineers, in April.

AI lets one person stay on top of a lot more people, because call transcripts, email and messaging get summarized and go straight to the CEO instead of passing through several people on the way up. Managers already average 12.1 direct reports in Gallup's 2025 numbers, up from 10.9 in 2024.

Software is barbelling, and the top end keeps getting pricier because the same supply of software capital is chasing a very limited number of really good companies. In public comps, the top five names went from 22.5x to 36.4x revenue since December, while the low-growth names went from 3.8x to 3.5x.