$285 Billion Vanished From SaaS in 48 Hours. The Pricing Model Was Always the Bug.
In the first week of February 2026, software stocks had the worst 48 hours in the sector's history. Roughly $285 billion in SaaS market capitalization evaporated. By the time the drawdown finished working through the market, JP Morgan called it the largest non-recessionary software selloff in more than thirty years — around $2 trillion in value, with software's weight in the S&P 500 falling from 12% to 8.4%.
The proximate cause was a product launch. Anthropic shipped Claude Cowork, and investors watched an AI system read files, organize folders, draft documents in Word, build models in Excel, run multi-source research, and handle compliance checks — tasks that had each justified a separate seat on a separate platform.
The market did not reprice software because AI got good. It repriced software because it finally understood what it had been buying.
Seats were a proxy, and the proxy broke
Per-seat pricing was never a theory of value. It was a convenient correlation.
For twenty-five years, the number of people at a company who needed to touch a system was a decent stand-in for how much value that system created. More salespeople meant more CRM value. More engineers meant more observability value. Headcount was observable, countable, and easy to put in a contract, so the industry priced on it and stopped thinking about it.
The correlation held right up until output stopped being a function of headcount. When one person directing agents produces what six people produced in 2024, a vendor charging per seat watches revenue fall while the value it delivers stays flat or rises. The pricing model is now actively working against the vendor — and every customer CFO can see it.
That is a structural break, not a cyclical one. You cannot discount your way out of it.
The industry already knew
The reaction was fast, because most operators had seen this coming and were waiting for cover to move.
Seat-based pricing fell from 21% to 15% of SaaS companies in twelve months. Hybrid models — a platform fee plus consumption — went from 27% to 41% in the same window. A survey of 300 SaaS CEOs conducted in April 2026, two months after the selloff, found 97% planning to retire seat-based pricing within two years.
Production examples are no longer hypothetical. Salesforce prices Agentforce at roughly $2 per conversation. Intercom's Fin charges about $0.99 per resolution — not per agent seat, not per contact, per problem actually solved.
Note what those two have in common: a countable unit tied to a thing the customer wanted to happen.
Outcome pricing is harder than it sounds
The temptation, reading the above, is to announce outcome-based pricing next quarter. We would counsel against moving that fast, for three reasons we have hit directly in our own portfolio.
You have to define the outcome, and the customer has to agree. "Resolution" sounds objective until you litigate what counts. The ticket closed — did the problem recur in nine days? The lead converted — would it have converted anyway? Every outcome metric has an attribution argument inside it, and you are now having that argument with the person who signs your invoice, every month.
Your COGS became variable and your revenue became lumpy. Under seats, you collected predictable revenue against roughly fixed infrastructure cost. Under outcome pricing, a heavy-usage customer can be gross-margin negative while a light one subsidizes them, and you will not know which is which until the quarter closes. You need per-customer unit economics instrumented before you change the price list, not after.
Forecasting gets genuinely harder. Seats renew. Outcomes fluctuate with your customer's business, which means you have imported their seasonality, their layoffs, and their demand shocks onto your own revenue line. Boards do not love this.
This is why hybrid is winning rather than pure outcome pricing. A platform fee covers the fixed cost of being available and makes revenue forecastable; consumption or outcome charges capture upside and keep pricing honest as the customer grows. The 27%-to-41% shift toward hybrid is the market finding that equilibrium in real time.
How we are pricing in the studio
We build and operate companies rather than advising from outside, so this stopped being an opinion and became a set of decisions we had to make.
Our working rules, as of this quarter:
- Price the unit the customer would describe to their boss. If they would say "it handled 4,000 tickets," price tickets. If they would say "it closed the books three days early," find a way to price the close. Internal metrics like API calls or tokens are a cost structure, not a value story.
- Instrument before you price. Ship with full per-account usage and margin telemetry for at least two quarters before committing to a public price. You are buying information about your own cost curve.
- Put a floor under it. A platform fee that covers infrastructure and support means a quiet month does not threaten the business. Nearly every durable model we see has one.
- Make the meter legible. Customers tolerate variable bills they can predict and audit. They churn over bills they cannot explain. A usage dashboard is a retention feature.
- Keep the escape hatch. Early contracts should let both sides revisit the unit after a year. You will be wrong about the first unit you pick. Everyone is.
The uncomfortable read
The February selloff gets discussed as a shock. It was closer to a correction of a mispricing that had been sitting in plain sight.
Software's value was never the login. It was the work that got done. For two decades, charging for the login was a serviceable approximation, and the industry built its entire financial architecture — net revenue retention, seat expansion, land-and-expand — on top of the approximation rather than the thing itself.
AI did not destroy that value. It severed the proxy. The companies in trouble are the ones whose product genuinely was the login. The companies that will be fine are the ones that can point at work that happened and charge for it.
If you are building now, you have an advantage the incumbents do not: no installed base of per-seat contracts to protect, and no board that needs seat expansion to hit a number. You get to price the work from the first invoice.
Do not waste that.
