For the first time in the history of modern software, publicly listed SaaS companies trade at a discount to the S&P 500. Not during the financial crisis. Not during the 2022 rate hike cycle. Not during the dot-com crash. Now. And it is not a market overreaction.

The SaaStr analysis published this month puts numbers on the collapse. Software forward P/E was 84x in 2021, roughly four times the broader market. By early 2026 it had fallen to 22.7x — below the S&P 500 multiple. The IGV software ETF is down 21% year to date and 30% off its September 2025 peak. Over $2 trillion in market cap gone.

The market is telling you something specific. If you are building software, you should understand what.

The two numbers that explain it

AI-native enterprise spend grew 94% year on year in Q1 2026. Traditional SaaS grew 8%.

That is not a slowdown. That is a split. Enterprise IT budgets are not unlimited. When one category grows 94%, the money has to come from somewhere. Anthropic and OpenAI now account for roughly $40–50 billion in combined annual enterprise spend — that is 5–6% of the entire enterprise software market, redirected from traditional vendors to two companies in two years.

The median SaaS revenue growth rate fell from 21% in 2023 to 12.2% by Q4 2025. Deals are getting smaller, slower, and more contested. And the customers doing the negotiating have a new piece of leverage: they can point at an AI agent and ask why they are paying for seats that agents are replacing.

The seat compression problem

This is the structural issue the valuations are pricing in.

Per-seat pricing worked for thirty years because headcount was a reliable proxy for software usage. One salesperson, one Salesforce seat. Ten support agents, ten Zendesk licenses. The math held because humans are the unit of work.

AI agents break that assumption permanently. When a single agent can handle the work of ten human operators, organisations do not need ten seats. They need one, maybe zero. The share of seat-based revenue in enterprise software contracts has already fallen from 21% to 15% in twelve months. Gartner forecasts that 35% of point-product SaaS tools will be replaced by AI agents or absorbed into larger agent ecosystems by 2030.

Klarna announced in 2024 that AI had let them reduce headcount by roughly 700 people. That is 700 fewer seats across every SaaS tool those people touched. Salesforce watched this happen and pivoted Agentforce to outcome-based pricing — charging per agent conversation rather than per human seat. ServiceNow moved toward consumption models tied to workflow executions. These are not incremental adjustments. They are the two biggest SaaS companies in the world signalling that the model has a shelf life.

The wrong response to a right signal

Most founders building SaaS right now are watching the valuation compression and concluding that the market is being irrational. It is not.

The market is discounting future cash flows, and the discount rate on per-seat revenue has gone up sharply because the addressable seat count is projected to decline. Even if your product is excellent, your revenue model may be structurally exposed to a shrinking denominator. That is not a product problem. It is a pricing architecture problem.

The founders who are getting this right are designing for outcomes from day one. Intercom charges $0.99 per resolved conversation. HubSpot dropped to $0.50 per agent resolution in April 2026. These are companies that used to own per-seat contracts. They made the switch because the economics of the next era do not reward seat growth, they reward demonstrated value at the moment it is delivered.

Outcome-based pricing has one property that per-seat pricing does not: it survives a world where software executes work autonomously. If your product solves a problem regardless of whether a human or an agent is doing the triggering, outcome pricing lets you grow with AI adoption rather than despite it.

What this means if you are building

If you are pre-product or early in your architecture, the decision is straightforward. Do not build a per-seat model and plan to migrate later. The migration is painful, the contract renegotiations are brutal, and the valuation market has already shown you what it thinks of the end state.

Design around outcomes, consumption, or a hybrid that starts with a usage floor and grows with value delivered. Figure out the unit of value your product creates — a resolved ticket, a contract reviewed, a lead qualified, a workflow executed — and price against that unit.

If you are already running per-seat contracts, the signal from the market is not that your product is bad. It is that your pricing model is carrying a structural discount that will compound. The customers renewing today are making renewal decisions in a world where AI agents exist. The customers renewing in eighteen months will have data on which seats they can eliminate.

The per-seat model had a thirty-year run. Public markets have decided that run is ending. The founders building the next generation of enterprise software are pricing for the world that is actually here.

If you are building SaaS and figuring out your pricing architecture, talk to us.

Sources