SaaS Got Cheaper. AI Got Pricier. Coincidence?

Most coverage of this year's record tech M&A focuses on the size of the deals. The number I keep returning to is smaller and stranger. According to FE International's mid year report, enterprise SaaS went from 4.9x trailing twelve month revenue at the end of 2025 to 3.3x in the first quarter of 2026. A third of the multiple, gone in one quarter, at businesses that by every ordinary measure are healthy.

Enterprise SaaS multiples fell from 4.9x to 3.3x revenue in a single quarter.

Healthy is the important word. The same report shows revenue growth holding at 12.7% and EBITDA margins expanding from 20.0% to a projected 22.6%. Growth is fine, margins are improving, and the price still fell. When earnings hold and the multiple drops, the market is not marking down what these companies do today. It is marking down what it thinks they will be worth once something else does the same job.

The premium went somewhere else

It did not leave the market. It moved. The same report notes that AI and cybersecurity assets still command premium pricing, and that nearly half of strategic tech deals above $500 million cite AI as an explicit benefit. Global announced M&A reached a record $2.8 trillion in the first half, with technology leading at $649 billion, and megadeals above $5 billion now make up 48% of deal value. Buyers are spending more, on fewer and larger bets, and they are paying up for AI while paying less for software.

Nearly half of strategic tech deals above $500 million now cite AI as a reason to buy.

Here is my working theory, and I want to be clear it is a theory. A lot of SaaS is a set of screens and workflows that let a person do a job, sold per seat. If AI can do more of that job directly, the screen matters less, and so does the seat. Investors do not need that to be fully true today to start pricing it in. They only need to be unsure that it is false. Software that hosts the work gets a lower multiple. Intelligence that does the work gets a higher one.

Pricier as an asset, cheaper as a service

There is a wrinkle worth sitting with. As an asset, AI is getting more expensive to buy. As a service, it is getting cheaper to use: this week's Enterprise Times roundup describes a price war between the major model providers, with the cheapest tier reportedly at $0.10 per million tokens. Falling usage prices are exactly what makes it easier for a customer to replace a software subscription with an AI workflow. So the thing investors are paying a premium for is also the thing making the incumbent products easier to leave. That is not a coincidence, and it is not obviously a bubble either. It is a repricing with a plausible mechanism behind it.

The AI feature bolted onto a SaaS product does not automatically sit on the premium side of that line. A premium tends to attach to something specific: data nobody else has, a workflow that gets better as the model improves, customers who would not leave. Buyers seem to be getting better at telling those apart.

What it does to a CFO's spreadsheet

I look at this from two seats. As a seller, or an investor in one, a lower multiple changes what a fair price looks like and how long the runway needs to be. As a buyer of software, the question is quieter: how many of the SaaS contracts on my books are priced per seat for work that something else is already doing? And can I name which of my critical processes now run through an AI layer inside a tool I already pay for, sometimes switched on by the vendor without much fanfare? If I cannot, I am carrying an exposure that neither I nor the vendor has priced. I wrote about the same trap from the deal room in My Gut Says This M&A Boom Is Fine. That Is Exactly Why I Do Not Trust It.: a big market signal makes it tempting to stop asking boring questions.

Once software does the work rather than hosting it, the questions change. Who is accountable for a decision the tool made? What data did it touch? Can the vendor show me, rather than tell me, what its AI features do? Those are diligence questions for an acquirer and procurement questions for a customer, and they look remarkably alike. That is the reason I keep coming back to insAIght: one register of the AI systems an organization actually runs, including the ones that arrived through a SaaS renewal, with an owner and evidence attached to each. For a fast first read on an AI tool already in use, ExplAIn is built for that.

The question I cannot shake

If the market is right and a good share of SaaS is being repriced because AI can absorb the work, then the multiple is not the story. The story is which of your tools, and which of your vendors, turn out to be the part that gets absorbed. And if the market is wrong and this is an overcorrection, who is quietly buying the bargains? I do not know the answer to either. I would still rather ask now, on my own terms, than have a lower multiple ask for me.

Learn more

• insAIght, Anove's AI governance and risk platform, for keeping one register of every AI system in use, including those embedded in SaaS tools.
• My Gut Says This M&A Boom Is Fine. That Is Exactly Why I Do Not Trust It., the same instinct seen from the deal room.
• ExplAIn, our free tool for checking what an AI system already in use discloses about itself.

If AI is spreading through your software stack faster than your inventory can track it, book a demo and Erik and the team will show you how insAIght keeps up.