SaaS Is Just Recurring Revenue. AI Broke the Rest.

Why the most valuable software companies in the world are about to have an identity crisis — and why the startups eating their lunch might already be winning.

There's a dirty secret in enterprise software that nobody wants to say out loud: SaaS was never really about the software. It was about the revenue model.

Strip away the mythology — the "cloud-native architecture," the "platform," the "ecosystem" — and what made SaaS companies worth 10x, 15x, 20x revenue was one thing: predictable, recurring subscription revenue with near-zero marginal cost. A customer signs a contract, pays every year, and serving them costs almost nothing beyond what you already spent building the product. Gross margins of 80–85%. Revenue that compounds. Wall Street's favorite financial profile.

That model is now breaking. And the market knows it.

The FactSet Paradox

On March 5, 2026, Reuters reported that Thoma Bravo and Hellman & Friedman had been running the numbers on acquiring FactSet after AI disruption fears drove a 39% collapse in its stock over six months. Morningstar and Gartner suffered similar drawdowns — 27.6% and 29.5% respectively. All three are textbook SaaS businesses: subscription revenue, high margins, predictable cash flows, deeply embedded in institutional workflows.

The private equity firms should have been salivating. These are exactly the kind of companies PE buys — stable, cash-generative, subscription-based businesses trading at cyclical lows. FactSet's EV/EBITDA ratio had cratered from 30 in 2022 to roughly 12. That's a 60% compression in the metric that drives every PE deal model.

But here's what happened: the same AI fears that made the prices attractive also made the firms hesitate. As Reuters put it, bankers couldn't accurately value the companies because executives couldn't predict whether their business models would evolve with AI or get overtaken by it. The very thing that created the buying opportunity — the market's belief that AI breaks the SaaS model — also made the buyout math unreliable.

This is the paradox that defines the SaaS industry in 2026. The subscription revenue that once made these companies premium assets is now a question mark. Not because the subscriptions are disappearing today, but because nobody can model what they look like in three years when AI has rewritten how information is packaged, delivered, and priced.

FactSet's revenue was still growing — 6.9% year-over-year in the latest quarter. But investors noted that most of that growth came from price increases on existing subscriptions, not new customers. Steady cash flow, but limited upside. For a leveraged buyout predicated on growth, that's a problem. For a market trying to figure out whether FactSet's product gets replicated by an AI tool at a fraction of the cost, it's an existential question.

And FactSet isn't alone. The selloff is hitting everything — software companies, data providers, consulting firms. It deepened after Anthropic released its latest Claude upgrade. The market isn't distinguishing between companies that will be disrupted and companies that won't. It's re-pricing the entire category.

This is the moment where we need to ask the uncomfortable question: what exactly was the SaaS premium for?

The Margin Problem Nobody Wants to Talk About

Here's what happens when a traditional SaaS company "adds AI" to its platform — which, at this point, is every SaaS company on the planet:

Every AI-powered feature introduces variable, per-use infrastructure cost. Every query hits a model. Every model consumes tokens. Every token costs money. Not a lot of money per query — but multiply it by millions of users, and suddenly your cost of goods sold looks less like a software company and more like a manufacturing business.

GitHub learned this the hard way. When Copilot launched at $10/month, Microsoft was reportedly losing $20–80 per heavy user per month in compute costs. The "unlimited AI assistance" model was hemorrhaging cash. By mid-2025, "unlimited" was quietly replaced with usage allowances and custom pricing tiers.

Replit's story is even more dramatic. Revenue rocketed from roughly $2M to $144M ARR in a single year — but gross margins cratered to single digits, even going negative during usage surges. They only stabilized margins (to around 20–30%) by abandoning flat-rate pricing entirely and moving to usage-based plans.

Bessemer's 2025 data makes it stark: traditional SaaS gross margins sit at 78–85%. AI-first companies average 25% (early stage) to 60% (mature). That's not a rounding error. That's a different category of business.

As one industry veteran put it: *"We're moving from the movie business model — invest millions to create a product you can distribute for free — to something more akin to manufacturing, where we need to care about COGS."*

The Pricing Identity Crisis

So what does a SaaS company do when its cost base becomes variable?

The honest answer: it stops being a SaaS company.

If every AI query costs you money to serve, you can't charge a flat subscription and hope usage stays manageable. You need to pass the variable cost through to the customer. That means usage-based pricing — charge per query, per token, per "resolution," per outcome.

This is already happening everywhere. A 2025 industry report found that 92% of AI software companies now use mixed pricing models. Intercom moved to charging per "resolution" rather than per seat. Notion added AI as a metered add-on. OpenAI and Anthropic price by token consumption. The entire industry is migrating from "pay for access" to "pay for what you use."

But here's the question nobody is asking loudly enough: if you're charging per use, with variable margins, on a consumption basis — what exactly makes you a SaaS company anymore?

The subscription was the whole point. Predictable revenue. Predictable margins. Predictable cash flows. That's what justified the multiples. Take away the subscription and you have a technology-enabled services business. A metered utility. An API with a billing layer.

There's nothing wrong with that. But the market prices it very differently.

The Real Moat Was Never the Software

This is where the argument gets uncomfortable for incumbents. If AI erodes the margin advantage that made SaaS economics special, then what's left?

Two things: customers and data.

A company like Veeva has 1,000+ life sciences customers running their clinical, regulatory, and commercial operations on Veeva's platform. Those customers aren't leaving — the switching costs are enormous, the data is deeply embedded, and the workflows are mission-critical. That's real. That's valuable.

But that's not a software moat. That's a distribution moat. It's the customer relationships and the proprietary data those relationships generate. The software itself is increasingly commoditized — if you can build a functional CRM in a weekend with AI tools, the defensibility of the application layer is evaporating.

So here's the provocative reframing: the most valuable SaaS companies aren't really software companies. They're data-and-distribution companies that happen to deliver their product through software.

And if that's true, then the valuation framework needs to change. We shouldn't be asking "what's the ARR multiple?" We should be asking "what's the value of the customer relationships and the proprietary data, and how defensible are they against an AI-native entrant with no legacy architecture?"

The Incumbent vs. The Startup

Take an AI startup in the same vertical as an established SaaS incumbent. Same industry. Same buyer persona. Same problem.

What does the AI startup lack? Customers, data, brand, distribution. It's selling into the same accounts but starting from zero.

What does the AI startup have? Purpose-built architecture for the AI-native world. No legacy codebase. No margin expectations set by a decade of 80%+ gross margins. No Wall Street analysts asking why margins contracted 300 basis points this quarter. And critically — the same variable cost structure that the incumbent is now being forced to adopt anyway.

The incumbent is migrating from flat subscriptions to usage-based pricing, watching margins compress, and trying to explain to investors why this is "strategic." The AI startup was born with usage-based pricing. It never had the margin illusion to begin with.

The only difference between the incumbent and the startup is the customer base and the data. That's it. The technology advantage has flipped. The margin structure is converging. The pricing model is converging. What's left is distribution.

And distribution, while powerful, is not permanent. It's a head start, not a moat. Especially when the AI-native entrant can demonstrate better outcomes at lower cost because it was designed for this world from day one.

The Uncomfortable Math

Let's make this concrete. Consider a life sciences analytics SaaS company with $50M ARR, 80% gross margins, and a 10x revenue multiple — a $500M valuation.

Now it introduces AI features. Margins compress to 65%. Revenue grows 20% because the AI features are genuinely better. But the market re-rates the multiple to 6x because the margin profile now looks like a different kind of business.

New math: $60M revenue × 6x = $360M valuation. Revenue went up. Valuation went down.

Meanwhile, an AI-native startup in the same space is doing $8M ARR with 55% gross margins and growing 100%+ year-over-year. It raises at 15x ARR because VCs see the growth rate, the AI-native architecture, and the fact that margins will improve as inference costs drop — which they are, rapidly.

Startup valuation: $120M on $8M revenue. The incumbent has 6x the revenue and only 3x the valuation. The market is telling you something.

As Shlomo Dovrat, co-founder of Viola Ventures, told Reuters: even if you grow 25%, a software business won't get the same valuation as a pure disruptive AI company with a $600 billion market opportunity ahead of it. The multiple premium has migrated from "predictable recurring revenue" to "AI-native with proprietary data." The incumbents aren't being rewarded for stability anymore. They're being penalized for it.

So Are SaaS Companies Still Valuable?

Yes. But not for the reasons they used to be.

The subscription revenue model is still attractive, but it's no longer the differentiator — everyone is moving to hybrid pricing anyway. The high gross margins are compressing toward a new equilibrium that looks more like 60–70% than 80–85%. The "build once, sell forever" economics are gone when every AI query has a cost attached.

What remains valuable is the accumulated proprietary data, the embedded customer workflows, the regulatory compliance infrastructure, and the distribution relationships that take a decade to build. These are real assets. They're just not software assets in the traditional sense.

The SaaS companies that will win in the AI era aren't the ones with the best software. They're the ones with the best data and the deepest customer lock-in — who also happen to execute well on AI. The software is the delivery mechanism. The data is the product. The customers are the moat.

And the ones that don't have proprietary data or deep workflow integration? The horizontal tools, the lightweight utilities, the "feature not a product" SaaS companies? They're the ones that AI startups will eat alive. Because when the margin advantage disappears, when the pricing model converges, and when the technology can be replicated in months instead of years — the only thing left is whether anyone needs you specifically, or whether a startup can do it better and cheaper from scratch.

PwC's Alex Baker framed it well: the software market is bifurcating. Some models will be existentially threatened. But the businesses in defensible positions that demonstrate revenue stability will use AI as an accelerant and meaningfully outperform. The key word is "defensible." Not "subscription-based." Not "high-margin." Defensible.

SaaS isn't dead. But the thing that made it special — the economic model — is being structurally altered by AI. What's left after the model breaks is what was always underneath: do you have data nobody else has, and do you have customers who can't leave?

If the answer is yes, you're still valuable. You're just not a SaaS company anymore. You're a data company with a subscription billing system.

If the answer is no, the AI startup with $3M in revenue and a better model is coming for you. And the market already knows it.

Go ask Thoma Bravo. They ran the numbers on FactSet, saw a 39% discount, and still walked away. Not because the business was broken today — but because nobody could tell them it wouldn't be broken tomorrow. That's not a valuation problem. That's an identity crisis.

Published by MaTrx | matrx.ai | March 2026

Reference: "AI fears temper interest as private equity firms weigh data company deals," Reuters, March 5, 2026. By Milana Vinn, Amy-Jo Crowley and Echo Wang.