Press play to listen to this article
Why AI is rewriting the economics of SaaS and shifting value from tools to outcomes
Over the past few months, more than $1 trillion in SaaS market value has vanished. That tends to get attention. Our clients are asking a very practical question: what’s actually happening, and what does it mean for how they choose partners, investments, and acquisition targets? Because the real risk isn’t missing the next big thing, but it’s backing companies whose advantage is quietly disappearing.
What’s Really Happening: A Reset in Confidence, Not Performance
The SaaS sector has shed more than $1 trillion in market value. That’s not a small correction; it’s a structural repricing, and valuation multiples have compressed dramatically too. Public software companies are now trading at around 3.9x forward revenue, down from 15x to 25x at the 2021 peak. That’s a 70 to 80 percent contraction. Free cash flow multiples tell a similar story. Salesforce, for example, has seen its multiple fall from roughly 30x to closer to 15x.
At first glance, this looks like a sector in distress. But the underlying fundamentals tell a different story.
Revenue growth has not collapsed. Customer churn has not spiked in any meaningful way. Net retention for strong companies remains healthy. In other words, the operating performance of SaaS businesses does not justify the magnitude of the repricing.
So what’s actually going on?
To understand it, you have to rewind the tape.
For years, SaaS was treated like a growth annuity. Predictable recurring revenue, high retention, strong visibility. These characteristics made future cash flows feel not just likely, but dependable. That perceived durability supported premium valuation multiples.
What has changed is not the revenue. It is the confidence in the durability of that revenue.
The market is no longer asking when SaaS cash flows might taper off. It is asking whether they are as defensible as once believed. That shift in the question is everything.
This is not a revenue collapse. It is a reset in terminal value assumptions.
What the Data Actually Shows: Expansion Under the Surface
Here’s the tension.
The market is pricing in disruption, but the operating data points to expansion.
Software engineering job postings are still up ~11% year over year, even as AI makes code easier to generate. AI capex is approaching ~2% of GDP, and nearly 2,800 data centers are in development to support growing compute demand. Adoption is progressing along a typical S-curve, not a sudden displacement event.
This is where Jevons Paradox comes into play.
When something previously constrained becomes cheaper and easier to use, we don’t use less of it. We use more. Lower cost of code does not reduce demand for software. It expands it, driving more company formation, more use cases, and more infrastructure.
That is exactly what the data is showing.
This is not a collapse in software demand. It is a recomposition of how and where that demand is created.
Both things can be true at once. Markets are repricing durability, while the real economy is accelerating investment.
The Next Era of Software Is Agentic
The shift underway is not subtle. Software is moving from enabling work to performing it.
For decades, enterprise software functioned as a productivity layer. It improved workflows but relied on human execution. Today, that boundary is now breaking, because AI systems are increasingly capable of completing tasks end-to-end.
This changes how software is built, priced, and evaluated.
The implication is straightforward: the unit of value is no longer the user. It is the work completed. Seat-based pricing reflects access, and agent-driven systems reflect output. As a result, pricing models are beginning to shift toward consumption and outcomes.
At the same time, a new layer is beginning to emerge. As workflows become automated, coordination across models to data becomes much more complex. The platforms that manage this orchestration, from routing tasks, allocating compute, to governing decisions, become structurally important. In this environment, the “user” of software is no longer always human. While the buyer remains human, the operator is increasingly not.
That distinction matters.
Where Value Accrues (And Why)
By the end of the decade, AI agents could represent more than 60% of the market, marking a shift from software that helps people do work to software that completes it.
The implication is not contraction, but redistribution and growth.
While the software market may expand by more than 20% in the coming years, that growth will not be evenly captured. Instead, value is bifurcating across three categories:
- Software that helps humans do work
- Software that completes work
- Software that performs work humans cannot do
Historically, nearly all enterprise value sat in the first category. That is no longer the center of gravity.
As execution shifts into software, value moves increasingly toward systems that control outcomes rather than interfaces. This is where the moat reset becomes clear.
Most software moats were never about code. They were structural: network effects, scale, brand, proprietary data, process power, and switching costs.
AI is now stress-testing those advantages.
Moats built on friction begin to weaken. Thin UI layers, seat-based bundle pricing (where only a fraction of features are actually used), model-only differentiation, and closed ecosystems that fail to integrate with broader workflows all become increasingly replaceable as agents intermediate usage and route work across systems.
As a result, switching costs may decline, and products that relied on being “good enough” within a closed environment face rising pressure.
But this is not uniform erosion. Rather, it is a sorting mechanism.
The strongest moats do not erode; they compound. Network effects deepen as more agents and systems interact within the same platforms. Proprietary data becomes more valuable as it trains and differentiates models. Process power, how work actually gets done across systems, becomes harder to replicate at scale.
This is what the market is actually repricing: not revenue, but durability.
The Bottom Line
The SaaS ecosystem is far from over. In fact, it is being rebuilt in real time.
The market is repricing durability, not demand. The assumptions behind premium multiples are being challenged, even as software usage continues to expand. The next winners will be defined by control over data, workflows, and outcomes—not interfaces.
The question is no longer which software people use. It is which systems that the work flows through.
