The Thesis

SaaS companies have spent two decades commanding premium valuation multiples. The logic was sound: building software was hard, distribution was defensible, switching costs were real, and scarcity drove pricing power. Investors paid 15x, 20x, 30x revenue because those dynamics looked durable.

They are not durable. As we argued last August, AI-generated software is collapsing the cost of creation toward zero. When that happens, the scarcity that justified those multiples evaporates. Software stops being a differentiated product and becomes a commodity input - like electricity, like bandwidth, like compute after AWS.

But here is the part most people are missing: the margin does not disappear. It migrates. And the most logical destination is downstream - to the businesses that use software to operate, manufacture, and serve customers.

The Mechanism

Think about what happened with cloud infrastructure. AWS did not simply lower hosting costs. It transferred the infrastructure premium that Sun, SGI, and IBM commanded for decades directly into the hands of every startup and operator who could spin up servers for pennies. The value did not evaporate. It moved - from hardware vendors to the businesses running workloads on top of them.

AI-generated software does the same thing one layer up. The application premium that Salesforce, ServiceNow, and Workday capture today migrates to their customers. A manufacturer can now build bespoke operational software at near-zero cost. A logistics company can generate custom route optimization without a seven-figure contract. A regional hospital system can have clinical workflow software that fits its actual workflows, not a vendor's reference architecture.

The margin implied in software company valuations does not vanish. It shifts to the consumers of software. And it compounds there, because operational businesses can deploy it directly against revenue-generating activity.

Why Operators Win

The New Scarce Inputs

If software is abundant, the bottleneck shifts to what is still scarce. And what remains scarce is exactly what traditional operating businesses already have: domain expertise accumulated over decades, proprietary data generated by real operations, customer relationships that took years to build, physical assets and infrastructure that cannot be conjured through a prompt, and - where it applies - regulatory positioning that took even longer to establish.

These are not software company assets. They are operator assets. And in a world of abundant code, they become the primary source of competitive advantage.

The Compounding Effect

So it is not just that operators capture the software margin. They are also suddenly better positioned than pure software companies, because they hold the inputs that actually matter when code is free. A logistics company with ten years of route data and carrier relationships does not just save on software costs - it now has a compounding advantage that no software-only competitor can replicate. The domain moat and the software capability reinforce each other.

The companies that win will not be the ones building the best software. They will be the ones deploying it most effectively against real-world problems where they already have domain advantage.

The Hard Part

None of this is automatic. Operators have to actually move. Most won't - at least not fast enough. The same organizational antibodies that slow enterprise software adoption will slow this too. Procurement processes designed for vendor contracts don't have a category for AI-generated internal tooling. IT departments structured around managing external software relationships face an identity crisis when software becomes internal.

The operators who close that gap first will look, from the outside, like they suddenly got much better at their core business. What will actually have happened is that they absorbed a decade of software margin in a few years and put it to work where it belongs - in the operation itself.

For incumbent SaaS vendors, this is both a threat and an instruction manual. The ones who survive will not compete on features. They will compete on implementation - software coupled with the deep service layer needed to embed AI in ways that match each client's appetite, educate them on what is possible, and help them prioritize what is plausible. That is predicated on trust and genuine understanding of the client's business. In other words, the surviving SaaS companies will look a lot more like professional services firms than product companies. Different margins, different talent, different sales motion. Not obviously better.

The Implication

If this thesis holds, capital allocation should follow. The rotation is from SaaS toward software-enabled operators: businesses in manufacturing, logistics, healthcare, and financial services that can absorb AI-generated software fastest and convert it into operational advantage.

The market will quantify this in phases, with increasing sophistication. The first wave is straightforward: takeout costs. Investors will credit operators for the software licenses and headcount they no longer need - easy to model, easy to verify. The second wave is harder to measure but more valuable: standalone productivity gains, where the same team does more and moves faster. The third wave is where it gets genuinely interesting - network-scale synergy, where proprietary operational data compounds across the business and benchmarking on what efficient operations actually look like becomes a structural advantage in itself. Most investor models today are still in wave one.

The valuation multiple attached to software does not go away. It migrates to wherever the real scarcity is. Right now, that is operators with domain depth, proprietary data, and the willingness to move before their competitors figure out what is happening.

The window for that arbitrage is not permanent. It never is.