The SaaSpocalypse was always coming — we just needed to see this chart
On frictionless markets, “chameleon code” and why AI valuations aren't as crazy as they look
“Chameleon code pus SaaS slomo.” © 2026. pitchhawk. Image generated with AI.
One chart
There is a chart that tells you everything you need to know about the current moment in technology investing.
It plots two companies against each other, both indexed to the same starting point.
One is the undisputed poster child of the SaaS era. The company that proved software could be delivered over the internet, that made "the cloud" a boardroom word, and that grew from nothing to a $176 billion revenue multiple over seven years.
The other is a seventeen-month-old data point that makes the first line look like it barely left the driveway.
The first company is Salesforce. The second is Anthropic.
And when you see them on the same chart, truly indexed, apples to apples, growth rate to growth rate, something shifts in your understanding of why the AI valuations that look insane in a spreadsheet are, in fact, the only rational response to what the data is showing.
Click on the Monthly (first 24m) and Annual (7 years) tabs to see what I mean. And if you prefer logarithmic scales, there’s provision for that as well.
Monthly zoom: Salesforce reaches only 5× by month 24 — its hockey stick comes later. Anthropic hits 44× by month 17.
Both indexed to 1× at start. Sources: HockeyStickPrinciples.com · SemiAnalysis newsletter
The gold standard in cloud, relativised
Salesforce was not a slow company. It was the defining hypergrowth story of its era. Marc Benioff built one of the most important enterprise software businesses in history, grew it through multiple recessions, and in doing so essentially invented the playbook that every SaaS founder has followed for the last two decades: land, expand, retain, repeat.
By month 24 from its founding, Salesforce had grown revenue roughly 5× from its starting point. That was extraordinary. That was the benchmark. That was what "hypergrowth" looked like for a generation of venture capitalists and growth equity investors.
Anthropic, in the same window, grew 44×.
Not 44% faster. Not 4× faster. 44 times the absolute starting value, in the same timeframe. By the time Salesforce had reached 5× growth at month 24, Anthropic had already blown past that at month 12 and kept accelerating. Its revenue run rate went from $1 billion in December 2024 to $44 billion annualised by May 2026 — a 44× expansion in seventeen months.
To put that in context, Salesforce took seven full years to reach a 176× index from its starting point. Anthropic covered a quarter of that journey in less than a fifth of the time.
Read that again 👆
This is not a small difference in degree. It’s a difference in kind.
Why AI has zero friction at the starting line
The reason SaaS grew the way it did, impressive but measured, is that SaaS still had a sales motion. It had procurement cycles. It had IT departments. It had security reviews, change management, user training, integration projects, and contract negotiations. Salesforce needed a field sales army. It needed Dreamforce. It needed relationships.
This friction was also a moat. It created switching costs. It created the "sticky" revenue that SaaS multiples were built on. But it also put a natural ceiling on how fast you could actually grow, because every new dollar of revenue required a proportional investment in humans, processes, and time.
AI arrived into a fundamentally different market structure.
A developer can integrate a frontier AI model in an afternoon. An enterprise can deploy an agent that replaces an entire workflow without buying new infrastructure. A product team can rebuild a feature they used to pay a SaaS vendor for, not in a six-month integration project but in a sprint. The marginal cost of adding a new customer to a model API is not a sales rep and a solutions engineer. It is essentially zero.
This is what "frictionless market entry" actually means in practice. Not that there are no barriers, i.e., trust, compliance, and hallucination risk are real, but that the velocity at which value can be delivered and captured is structurally different from anything the software industry has produced before. The growth curve is not just steeper because AI is better software. It’s steeper because the distribution mechanism has been fundamentally decompressed.
“Chameleon Code” and the software obsolescence problem
Here is the part of the AI story that makes SaaS investors genuinely uncomfortable. And should.
Traditional software is vertical. Salesforce does CRM. ServiceNow does workflow. Workday does HR. Each company carved out a domain, built deep integrations, accumulated data, and then defended its position through exactly the kind of friction described above. The moat was the complexity. You didn't leave Salesforce because leaving Salesforce was a multi-year project.
Agentic AI is the opposite of vertical. It is, in the most literal sense, what I like to call chameleon code — software that can reason about a problem domain, learn the shape of a workflow, and then execute against it without being pre-programmed for that specific use case. An agent deployed to manage customer relationships doesn't need Salesforce's data model. It can read your emails, your call transcripts, your support tickets, and synthesise a richer picture of customer health than any CRM dashboard ever could. It doesn't replace the data. It makes the interface, the workflow layer, and the licensing agreement that sits on top of that data optional.
This isn’t theoretical. It’s already happening at the edges of the enterprise software market, and the edges tend to predict the centre with about an eighteen-month lag.
The SaaSpocalypse, which is to say the compression of SaaS multiples from the 20–40× revenue peaks of 2021 to the 5–10× range where many names now trade (and 3x where the emerging cloud companies are now valued according to Bessemer data) has been widely attributed to rising yields and multiple compression across risk assets.
That story is true, and according to your own research, a rising 10-year Treasuries yield supports this compression, but it’s an incomplete picture. The deeper post-Anthropic driver is that the market is beginning to price in what happens to recurring software revenue when the friction that made it sticky starts to dissolve.
When a sufficiently capable AI agent can approximate the function of a point solution, the question investors are asking is no longer "what is this revenue worth in perpetuity?" It is "how long does this revenue last?"
The game has changed, profoundly.
The value trap hiding in plain sight
Low multiples feel like safety. A SaaS company at 5× revenue, with 80% gross margins and a $10 billion installed base, looks like a value play to a classically trained software investor. The cashflow is real. The customers are real. The NRR seems defensible.
But a value trap is not a company with bad fundamentals today. It’s a company whose fundamentals are about to become structurally impaired, trading at a price that reflects neither that impairment nor the timeline on which it arrives.
Our chart is the argument.
If Anthropic and the two or three frontier model companies that will ultimately matter continues on even a fraction of its current trajectory, the question for every enterprise software incumbent is not whether AI will affect their market. It’s whether their market will exist in its current form in five years. And that question lands very differently when you look at the indexed growth comparison and ask which of these two curves represents the direction of capital allocation in large enterprises over the next decade?
The answer is not ambiguous.
None of this means that every SaaS company dies immediately. Vertical depth, proprietary data, and regulatory complexity will insulate some categories (healthcare records, financial compliance, government systems) for longer than the bull case for AI disruption implies. There will always be survivors. There will even be acqui-hires and pivots that look like successes.
But "some will survive" is not the same as "this is a safe asset class at current prices." A 5× revenue multiple on a business with structurally declining pricing power and a customer base that is quietly evaluating whether chameleon code, i.e., an agent can do the same job is not obviously cheap. It may be exactly the right price for the risk. Which is to say that buying the dip might be a trap.
What the valuations are actually saying
When a frontier AI company raises at a valuation that looks disconnected from current revenue, the market is not being irrational, it’s doing a different calculation.
It’s most likely looking at a growth curve that makes the SaaS gold standard look as pedestrian as the proverbial tortoise, and pricing in a distribution model with near-zero marginal cost of delivery. It’s probably anticipating a market size that’s not "the enterprise software market" but something closer to "every knowledge workflow that exists in the global economy." And what’s not profound about that.
And it’s doing all of this against the backdrop of a chart that shows 44× growth in seventeen months from a company that did not exist as a serious revenue business two years ago.
The numbers are large because the opportunity is large. The opportunity is large because for the first time in the history of software, the chameleon code can genuinely shapeshift to fit almost any problem domain. Not through configuration and customisation, but through reasoning.
That’s not a normal software company. It’s something new, and whatever the markets flaws are, it’s pricing it as such. And it’s also why energy, capex, data centre and chip stocks are second, third and fourth order ways to play this race to every knowledge workflow that exists in the global economy.
The chart was always going to look like this, so perhaps the only surprise is that it took this long for the rest of the conversation to catch up. The question for us all now is whether these tectonic forces are fuelling our businesses — or turning them into museum pieces.
Reach out if you’d like to discuss any of the above.
Mike 🖐
The indexed growth data referenced in this piece uses Salesforce revenue from HockeyStickPrinciples.com and Anthropic revenue run rate data from the SemiAnalysis newsletter. Both series are indexed to 100 at their respective starting points for comparability.
Innovation doesn't stall for lack of ideas. It stalls in the gap between a great innovation and an investable business.
A sell-side industry was never going to close it — nobody was looking from the investor's side.
pitchhawk is.
Copyright, pitchhawk, 2025-6. All rights reserved.