At the core of SaaSmageddon, the workflow still exists, but fewer license-paying humans may be required to run it
In 1812, the Luddites smashed the looms.
Not because they were acting irrational or out of fear… They simply saw what was coming.
The new machines on the horizon could make cloth faster, yes, but they also made skilled labor less necessary. The master weaver, who was once indispensable, became optional. Meanwhile, the power and wealth shifted, and those who saw it coming were able to ride the wave.
By 1844, Friedrich Engels had walked industrial Manchester and saw a paradox: machines made the nation richer, yet workers weren’t sharing the gains. Economic historian Robert C. Allen later called that mismatch “Engels’ Pause”: a period when productivity rose while wages lagged
Today, a different loom is humming and a new Engels’ Pause is approaching. And this time, it isn’t coming for factory workers…
It’s coming for Software-as-a-Service and the middle class.
For 15 years, SaaS was one of the greatest business models ever invented. Build a sleek dashboard, connect it to a database, and charge $30 to $100 per seat, each month. As customers hired more employees, SaaS revenue compounded automatically.
It was the perfect toll booth on the information highway.
But here’s the problem: AI doesn’t need a seat.
And if AI doesn’t need a seat… neither does the software built around one.
That’s the heart of what we’ve been calling “SaaSmaggedon” — the slow-motion collapse of the traditional seat-based SaaS model under the weight of AI agents, automation, and autonomous workflows.
And most investors are not ready. That’s the heart of our discussion this week in Being Exponential With Luke Lango, which you can check out below:
Why AI Is Collapsing the Seat-Based SaaS Model
The old SaaS equation was simple: More employees equal more licenses which equals more revenue.
But agentic AI systems — like Anthropic’s Claude Cowork, OpenAI-powered agents, and other autonomous frameworks — are now capable of completing multi-step workflows independently. From research to draft to data analysis to execution.
Not assisting… Replacing.
If one AI agent can do the work of five junior analysts, a company doesn’t just reduce payroll. It also reduces software seats.
That’s how revenue gets compressed, and it’s hitting the middle layer first.
The Flattening of the Middle Software Layer
Most horizontal SaaS platforms are essentially intermediaries. They sit between a human and a structured database. The user clicks buttons as the dashboard visualizes data and the system logs activity.
But with Model Context Protocol (MCP), API-native execution, and function calling, AI models can now retrieve, update, and reason over structured data without a graphical interface at all.
In other words: no dashboard required.
Why pay $70 per month per employee for access to a workflow tool if an AI agent can access the same database directly and execute the workflow itself?
This is why we’re seeing pressure on companies that sell into SMBs — firms like LegalZoom (LZ) or Unity Software (U) that rely on broad, seat-based adoption. Their customers are exactly the ones most likely to replace labor with AI.
At the very top end — think Palantir (PLTR) — deeply embedded enterprise systems with mission-critical data moats may survive. But the middle? It’s getting squeezed.
Project Genie and the Death of Creation Moats
Then there’s the creative layer.
Google’s Project Genie demonstrated something profound: high-fidelity environments — even video games — can now be generated from prompts.
The barrier to entry for building sophisticated applications is collapsing.
If small teams can spin up production-grade tools in days instead of months, legacy software vendors lose pricing power. Creation itself is becoming commoditized.
For years, SaaS companies justified premium valuations on the idea that their software was hard to replicate.
AI just changed that math.
The ‘Show Me the Money’ Pivot
After three years of AI hype, Wall Street has shifted from asking, “What’s your AI strategy?” to “Where are the profits?”
And the profit pools are not where many expected.
Chipmakers like Nvidia (NVDA) and Micron (MU) are printing cash. Demand for GPUs, high-bandwidth memory, and AI infrastructure is insatiable.
Meanwhile, software giants like Salesforce (CRM) and Adobe (ADBE) face a painful dilemma: integrate AI and risk cannibalizing seat revenue… or don’t integrate it and risk irrelevance.
Multiples are compressing in the application layer.
Margins are expanding in the infrastructure layer.
That’s not an accident. AI shifts economic power downward… into compute, memory, and silicon.
The Real Opportunity: AI Hardware and Memory Bottlenecks
If SaaSmaggedon is the destruction phase, AI hardware is the buildout phase.
The AI revolution is infrastructure-heavy. Massive data centers. High-performance GPUs. Advanced networking. And, critically, memory.
Right now, AI memory bottlenecks are one of the most important investment themes in the market. Training and inference workloads require enormous bandwidth and storage throughput.
Nvidia’s financial releases show the scale: record revenue driven by data center demand and guidance implying mid‑70% gross margins in fiscal 2026. The Wall Street Journal reports Micron is investing heavily to break an AI memory bottleneck, describing shortages and rising prices as high-bandwidth memory (HBM) demand accelerates.
That’s why companies like Seagate (STX) and SanDisk (SNDK) have rallied — and may still have room to run. As AI models grow larger and edge AI expands into devices, cars, and robotics, memory demand compounds.
After memory? Edge AI is the next likely bottleneck.
As inference moves closer to the user — into smartphones, PCs, industrial systems — the next wave of hardware winners will emerge in a hardware-first supercycle.
A Narrow Window
History suggests that when technological shifts occur, capital reallocates quickly but unevenly.
Right now, we may have a 12- to 24-month window where infrastructure beneficiaries surge before broader economic disruption sets in.
SaaS investors who ignore the seat-based revenue risk may find themselves holding yesterday’s business model.
Hardware investors who understand where the bottlenecks are forming may capture the upside of the AI buildout.
The looms are humming again.
The question isn’t whether AI will reshape software economics… It’s whether you’ll be positioned on the side building the machines… or the side being replaced by them.
In the full podcast, we go much deeper into which software companies might survive SaaSmaggedon, where the next AI hardware choke points are forming, and how to position a hypergrowth portfolio for what could be the most dramatic capital rotation since the dot-com era.
This time, the machines aren’t just making cloth, they’re rewriting the economy.
