We’ve Seen This Pattern Before — And It Never Looks Rational in Real Time
- JENNY LEE
- Jan 30
- 4 min read
I remember the early 2000s clearly. A lot of internet companies died. A lot of websites disappeared. Entire sectors were written off.
But people didn’t actually say “the internet was a mistake.” What they said was more subtle — and, in hindsight, more misleading:
“This industry can’t really be replaced by the internet.”
“There’s no sustainable revenue model here.”
“Real business still needs the old structure.”
At the time, those arguments sounded reasonable. They were grounded, cautious, and defensible. They were also wrong.
The internet didn’t fail to generate value. It generated value in ways the market wasn’t ready to recognize or price. Many early companies disappeared, yes. But the network itself kept expanding — quietly — until it became embedded in daily life, logistics, finance, procurement, and even areas we once believed required physical presence and experience.
That memory keeps resurfacing as I watch today’s debate around AI and software.
Right Now, the Market Isn’t Rational — It’s Afraid
There’s a noticeable shift in tone. People aren’t calmly debating software business models anymore. They’re reacting.
“Software is dead.”
“AI replaces everything.”
“SaaS has no future.”
This isn’t analysis. It’s a fear phase. Markets behave this way when a new technology breaks old assumptions faster than pricing frameworks can adjust. The conversation stops being about how things change and turns into whether they survive at all. That’s usually the least accurate moment.
AI Forces a Real Question — Just Not the One People Think
AI clearly lowers the cost of building software. It compresses development cycles. It challenges seat-based pricing models. Those pressures are real, and they matter.
But the leap from pricing power is changing to software becomes obsolete is too large. We’ve made this mistake before — confusing economic disruption with existential replacement.
Software was never valuable simply because writing code was hard. It was valuable because complex systems needed to be governed, stabilized, and owned. AI can help build tools. It still doesn’t replace responsibility.
Software Was Never Just About Code
Enterprises don’t really buy features. They buy structure. They buy:
Accountability
Permission systems
Audit trails
Uptime guarantees
Responsibility when something breaks
AI is very good at generating solutions. It is still very bad at owning consequences. That gap matters more than current narratives want to admit.
Microsoft Is Being Misclassified — And That Matters
One thing especially bothers me in the current discussion: Microsoft is being treated as just another “software company.” It isn’t. Even in a hypothetical world where many traditional software vendors struggle, Microsoft does not belong on that list. Why? Because its core products are not optional layers — they are preconditions.
Operating systems are required for AI to exist and function.
Cloud infrastructure is required for AI to run, scale, and deploy.
Enterprise environments still need identity, access control, security, and orchestration.
AI doesn’t replace these things. It depends on them. Microsoft is not simply selling software licenses. It provides the operating substrate on which AI is generated, trained, deployed, and managed.
If companies are being categorized by vulnerability to AI-driven software compression, Microsoft should be excluded — not because it’s immune to cycles, but because it sits beneath them.
Which Types of Software Are Most Vulnerable First?
If AI truly replaces parts of software, it won’t happen randomly. It will follow the path of least resistance. The most vulnerable software tends to share a few characteristics:
Products built primarily as interfaces over logic, rather than systems
Tools whose value comes from manual configuration or repetitive workflow design
Software that lives above core infrastructure, not embedded within it
Platforms heavily dependent on seat-based pricing without deep usage coupling
Products that can fail quietly without serious operational consequences
These are areas where AI-generated solutions can become “good enough” very quickly. Not because AI is perfect — but because the cost of experimentation has collapsed.
By Contrast, What Feels Structurally Safer?
On the other side are systems that:
Sit at the foundation of enterprise environments
Carry operational, legal, or security responsibility
Become more valuable as complexity increases
Cannot fail quietly
Are painful and expensive to replace
These systems don’t escape disruption. But they tend to absorb AI rather than compete with it. They become the rails AI runs on.
This Feels Less Like a Verdict, More Like a Transition
What we’re seeing doesn’t feel like an endpoint. It feels like:
Old assumptions breaking
Pricing power being renegotiated
Investors pulling confidence forward too aggressively
That happened during the early internet era. It happened again during other major platform transitions. The industries didn’t disappear. They reorganized. Some companies didn’t survive. Others emerged stronger — but different.
Five Years From Now
Five years from now, I’m very aware that I could be wrong. Not wrong in the sense that software disappears overnight — but wrong about how far AI goes, and how quickly organizations adapt to it.
If I’m wrong, it probably won’t look like chaos. It will look quieter than expected:
More companies will realize they don’t need to buy as much software as they used to.
Internal tools will quietly replace external licenses.
Procurement committees will shrink stacks without headlines.
Entire categories won’t “die” — they’ll just stop growing.
If that happens, the danger won’t be that software vanished. The danger will be that we underestimated how quickly good enough + cheap + flexible becomes acceptable at scale.
That’s where real displacement would come from. Not from a single breakthrough — but from thousands of small decisions no one bothers to announce. Even then, what survives won’t be hype-driven platforms or narrative leaders.
What survives will be what always survives in complex systems:
Infrastructure everything depends on
Systems that cannot fail quietly
Layers where responsibility still has to land somewhere
If five years from now software looks very different, it won’t be because AI destroyed it. It will be because we finally stopped confusing convenience with necessity.
And markets — as usual — will have panicked far earlier than they needed to.
Equity Regime X Commentary— Jenny Lee 1/30/2026
