I've been watching the software sector meltdown for the past month and I think the market's misdiagnosing what's actually happening here.
What everyone thinks is going on:
AI is replacing enterprise software. Agents are making Salesforce obsolete. The whole SaaS model is dead because why pay for software when AI can just do the work directly? That's the narrative anyway.
What's actually happening (and why it matters):
The software isn't being replaced. The people using it are getting fired.
Think about it - when Block cut 4,000 employees last week, those 4,000 people don't need Salesforce seats anymore. Don't need Workday logins. Don't need ServiceNow licenses. The software still works fine. It's just that the humans who needed it are gone.
This isn't product obsolescence. It's demand compression.
And that distinction completely changes which companies survive this.
---
Here's the pattern I'm seeing:
Companies announce AI-driven layoffs → stocks jump (like Block up 25%) → but those eliminated workers were using enterprise software → software companies lose seats → revenue projections get cut → software stocks crash.
The $660B+ going into AI infrastructure this year? A lot of that came straight from enterprise software budgets. Same CIOs who used to expand their Salesforce contracts are redirecting that money to AI compute instead. They're not replacing Salesforce with agents. They're replacing the people who needed Salesforce, then cutting the software budget.
Two-stage compression happening at once. First the labor layer gets compressed. Then the software layer that served those workers gets compressed right after.
Markets saw the first part coming. The second part hit all at once on Feb 3rd when Anthropic dropped Cowork - $285B gone in one trading session. A trillion in software market cap evaporated over six weeks. P/E multiples collapsed from 39x to 21x (sharpest drop since dot-com).
---
Where I think the market's getting it wrong:
The selloff's been totally indiscriminate. Everything software-related got hammered. But there's actually a pretty clear split between what survives and what doesn't.
What gets destroyed:
Seat-sensitive SaaS. Software that's priced per user, where the whole point is serving human workflows. When the humans disappear, revenue goes with them. Workday HR modules, collaboration tools, productivity software - these are directly tied to headcount. No workers = no seats = no revenue.
What actually survives (or grows):
Systems of record. The databases and workflow engines that enterprises can't replace. 20 years of Salesforce customer data doesn't just get rebuilt in some new AI tool. The compliance infrastructure, audit trails, integrations with 500 other systems - these don't vanish when you fire people. If anything they become MORE critical because now AI agents need to run through them.
Plus infrastructure - compute, data pipelines, security. Agents need more of this stuff, not less.
The problem with how this got priced:
Salesforce is down 26% YTD. ServiceNow down 28%. But Salesforce's Data Cloud just hit $1B ARR. ServiceNow's positioning itself as the orchestration layer for AI agents. These aren't the same as Workday processing HR for employees who don't exist anymore.
Market priced the fear. Hasn't priced the differentiation yet.
---
Management behavior worth watching:
Every enterprise software CEO is announcing their own AI product right now. Benioff's got Agentforce. ServiceNow's got Agent Control Tower. Workday's pushing AI-powered HCM.
Question is - are they doing this because the products actually work, or because their stocks need the narrative?
Look at Salesforce charging $0.10 per agent action vs $150/month for a human seat. That math only works if agents generate way more billable actions than humans did. Not proven yet. The pivot's real but revenue replacement is still TBD.
When every CEO makes the exact same pivot under the exact same market pressure at the exact same time... that's not conviction. That's stock management.
---
Testing my own thesis (because I could be wrong):
Strongest counter-argument: maybe the "vibe coding will replace enterprise software" take is just wrong. Building v1 is maybe 2% of what enterprise software actually does. Maintenance, compliance, integrations, scaling - that's the other 98%. Nobody's rebuilding 20 years of Salesforce in Cursor.
Seat compression is definitely real. But the product replacement thesis might be way premature.
Which means... the panic selloff created opportunities. Companies with real data moats and deep integrations got oversold. Companies whose whole value was in per-seat revenue - those got correctly repriced (or are still overpriced).
So revised take: don't short all software. Short the pure seat-plays with no moat. Look for entry points on the infrastructure survivors that got caught in the panic.
---
What hasn't repriced yet:
Mid-cap SaaS and collaboration tools already took their hit. Might even overshoot down from here.
Large platforms (Salesforce, ServiceNow, Workday) - market's still figuring out what survives. This is where the opportunity is.
But the next wave? Staffing and consulting. If enterprises need fewer people and less software, they need way fewer consultants implementing that software. Accenture, Cognizant, Infosys are already cutting but the market hasn't fully priced the contraction of that whole layer yet.
And the final cascade that's not even visible yet - commercial real estate tied to office space for workers who got eliminated. Consumer spending from all those displaced white-collar salaries. Those are 12-24 month signals. Already baked into the structure but markets aren't there yet.
---
How I'd actually trade this:
Short: pure seat-count SaaS with no data moat, no agent strategy. Anything whose revenue assumes enterprise headcount keeps growing.
Long: Salesforce Data Cloud specifically (the data asset's irreplaceable even if seats compress). Palantir (already positioned as AI infrastructure). Security (agents create more attack surface not less).
The real trade: long the infrastructure layer agents need. Short the human-serving layer agents replace. Not betting on the sector - betting on structural compression within it.
Key monitoring signal: net revenue retention dropping below 100% consistently. When that breaks across multiple companies, compression's structural not cyclical. Starting to see that number crack.
---
Bottom line:
The SaaSpocalypse is real. Just not what people think it is.
Market priced it as "AI replaces software." Reality is "companies fire workers, then cut the software those workers used." The software's not obsolete. The humans got removed from the equation.
That difference determines who survives.
The panic selling created a gap between fear and structure. Companies with irreplaceable infrastructure got lumped in with seat-count plays. That gap's the opportunity.
And the repricing's not done. Three more cascade waves coming - consulting layer, real estate, consumer spending. Those haven't hit yet.
---
- Zahaviel