Jason Hauer | CEO, HauerX Holdings | February 2026
Last week, the stock market did something it almost never does. It told the truth in real time.
More than $400 billion in market value disappeared from software, data, and financial technology companies in a single week. Not because of a recession. Not because of a rate hike. Because Anthropic released a set of AI plugins for Claude Cowork that can perform legal contract review, draft and manage marketing campaigns, research prospects and run sales workflows, automate compliance, and build production-quality software. And because OpenAI launched Frontier, an enterprise platform that deploys AI agents as coworkers across an organization's entire tech stack.
The market looked at what these tools can do right now, today, and repriced an entire sector.
Thomson Reuters dropped 16% in its largest single-day decline ever. RELX, which owns LexisNexis, fell 14.6%, its worst day since 1988. The S&P 500 Software & Services Index shed roughly $830 billion across six sessions. Hedge funds had already built $24 billion in short positions against software stocks heading into the week.
This was not panic. This was price discovery.
Some of those stocks bounced on Friday. Thomson Reuters and LegalZoom each recovered a point or two. The broader market surged, with the Dow hitting 50,000 for the first time. But the software sector ETF is still down more than 24% for 2026, and analysts at BTIG noted that while a bounce was likely, the sector will "take a long time to repair and build a new base." The recovery doesn't change what the selloff revealed.
The market is telling you something most enterprise leaders haven't accepted yet: AI isn't just replacing individual tools. It's creating an entirely new layer of the enterprise stack, and that layer is where growth, margin, and competitive advantage will live going forward. The value layer is moving. The market sees it. Most boardrooms don't. And the gap between those two realities is where all the value creation and destruction will happen over the next 18 months.
The mechanism behind the repricing
Brad Gerstner of Altimeter Capital broke down the mechanics of this selloff the way he always does, directly and precisely. Three years ago, he noted, investors would pay 35 times free cash flow for a company like Salesforce because they could see 35 years into the future with near-certainty. The predictability of enterprise software revenue was like a government bond. You were going to get those cash flows. It was a sure thing.
Then AI changed the visibility window.
Gerstner's point is critical for understanding what happened this week: "I'm not penalizing them because they're missing their numbers today. I'm just putting it in the too hard bucket because I can't predict those future numbers." Investors aren't saying these companies are broken. They're saying the future got harder to model. So they pay less for terminal value. The stock price drops not because the present is failing, but because the future is no longer a straight line.
The only way it reverses, Gerstner argued, is if those companies accelerate their core revenues and prove they are beneficiaries of AI, not casualties of it. Companies like Databricks and Snowflake, growing 60%+ because AI infrastructure depends on them, will be fine. Application software, where the future is now opaque, will carry lower multiples until it proves otherwise.
And the repricing is compounding. Every new capability release resets the math. As Gerstner put it: "When you're in the middle of exponential change, investors say, fog of war. I can't predict 35 years into the future those free cash flows, so I have to pull it in a little bit." A little bit, multiplied across every SaaS company on the planet, is $400 billion in a week.
So what, specifically, caused investors to recalculate those future cash flows in a single week? What did they see that turned fog into a storm?
The shift that triggered it
There's a meaningful difference between AI that helps you do your work and AI that does the work.
For the past three years, most enterprises have treated AI as a productivity layer. Summarize this document. Draft this email. Analyze this dataset. Useful, but incremental. The tools released in the first week of February 2026 are not incremental.
Claude Opus 4.6 processes a million tokens of context in a single pass. That's an entire codebase. A year of contracts. A full regulatory filing. SentinelOne's Chief AI Officer said it handled a multi-million-line codebase migration "like a senior engineer." Claude Cowork lets non-technical users automate workflows across their desktop through a Chrome extension, no code required. Claude Code, already in production, grew revenue 5.5x within months and is used by engineers at Microsoft, Google, and Nvidia.
OpenAI's Frontier platform connects an organization's CRM, data warehouses, ticketing systems, and internal tools, then deploys AI agents that work across all of them. Not as assistants. As operators. Initial customers include Intuit, State Farm, Thermo Fisher, Uber, and Oracle. OpenAI's internal data shows a manufacturer reduced production optimization from six weeks to one day. A global investment company freed up more than 90% of salesperson time.
The market didn't see better chatbots. It saw AI performing the actual functions that billion-dollar software categories were built to organize. That's why the selloff hit application software, not infrastructure. This wasn't a repeat of the DeepSeek scare in January 2025, which threatened the hardware layer. This one hit the work layer. The place where enterprises spend, operate, and compete.
And this is exactly what Gerstner's framework predicts. When investors can no longer model what the application layer will look like in five years, let alone thirty-five, they compress the multiple. The tools that launched this week didn't just demonstrate new capabilities. They demonstrated that the future of enterprise software is being rewritten in real time, by companies that didn't exist five years ago.
Where the moats are moving
For decades, competitive advantage lived in proprietary data, distribution networks, scale, and switching costs. AI is compressing all of them. Foundation models trained on internet-scale data have made proprietary datasets less distinctive. AI coding tools have collapsed the cost of building software products. Agentic systems are beginning to automate the enterprise migrations that used to require 18 months of consulting fees, the exact switching costs that protected incumbents.
But the disruption is more nuanced than the headlines suggest. David Sacks offered perhaps the sharpest analysis of where the real threat lies. It's not that SaaS companies get replaced overnight. A system like Salesforce has had millions of bug reports filed over 25 years. It's been tested across thousands of enterprises. The idea that you rip that out and replace it with probabilistically generated code maintained by a small internal team, Sacks argued, "just doesn't seem realistic."
The real risk is subtler and, for incumbents, potentially worse.
Every major SaaS company is building AI copilots inside their tools. Some work well. But they're limited to playing in their own sandbox. Meanwhile, Claude Cowork already has connectors across multiple SaaS platforms. It pulls data across tools. It operates seamlessly across databases and applications. As Sacks put it: "Which one of these products is going to be your workspace? It seems to me you're going to want your workspace to be the one that spans across and gives you AI across the most data and context, as opposed to having a bunch of separate AIs inside of your existing tools."
This is the structural insight that matters most. The threat to SaaS companies isn't extinction. It's becoming a lower layer of the stack while the value capture moves to a new layer built on top of them. They become legacy infrastructure, still necessary, still running, but no longer where the growth happens. The action moves to whoever controls the AI layer that orchestrates across all of them.
And that is exactly what those companies were banking on as their next growth chapter. Every product roadmap in enterprise software is AI-first right now. If that value capture happens somewhere else, at the orchestration layer rather than the application layer, the entire forward earnings story collapses. Which is precisely what the market repriced this week.
Sacks also drew a useful distinction on vulnerability. Expensive horizontal software where users only touch a handful of features is a rip-out target. The ROI math doesn't hold when a bespoke AI-built alternative costs a fraction. But deeply embedded vertical software that serves as a true system of record, the kind that's hard to rip out, has more runway. The question for every enterprise leader and every technology company is honest assessment: which category does your stack fall into? And are you building for the layer that's gaining value or the one that's losing it?
The new moat is execution velocity. Who ships AI capabilities fastest. Who embeds them deepest into customer workflows. Who builds compounding learning loops that improve with every interaction. The companies winning right now aren't the ones with the most data or the biggest teams. They're the ones that move fastest and compound hardest.
This is what I mean when I say compounding beats campaigns. The winners in this cycle aren't running AI experiments. They're building systems that get better every day, automatically, as a function of operating across the stack.
The evidence is already in
I wrote about the AI capability overhang in my last article. The concept is simple: AI capability has leapt forward. Adoption hasn't. And the distance between those two things is growing, not shrinking.
The data confirms this. BCG studied 1,250 senior executives and found that just 5% of companies are "future-built" for AI. That 5% is pulling away fast, with 1.7x the revenue growth and 3.6x the total shareholder return of their peers. Meanwhile, 42% of companies scrapped most of their AI initiatives in 2025, up from 17% the year before. Not because the technology failed. Because they bolted AI onto existing workflows instead of redesigning how work gets done. The five major hyperscalers are projected to spend more than $600 billion on AI infrastructure in 2026. The infrastructure is built. The capability exists. The bottleneck is us.
But the companies that have committed fully are producing results that would have been unimaginable three years ago.
Klarna cut its workforce from 5,500 to under 3,000 while doubling quarterly revenue to $903 million. Revenue per employee up 152%. Its AI handles two-thirds of all customer service, resolving issues in 2 minutes instead of 11. The CEO called it "our strongest quarter ever."
Cursor, an AI-native code editor, hit $1 billion in ARR with roughly 300 employees, reaching $100 million faster than any SaaS product in history. Salesforce, with 40,000 engineers, now runs 90% of its developers on it.
Harvey AI went from a prototype to an $8 billion valuation serving 100,000 lawyers across 60 countries. That's why Thomson Reuters lost 16% in a day. The market wasn't reacting to a product launch. It was repricing what happens when that capability becomes a commodity.
Replit hit $100 million ARR in 5.5 months with 65 employees. Midjourney generates $200 million annually with 40 people. These aren't outliers anymore. They're the new math. Every one of these companies built from the ground up with AI as the operating system. The result is revenue-per-employee ratios that legacy organizations cannot match with traditional structures.
And the gap compounds. Every month an AI-native competitor runs production workflows, its systems learn, its cost structure improves, and its speed increases. Every month an enterprise stays in pilot mode, the distance grows.
The disruption extends well beyond software. John Gray, President of Blackstone, the world's largest alternative asset manager, has been explicit: "Everyone's focused on these bubble risks. I think the biggest risk is actually the disruption risk. What happens when industries change overnight." In January, JP Morgan announced it would stop using proxy advisors and use AI instead. Lemonade said it would cut insurance rates by 50% for miles driven by AI-powered self-driving Teslas. Gray connected the dots: what does that mean for insurance companies? For collision repair? The disruption cascades into every adjacent industry.
Blackstone now puts AI risk assessment on the first two pages of every deal memo. They're spending "tons of time thinking about rules-based businesses" across accounting, legal, IT services, and any business that functions as an intermediary. If the core value of a business is organizing, processing, or intermediating information according to defined rules, agentic AI can perform that function. The question is when, not if.
Dario Amodei, Anthropic's CEO, put it in historical context. The disruption coming from AI is "not different in kind" from past technological shifts, from farming to factories to knowledge work. But it's deeper and it's faster. AI can perform entry-level law, finance, and consulting simultaneously. It hits multiple points across the knowledge economy at once. The challenge for leaders isn't understanding that disruption is coming. It's understanding that the timeline has compressed.
What this means and where the value moves next
The market just told you, in $400 billion worth of price action, that the operating model you're running is being repriced. Not in five years. Now.
The tools launched this month can review contracts, build production software, run sales workflows, plan and execute marketing campaigns, and deploy AI agents across your enterprise tech stack. The capability is here. The question is whether your organization is built to absorb it, or still treating AI as a side project.
"Move at your own pace" is comfortable advice. It's also wrong. The pace is being set for you. By the 5% pulling away. By AI-native companies hitting $1 billion in revenue with 300 people. By a market that just wiped $400 billion off the board in a week because it can see what's coming, even if most boardrooms can't.
Every quarter you spend in pilot mode, the companies that committed are compounding. Their systems are learning. Their cost structures are improving. Their speed is increasing. Yours isn't.
The $400 billion wasn't destroyed. It was transferred. And understanding where it went reveals where the opportunity lives.
Sacks' framework presents every technology company, every startup, every solutions provider with a strategic fork. You're either building the new intelligence layer, the AI-native workspace that spans across enterprise tools, data, and workflows, or you're part of the infrastructure underneath it. Both can be valuable positions. But you have to know which one you're in and build accordingly.
If you're in the legacy layer, the move isn't to bolt an AI copilot onto your existing product and hope the market gives you credit. The market just told you it won't. The move is to fundamentally rethink your business model. That could mean shifting from per-seat pricing to outcome-based pricing, the way Salesforce's Agentforce charges $2 per conversation instead of per user. It could mean opening your platform so that AI agents can operate on top of it, making yourself essential infrastructure for the new layer rather than competing with it. It could mean repositioning from "tool" to "system of record" and doubling down on the data gravity and workflow entrenchment that makes you genuinely hard to rip out.
If you're building the new layer, the opportunity is the largest in a generation. But building at this layer doesn't necessarily mean becoming the layer itself. For many companies, whether you're a data company, a workflow-specific platform, or a vertical solutions provider, the opportunity is understanding how you fit within the stack. How you interconnect. How you become an essential module within the intelligence layer that enterprises can't operate without. Some companies will be the orchestration point. Most will be critical components of it. Either way, the mentality is the same: you're building for the new architecture, not defending the old one. The companies that get this right will capture the value that the legacy application layer is shedding. That's the $400 billion story. It moved from the old layer to the implied valuation of the new one. And the new layer is still being built.
This is not about running pilots. It's about rebuilding the growth engine from the operating layer up. The organizations that treat this moment as a signal to fundamentally rewire how they operate will capture disproportionate value. The ones that wait will wonder what happened.
This is where I spend my time. The companies in our portfolio are building at this layer, AI-native companies designed to deliver compounding enterprise growth from the operating system up. Not tools that sit on top of old workflows. Systems that replace the workflow entirely and get better every day as a function of running.
The window is open. It won't stay open forever.
Jason Hauer is CEO and Founder of HauerX Holdings, backing and building AI-native companies that deliver compounding growth for the now.
