Two trillion dollars.
That's not a typo. That's how much market value evaporated from software stocks between February 3 and February 17, 2026. In fourteen trading days, the SaaS industry experienced the kind of reckoning that financial historians will write dissertations about — and it happened so fast that most business owners are only now realizing what occurred.
Forrester published a report titled "SaaS As We Know It Is Dead" — the kind of headline that would have been dismissed as clickbait three months ago. Bloomberg ran a piece arguing that Salesforce and other SaaS companies deserve their AI reckoning. JPMorgan fired back, calling the entire selloff "broken logic." Deutsche Bank pegged the total damage at $2 trillion.
Everyone has an opinion. Not everyone has the data. Here's what actually happened, where the smart money disagrees, and — most importantly — what this means if you're a business owner who depends on software to run operations.
Two Weeks That Shook the Software Industry
The trigger point everyone points to is February 3, when Anthropic launched Claude Cowork — a product that lets AI agents operate software the way a human employee would, navigating interfaces, filling forms, moving data between systems, executing multi-step workflows. Within 48 hours, $285 billion was wiped from SaaS stock valuations. Not because Claude Cowork was perfect. Because investors suddenly realized what it represented: the early infrastructure for a world where the software layer between humans and outcomes gets radically thinner.
The bleeding accelerated. By February 5, Operant AI had launched "Agent Protector" (security tooling for autonomous agents), and Anthropic's Opus 4.6 went live. By February 9, OpenAI was showing ads inside ChatGPT at a $60 CPM — a signal that AI interfaces were already generating enough traffic to support advertising. The old SaaS model charges per seat. The new model might not need seats at all.
February 12 brought the first wave of institutional analysis. Bloomberg Opinion didn't mince words: Salesforce and other SaaS companies deserve their reckoning. Search Engine Land reported that SaaS companies had seen a 53% drop in AI-related website traffic — but that Microsoft Copilot usage had grown 20x in the same period. Users weren't abandoning software. They were bypassing the traditional interfaces entirely.
The contrarian case showed up on February 13, when Apollo Global Management published a note arguing that AI would actually proliferate software usage by 10x, creating more demand, not less. The bull case: AI agents need more software endpoints to talk to, not fewer. It was a plausible argument. The market ignored it.
By Valentine's Day weekend, the real story crystallized. The full agent commerce stack — the technical plumbing that lets AI agents discover, negotiate with, and pay for services autonomously — shipped in roughly six weeks. The Universal Commerce Protocol (UCP) went live with Etsy and Wayfair as launch partners. WebMCP gave agents the ability to read any website semantically. The x402 protocol handled machine-to-machine payments. Shopify reported that orders originating from AI-powered search had increased 15x.
This wasn't theoretical anymore. The infrastructure for an agent-driven economy was being laid in real time, and the market noticed.
The Week Everything Broke
Monday, February 16 was the day the selloff went from "correction" to "existential renegotiation."
Polar Capital, a $12 billion London-based fund, used the word existential to describe the AI threat to software companies — a word fund managers avoid the way surgeons avoid the word "oops." The cumulative selloff hit $800 billion. A Base44 customer publicly disclosed terminating a $350,000-per-year Salesforce contract, replacing the functionality with AI-native tooling. Publicis Sapient, one of the world's largest digital consultancies, announced it was cutting SaaS licenses by 50%. Zoho CEO Sridhar Vembu called the moment "the pin popping this inflated balloon."
Salesforce and Workday — the bellwethers of enterprise SaaS — were both down more than 40% over the trailing twelve months. The iShares Software ETF (IGV) was down 22% year-to-date, with 8% of that coming in February alone.
Tuesday morning, February 17, brought the counterattack. JPMorgan published what became the most-cited rebuttal: the selloff was "broken logic," because investors were simultaneously selling software stocks and AI stocks. If you believe AI is going to eat software, you should be buying AI companies. If you believe AI is overhyped, you should be buying software companies. Selling both makes no sense. Goldman Sachs, Citi, and AWS CEO Matt Garman all pushed back on the narrative. PitchBook called the AI threat to software "overblown."
The market didn't care. CNN reported that the "AI scare trade" was spreading beyond software into insurance, real estate, logistics, and wealth management. Bloomberg reported that investors were "snubbing the dip," actively rejecting the counter-narrative and bracing for deeper disruption. Deutsche Bank strategist Jim Reid wrote that "a more realistic differentiation is now rippling into the broader economy with surprising speed."
Then, at 2:47 PM Eastern, Forrester dropped the bomb.
Forrester's Three-Tier Framework: The Analytical Backbone
The report — "SaaS As We Know It Is Dead: How To Survive The SaaS-pocalypse" — did something that most selloff commentary had failed to do: it differentiated. Rather than declaring all software dead or all software fine, Forrester broke the SaaS universe into three tiers with genuinely different risk profiles.
Tier 1: Horizontal Point-Solution SaaS — Challenged. These are the tools that do one thing across industries: project management apps, basic CRM, scheduling software, email marketing platforms, form builders. Forrester's assessment was blunt. These products compete on workflow convenience, not proprietary data or regulatory depth. An AI agent that can orchestrate the same workflow across native APIs doesn't need a middleman UI. If your product is essentially a pretty interface on top of commodity functionality, your moat is evaporating.
Tier 2: Vertical and Domain-Specific SaaS — Survives. Healthcare compliance platforms. Construction management systems. Legal document automation built on jurisdiction-specific regulatory knowledge. Forrester projected this segment growing from $133.5 billion to $194 billion by 2029. The logic: domain expertise, regulatory compliance, and industry-specific data create genuine defensibility. An AI agent can fill out forms, but it can't easily replicate twenty years of HIPAA compliance logic or the network effects of a construction supply chain marketplace. Anthropic CEO Dario Amodei reinforced this point, noting publicly that "there's a big gap between an AI model that works in a demo and one that works in a regulated industry."
Tier 3: Enterprise Suite SaaS Racing to Become AI Companies — Mixed. This is where it gets interesting. Salesforce, ServiceNow, Workday, HubSpot — the enterprise behemoths are all pivoting hard toward AI. Some will make it. Some won't. Forrester's sharpest observation was this: "As agentic control planes, orchestration layers, and semantic data layers increasingly sit above these suites, even AI-first SaaS vendors risk disintermediation." In other words, even if Salesforce successfully embeds AI into its platform, there's a risk that customers will use an orchestration layer above Salesforce that treats it as one data source among many — stripping away the lock-in that justifies premium pricing.
Goldman Sachs, in its own analysis, identified the moats that could protect enterprise software: proprietary data, network effects, regulatory compliance, and deep integration into customer workflows. The overlap with Forrester's Tier 2 characteristics is not a coincidence.
The Technical "Why": Agent Commerce Is Already Here
For business owners who don't track infrastructure protocols, the technical story behind the selloff is worth understanding — because it explains why the market is reacting now, not six months from now.
Between mid-January and mid-February 2026, a complete agent commerce stack went from concept to production:
- Model Context Protocol (MCP) — an open standard (originally from Anthropic) that lets AI agents interact with software services through structured, discoverable APIs. Think of it as a universal adapter.
- WebMCP — extends MCP to the open web, letting agents read and interact with any website semantically rather than scraping HTML.
- Universal Commerce Protocol (UCP) — enables agent-to-agent commerce. Already live with Etsy and Wayfair. An AI agent can discover products, compare prices, and complete purchases without ever opening a browser.
- x402 — a machine-to-machine payment protocol using HTTP 402 status codes. Agents can pay for API calls, content access, and services programmatically.
- adMCP — the advertising layer for the agent economy. If agents are the new "users," advertisers need a way to reach them. This is that mechanism.
As one widely shared post on X put it (garnering over 1,000 engagements): "The entire internet is being rebuilt for a new type of user." That's not hyperbole. The infrastructure for agents to operate as autonomous economic actors — discovering services, evaluating options, executing transactions — went from vaporware to production in six weeks.
This is why the selloff accelerated so dramatically. It's not just that AI could disrupt SaaS in theory. The plumbing is live. Agents are already placing orders on Shopify (15x increase), browsing Etsy through UCP, and paying for services through x402.
We covered this infrastructure shift when SaaStr rebranded around agentic workflows — but even that analysis from just weeks ago understated the speed at which production infrastructure would arrive.
The Two Camps: Who's Right?
The honest answer is: both sides have a point, and neither has the full picture.
Camp A — "SaaS is dead" (Polar Capital, Bloomberg, Forrester, Zoho): The existential threat is real for a specific category of software. Horizontal point-solution SaaS — the tools that survive on workflow convenience rather than proprietary data — face genuine disruption. When a $729/month stack of AI tools lets one person operate as developer, designer, researcher, and content creator simultaneously ($8,748/year), the per-seat SaaS model for commodity functionality starts to break. When a DTCP Ventures partner says publicly that "an entrepreneur who approaches a VC fund today with a SaaS startup won't even reach the pitch stage," the funding pipeline for these companies is drying up too.
Camp B — "This is overblown" (JPMorgan, Goldman Sachs, Citi, AWS, PitchBook): JPMorgan's "broken logic" argument is technically correct — selling both software and AI stocks simultaneously is incoherent. Goldman's moat framework identifies real defensibility: proprietary data, network effects, regulatory compliance. PitchBook's data shows that enterprise software spending is still growing in absolute terms. And there's a legitimate argument that AI agents need more software infrastructure, not less. Apollo Global's 10x proliferation thesis isn't crazy.
The key insight from SaaStr founder Jason Lemkin might thread the needle: even the "systems of record" defense — the idea that companies like Salesforce are safe because they hold the data — may not protect long-term. If an orchestration layer sits above your system of record and treats it as a commodity data store, the value shifts to the orchestration layer. Salesforce becomes a database with a very expensive UI.
Jamie Dimon may have captured the meta-dynamic best when he noted that "some asset prices are high, in some form of bubble territory." The $2 trillion selloff might be overdone in magnitude, but the directional signal — that the SaaS pricing model and value proposition are being renegotiated — is real.
Mark Cuban said it more bluntly: software as a packaged product is dying. What replaces it is software as a capability layer — consumed by agents, not humans.
What This Means for Your Business
If you're a small or mid-sized business owner reading this, the selloff itself doesn't directly affect you. Stock prices are Wall Street's problem. But the forces driving the selloff — AI agents that can replace human-operated software workflows — are your opportunity and your risk.
Audit your SaaS stack against Forrester's tiers. Pull up your monthly software spend. For each tool, ask: does this product survive because it holds proprietary data or regulatory expertise (Tier 2)? Or does it survive because no one has bothered to build a better alternative yet (Tier 1)? The Tier 1 tools are the ones most likely to be disrupted by AI-native alternatives within 12-18 months. That doesn't mean cancel them tomorrow — but it means you should be watching the alternatives.
Watch the agent commerce stack. UCP, MCP, x402 — these aren't buzzwords. They're production infrastructure. If your business involves e-commerce, services, or any kind of transaction, the way customers find and buy from you is changing. Shopify merchants who enabled AI search saw orders increase 15x. That's not a marginal improvement. That's a channel shift.
Treat AI as a force multiplier, not a replacement. The $729/month AI toolkit that gives one person the output of a small team isn't a threat to your business — it's a weapon for it. The businesses that thrive in the next 24 months will be the ones that equip their existing teams with AI tools, not the ones that wait for perfect clarity on which SaaS products will survive.
Don't panic, but don't ignore. The JPMorgan counter-narrative has merit. AI won't replace all software overnight. Regulated industries will move slowly. Complex enterprise workflows have genuine switching costs. But the direction is clear, and the infrastructure is live. The businesses that start experimenting now will be 18 months ahead of those that wait for a consensus view.
The Bottom Line
The $2 trillion SaaS reckoning isn't a single event — it's a repricing of the entire software industry against a new reality. The Forrester framework gives us the most useful lens: horizontal point solutions face genuine existential risk, vertical and domain-specific platforms have real defensibility, and enterprise suites are in a race to reinvent themselves before orchestration layers commoditize them from above.
Both camps — the bears who say SaaS is dead and the bulls who call it broken logic — are looking at the same elephant from different angles. The bears are right that something fundamental has changed. The bulls are right that differentiated software still has value. Where they converge is the most actionable insight: if your software's primary value proposition is a convenient interface for a workflow that AI agents can now execute directly, the clock is ticking.
For business owners, this isn't a crisis. It's a transition. The companies that navigate it well — by auditing their software dependencies, experimenting with AI-native alternatives, and treating AI agents as productivity multipliers — will come out of this stronger. The ones that assume their current stack will look the same in 2028 may find themselves paying premium prices for commodity tools while competitors operate leaner and faster.
If you're unsure where to start — which tools in your stack are vulnerable, how to evaluate AI-native alternatives, or how to build an AI integration roadmap — we can help you figure that out. This is exactly the kind of strategic assessment that separates businesses that ride disruption from those that get swept by it.
The SaaS reckoning isn't coming. It's here. The question is whether you're watching it happen or shaping what comes next.