The SaaS Reckoning

The Setup

On April 23, 2026, enterprise software had one of its ugliest days in recent memory. ServiceNow fell 17%, on track for the worst single-session decline in the company's history. IBM dropped 9%. Salesforce and HubSpot each fell about 9%. Adobe and Intuit lost roughly 7%. Workday slid 10%, extending its year-to-date decline past 45%. The iShares Expanded Tech-Software ETF (IGV) dropped roughly 5% on the session alone.

The trigger was earnings. ServiceNow narrowly beat estimates but flagged a headwind from deal slippage in the Middle East. IBM beat on both revenue and earnings but reported that growth in its software segment, anchored by Red Hat, had slowed to 11.3%. Neither result was catastrophic in isolation. But the market reaction was violent because it confirmed a fear that has been building all year: enterprise software companies are losing pricing power, and AI is the reason.

The numbers tell the story. The software ETF (IGV) is down roughly 18% year-to-date. The semiconductor ETF has surged more than 43% over the same period. Wall Street has coined a term for what is happening: the SaaSpocalypse. Since early 2026, more than $2 trillion in enterprise software market capitalization has been wiped out. The Claude Cowork launch on February 24 alone erased $285 billion in SaaS market value within 48 hours.

But here is the part the headlines miss. This is not simply a story about AI companies competing with SaaS companies. The deeper, more structural threat is that AI has collapsed the cost of building custom software to the point where SaaS customers are starting to ask a question they never asked before: why are we paying for this when we could build it ourselves?

The Math That Changed

The disruption story is simple once you see the two cost curves. The cost of buying SaaS has been climbing steadily. The cost of building a replacement has been falling off a cliff. Those curves crossed sometime in 2025, and 2026 is the year the market noticed.

On the buy side, enterprise software pricing has outpaced inflation for years. According to SaaS Capital's annual survey of more than 1,000 respondents, the median annual contract value for private SaaS companies climbed from $22,357 to $26,265 between 2023 and 2024 alone. That is a 17% jump in a single year. Gartner reports that enterprise software spending will increase roughly 15% in 2026, but much of that increase is going toward price hikes on existing contracts and AI infrastructure, not toward new SaaS purchases. Renewal conversations routinely feature 10% to 20% price increases, and many CIOs report that their SaaS costs are rising faster than any other line item in the IT budget.

On the build side, AI coding tools have compressed development timelines from months to days. Internal tools that would have cost $20,000 to $100,000 to commission from a consulting firm can now be built by a single developer (or in some cases a non-developer with engineering intuition) in weeks. The UK's National Cyber Security Centre documented a case where a startup received a SaaS renewal quote at double the current price, and one of their engineering leads replaced the core functionality in a couple of hours using AI coding tools. That is not a future projection. That is happening now, repeatedly, across industries.

The asymmetry is what matters. SaaS vendors are locked into a model where they have to raise prices every year to satisfy Wall Street's growth expectations. But every price increase makes the build alternative more attractive. The higher the renewal quote, the more rational it becomes for the customer to invest a few weeks of internal engineering time instead. SaaS companies are inadvertently funding their own disruption with every annual rate hike.

The Three Forces Eating SaaS Alive

Fortune recently co-hosted roundtables with senior business leaders in San Francisco and New York. The threat to SaaS was the recurring theme. They identified three structural forces that enterprise software companies can no longer ignore, and each one reinforces the others.

Force 1: Market vulnerability from captive customers. SaaS margins have been high for decades, propped up by switching costs that keep enterprise customers locked in regardless of satisfaction. Many companies pay substantial sums for ERP, CRM, and other business-critical platforms not because they love the product but because migrating away is painful. That kind of captive market is an invitation to disruption. SaaS vendors have raised prices annually for years, often well above inflation, because they could. When customers have no alternative, there is no ceiling on what you can charge. AI is creating the alternative.

Force 2: Collapsing barriers to entry. Building enterprise-grade software used to require enormous capital and specialized engineering talent. Today, AI coding agents have made it cheaper and faster by orders of magnitude. That means more competitors, more alternatives, and more pressure on the margins that SaaS companies have long taken for granted. The tool that a 10-person team and $2 million in seed funding would have built five years ago can now be prototyped by a single competent engineer in weeks.

Force 3: The rethinking of workflows. SaaS companies built their empires by standardizing processes across industries. One CRM for every company. One finance platform for every CFO. One ticketing system for every IT department. AI is inverting that logic. Instead of forcing your business into a vendor's workflow, you can build the workflow that fits your business. The standardization that made SaaS scalable is the same standardization that makes it replaceable.

These three forces compound. When building is cheap, switching costs drop. When switching costs drop, pricing power erodes. When pricing power erodes, the margins that justified those premium multiples can no longer support them. The market is not being irrational. It is being early.

The Build Cost Collapsed

This is the part of the story I can speak to personally.

Wealth Engine Pro is a full investment research platform. It scores 5,500+ stocks across financial health, trend strength, and valuation. It runs a daily AI analysis pipeline that identifies put-selling opportunities. It has brokerage integration, portfolio tracking, options analysis, ETF tools, a subscription billing system, and a conversational AI assistant with 29 integrated tools. The backend runs 53+ automated Python scripts on a daily cron schedule. The frontend is a Next.js application deployed on Vercel.

I built it. One person. No engineering team. No venture funding. No $2 million seed round.

Five years ago, this would have required a team of 8 to 12 engineers, a product manager, a designer, and at least 18 months of runway. The total cost would have been somewhere north of $1.5 million before the first customer saw the product. Today, AI coding tools, open-source infrastructure, managed databases, and serverless hosting have collapsed those barriers to the point where a single person with domain expertise and the willingness to learn can build what used to require a funded startup.

The key phrase there is domain expertise. The SaaS model was built on the assumption that the people who understand the problem (the customer) and the people who can build the solution (the engineering team) are different groups, and that a vendor sits profitably between them. AI is collapsing that gap. When the domain expert can build the tool directly, the vendor layer becomes optional.

Constellation Research put it plainly in their 2026 enterprise technology outlook: build will beat buy. The build versus buy debate is being reframed by AI agents that make it easier to create applications you used to purchase. CIOs are increasingly convinced that applications custom-built for their specific use cases offer a greater competitive advantage than off-the-shelf software. The sunk costs in existing enterprise systems will be abstracted away using agentic AI as a user interface.

Gartner projects that 35% of point-product SaaS tools will be replaced by AI agents by 2030. That forecast was published before the current selloff accelerated. Given the pace of AI capability improvement in the first four months of 2026, it may prove conservative.

The Credible Threat

Here is the part that the software companies really do not want investors to understand: customers do not actually have to build anything to destroy pricing power. They just have to be credibly capable of building it.

I have seen this firsthand. At my day job, we evaluated whether to build an in-house replacement for KnowBe4, a security awareness training platform. We put together a credible plan for an internal solution. We did not have to execute it. The moment KnowBe4 understood that the threat was real, they cut the contract price in half.

That is not an isolated anecdote. It is the new dynamic in every enterprise software renewal conversation. The customer walks in with leverage they never had before, because the vendor knows the customer's internal engineering team (or even a single senior engineer with AI tooling) could plausibly replicate the core functionality.

This creates three layers of pain for SaaS revenue, and none of them show up clearly in traditional churn metrics until the damage is done:

Layer 1: Some customers actually leave and build in-house. This is the visible churn. It is also the smallest category today, because most enterprises move slowly. But the direction is clear.

Layer 2: Some customers use the credible threat to negotiate massive discounts at renewal. The contract renews (so it does not count as churn), but the revenue per customer drops by 30% to 50%. This is the silent revenue compression that investors will not see until it flows through guidance.

Layer 3: Every renewal conversation now starts from a position of buyer strength instead of vendor lock-in. Even customers who have no intention of building anything benefit from the changed power dynamic, because the vendor cannot distinguish a credible threat from a bluff. The negotiating floor has shifted permanently.

A Fortune 50 company's internal memo surfaced in January 2026 outlining plans to cut Salesforce and ServiceNow license spending by 60% by year-end, replacing the functionality with API credits from AI providers. The memo did not describe a plan to eliminate software. It described a plan to eliminate the human workers the software was built around, and once those workers are gone, the per-seat licensing model collapses with them.

Jason Lemkin of SaaStr described this as a structural reversal of the growth engine rather than a cyclical slowdown in purchasing. The math is direct: if 10 AI agents can perform the work of 100 sales representatives, a company no longer needs 100 Salesforce seats. It needs 10, or perhaps none, depending on how the workflow is restructured.

The Vulnerable Middle

Not all SaaS is equally exposed. The disruption sorts the market into three tiers, and the pain is concentrated in the middle.

The top tier survives. Salesforce CRM for a 10,000-person sales organization. Oracle Financials running global supply chain operations. SAP ERP embedded in the operating rhythms of a multinational manufacturer. These are systems of record where the switching cost is measured in years and the compliance infrastructure alone would take millions to replicate. No CIO is going to ask an engineer to replace their general ledger system with an AI-coded prototype, regardless of how good the tools get. The switching costs here are not just technical. They are regulatory, contractual, and organizational.

The bottom tier was never worth building. A $50-per-month project management tool. A $20-per-seat document editor. A $10-per-user calendar app. These are commodity products where the annual cost is so low that the internal engineering time required to build a replacement exceeds the subscription savings. Nobody is going to spend two weeks of developer time to save $600 a year.

The middle tier is where the disruption hits. The $30,000-to-$150,000-per-year tools that do one thing reasonably well for a broad audience. Internal reporting dashboards. Workflow automation platforms. Security awareness training. Data pipeline managers. Customer portals. Onboarding systems. These products are expensive enough to justify the build investment and standardized enough that a competent internal team can replicate the core functionality with AI assistance in weeks rather than months. And critically, they are the tools where the vendor's one-size-fits-all approach creates the most friction with the customer's actual workflow.

The seat-based pricing model amplifies the vulnerability. According to Deloitte Insights, seat-based pricing is already declining from 21% to 15% of software vendors in just twelve months, and by 2030, at least 40% of enterprise SaaS spend is projected to shift toward usage-based, agent-based, or outcome-based pricing. The per-seat model worked because headcount growth drove software growth. AI breaks that link. When 10 AI agents perform the work of 100 employees, the company does not need 100 software seats. It needs 10, or perhaps none, depending on how the workflow is restructured.

The mid-tier SaaS companies that survive this transition will be the ones that own proprietary data their customers cannot replicate, serve as genuine integration hubs across multiple systems, or pivot their pricing models before the renewal conversations turn adversarial. The ones that are simply selling a standardized workflow at a premium price are running out of time.

Where the Incumbents Are Still Safe

The SaaS-is-dead narrative, like most binary narratives, is incomplete. There are significant categories of enterprise software where the incumbents have structural advantages that AI does not easily replicate. Being honest about where the moats are real is just as important as identifying where they are eroding.

Systems of record survive. Software that serves as the canonical source of truth for regulated processes (financial reporting, HR compliance, healthcare records) is not going away. The switching costs are not just technical but legal and regulatory. An AI agent can automate the workflow around these systems, but it cannot replace the system itself without recreating the compliance infrastructure. That is a multi-year, multi-million dollar undertaking that most enterprises will not attempt.

Network-effect platforms persist. Slack, Teams, and similar collaboration tools derive value from the number of people using them. An in-house replacement is useless if your customers, vendors, and partners are not on it. The same logic applies to marketplaces and ecosystem platforms.

Maintenance is the hidden cost of building. A university technology professor told Fortune that migrating mission-critical systems using AI-generated code is like changing flat tires on a car driving at 60 miles per hour. Enterprise platforms carry years of integrations, compliance infrastructure, and operational complexity that AI coding tools do not magically replicate. An in-house tool is nobody's product, which means it degrades unless someone owns it permanently.

AI agents still need SaaS. This is an underappreciated counterargument. AI agents require licenses for the tools they operate through: Salesforce, Microsoft 365, Slack. Deploying agents at scale could actually increase SaaS expenditure as digital headcount replaces human headcount but still needs the underlying platforms to execute workflows. The seat count drops, but the infrastructure spend may not.

The honest assessment is that foundational software layers will still be needed, and enterprises will resist becoming wholly dependent on any single AI vendor. But margins will likely compress. The shift from per-seat pricing to outcome-based or consumption-based pricing will erode the economics that made SaaS one of the most profitable business models in technology history.

What Could Go Wrong

The bear case on SaaS is compelling, but it is worth identifying the scenarios where the selloff turns out to be overdone.

AI hype exceeds AI delivery. The MIT Project NANDA study, based on 150 executive interviews and 350 survey responses, found that over half of enterprise AI budgets were going to sales and marketing tools where ROI was lowest. The gap between AI demos and AI production deployments remains substantial. If enterprises discover that building in-house is harder than it looks in a proof-of-concept, the pendulum could swing back toward buy.

The circular financing problem breaks. Nvidia invests in OpenAI. OpenAI commits to buying Nvidia chips. The hyperscalers spend $470 billion on AI infrastructure in 2026. A meaningful portion of the revenue that AI companies report is flowing in a loop between a small number of very large companies. If that circular dynamic cracks, the AI tools that threaten SaaS could themselves face a reckoning.

Security and compliance gaps emerge. In-house AI-built tools may introduce vulnerabilities that packaged SaaS solutions have spent years patching. Regulated industries (healthcare, financial services, government) may discover that the cost of compliance for custom-built software exceeds the licensing savings. The first major security breach attributed to AI-generated in-house code could reverse the build-versus-buy calculus overnight.

SaaS incumbents adapt faster than expected. The smarter software companies are already pivoting. Salesforce introduced Agentic Enterprise License Agreements, a flat-fee model designed for AI agent deployment. ServiceNow is embedding autonomous workflow AI into its ITSM environments. If incumbents can reframe themselves as AI platforms rather than seat-based software vendors, the disruption narrative becomes a transformation narrative, and the stocks reprice upward.

Every one of these scenarios is plausible. The question is whether they are probable enough to offset the structural forces already in motion. Based on the data, I believe the build-versus-buy shift is real and durable, even if the pace of the transition is slower than the most aggressive predictions suggest.

The Bottom Line

The SaaS business model was built on three assumptions: that building software is expensive, that customers lack the expertise to do it themselves, and that switching costs make the relationship permanent. AI has weakened all three.

The historical arc is worth remembering. Before standardized ERP platforms became widespread in the 1990s, businesses mostly used custom-built internal software. As Booz Allen's chief technology officer recently put it: companies built custom systems for the Army, for Bank of America, for BMW, each one tailored precisely to how that organization worked. Then the industry decided that custom was too expensive and standardized on SAP, PeopleSoft, and Oracle. It was painful. Every business person hated it. But the economics were overwhelming. Now AI is swinging the pendulum back toward customization, except this time the custom tools can be built in days instead of years, and the people building them are domain experts rather than armies of consultants.

Building is no longer expensive. AI coding tools have compressed the cost and timeline by an order of magnitude. Customers no longer lack the expertise, because the domain expert with AI assistance can now build what used to require a dedicated engineering team. And switching costs, while still real for systems of record, are rapidly declining for the mid-tier workflow tools that represent a huge portion of enterprise SaaS spending.

The most important shift is not happening in the stocks. It is happening in the renewal conversations. Every enterprise CIO now walks into a software negotiation knowing that the credible alternative to renewing is building. Even when they choose to renew, they are negotiating from a position of strength that did not exist two years ago. That power shift compresses revenue, which compresses margins, which compresses multiples.

This does not mean every SaaS company goes to zero. Systems of record, network-effect platforms, and deeply embedded compliance infrastructure will persist. But the mid-tier, the $30,000-to-$150,000-per-year tools that do one thing reasonably well for a broad audience, is where the disruption hits hardest. Those are the products that a competent internal team, armed with AI, can replicate and customize in weeks rather than months.

The software sector is not repricing on fear. It is repricing on a structural reality that the numbers now confirm. The divergence between software stocks and semiconductor stocks in 2026 is the market telling you, in real time, where the value is flowing and where it is draining. UBS described today's action in software as just the beginning of a longer repricing.

At Wealth Engine Pro, we follow the numbers, not the narrative. The narrative says AI is eating software. The reality is more nuanced: AI is giving customers the tools to build what they used to buy, and that changes the economics of every software contract, whether the customer builds or not. The credible threat is the disruption. And right now, the threat has never been more credible.