Something quietly remarkable is happening. The distance between an idea and a working product has never been shorter.
Empathize, Define, Ideate, Prototype, Test. That entire design-thinking loop that used to fill a quarter for a cross-functional team can now move at the speed of a small one's judgment. The craft hasn't gotten easier. Taste, knowing what to build, and the call on what to leave out still do the heavy lifting. What's changed is how quickly that judgment can land in real product. The ceiling on what a thoughtful team can create has lifted in a way most of us are still catching up to.
The Market
The iShares Expanded Tech-Software ETF (IGV) is down 24% year-to-date as of February 20, 2026. Forward P/E multiples across the software sector have moved from 39x to 21x in twelve months. ServiceNow beat earnings estimates and the stock still pulled back 55% over the year. Atlassian is down 70%. monday.com, 75%.
But the reset isn't even. The companies in this analysis are some of the strongest software businesses ever built, and investors are sorting them carefully rather than selling them off as a category. I mapped roughly twenty SaaS companies across two dimensions, coordination complexity and AI revenue exposure, and the spread is wider than I expected:
- AI-Native Infrastructure: averaging just −1% over one year
- Domain Defenders: averaging −18%
- Platform Transformers: averaging −47%
- In Transition: averaging −58%
A 57-point spread between the best and most pressured categories. Investors are drawing a clear distinction between businesses where AI is a tailwind and businesses where AI changes the shape of the business model itself.
The Bifurcation
Companies whose products power AI workloads (databases, observability, streaming) are basically flat. Companies with deep vertical moats in security and compliance are holding up. Horizontal SaaS that coordinates work is where the reset is sharpest.
A lot of what software-of-record products do today is administrative glue, the routing, the reminders, the status rollups. That layer is increasingly handled inside AI agents and assistants, so tools built primarily on top of it have to rethink what they're really for. The substantive parts of work, the design, judgment, building, conversations that move things forward, aren't going anywhere. The glue around them is changing.
The Demand-Side Moat Holds
The Bureau of Economic Analysis publishes national accounts data on software investment by type (NIPA Table 5.6.5). It shows self-built ("own-account") software as a share of total software spending has declined from roughly 30% in the 1990s to about 15% today.
Even as development tools improved dramatically every decade, from mainframes to PCs to cloud to low-code, companies subscribed to more software and outsourced more development, not less.
The Economist and the Bank of France have documented a complementary finding: a 10% decline in software prices leads to roughly 20% more spending on software. Price elasticity greater than one. Cheaper software means a bigger market.
Enterprises with Cursor and Claude can build domain-specific software fast. But the harder thing to copy is the talent flywheel, the retention loops, and the equity-driven culture you find at AI-native firms where engineers eat, sleep, and breathe the product. The demand-side moat probably holds.
Where the New Energy Is
The pressure on SaaS isn't really coming from customers building their own tools. It's coming from where new energy usually comes from in software: small teams of motivated builders.
What's actually changed is the leverage curve for engineers. Strong engineers now ship 10x to 100x more than they could three years ago, because the parts of the job that used to bottleneck them (boilerplate, integration plumbing, the fifth refactor) are largely automated. Peter Steinberger built OpenClaw, an autonomous AI coding agent, mostly solo. It hit 140,000 GitHub stars in three months, more than React accumulated in eight years. OpenAI acqui-hired him for it.
This isn't a story about teams shrinking. It's a story about what becomes possible when good engineers stop being held back by the unglamorous parts of the work. AI-native startups are organized around that leverage from day one: equity incentives, shipping culture, a clear mission, real skin in the game. The interesting question for incumbents isn't whether they can match the speed. It's whether they can rebuild that kind of culture inside a 10,000-person company.
What Still Defends Great Software Businesses
If the new energy is on the supply side, what holds up for the businesses already in the market? A few things, and they're substantial.
Distribution. Selling to Fortune 500 procurement is still hard, in the way that any work involving humans, trust, and risk is hard. Sales cycles, security reviews, compliance, legal. AI will help here over time, but the relationships are real assets.
Data network effects. Products that get smarter with scale, where more customers make the product better for everyone, compound in ways a new entrant can't replicate quickly.
Workflow integration. The switching cost isn't the software itself. It's the hundreds of integrations, automations, and customizations enterprises have built on top of it. That's accumulated work, and it has value.
Trust and compliance. In security, healthcare, and finance, the product isn't just the features. It's the audit trail, the certifications, the regulatory relationships, the institutional trust built over years. None of that is easy to copy.
Notice what isn't on the list: features, UI design, or the code itself. Those are exactly the things AI makes easier to replicate, which is liberating in its own way. It pushes companies to compete on the things that actually matter to customers.
What to watch
The reset is real, but the thesis driving it (that AI will destroy SaaS) is too blunt. Companies with distribution, data gravity, and enterprise trust will likely come through fine, often better than fine, because they get to redirect engineering energy toward the parts of the product that actually compound. The companies with the harder road ahead are the ones whose value was concentrated in the glue layer that AI now handles. That's a real challenge, and it's also a chance to figure out what they can be useful for next.
The bigger picture: software is in one of those moments where the floor on ambition just dropped. The good engineers, good operators, and good teams in your life are about to do their best work. That's worth being excited about.
Sanej Bandgar is the founder of turingly.com.