Is SaaS Dead? No. The SaaS Renaissance Is Just Starting

SaaS is not dead. What died in February, when roughly $285 billion in software market value evaporated in a single 48-hour sell-off, was the monolith: one bloated product built for a median user who does not exist. Software is entering its best decade, because AI has collapsed the cost of building it, and cheap production unlocks a long tail of custom software made for one company, one team, or even just one person. Space (sp-ce.co), the AI-native brand agency in New York, calls this the SaaS renaissance.

Here is the argument.

Why does everyone think SaaS is dead?

The bear case has two parts. First, AI drives roughly a 10x decrease in software production costs. Second, cheap production floods the market with competitors, both VC-funded startups and in-house builds, which destroys pricing power. Wall Street named the sell-off the SaaSpocalypse, and for the first time in history, software traded at a discount to the average S&P 500 earnings multiple.

Both premises are true. The conclusion does not follow. Lowering the cost to produce software has never meant software revenue shrinks. History says the opposite.

The Jevons Paradox, or why the software market grows from here

In 1860s Britain, economists assumed more efficient coal engines would make coal reserves last longer. William Stanley Jevons predicted the reverse: efficiency lowers cost, lower cost unlocks demand, and demand grows faster than efficiency saves. Britain burned through its coal faster. The same pattern repeated with datacenters, where a modern rack delivers 20,000 times the compute of a 2005 rack and the world still cannot build capacity fast enough.

Bob Morse of Strattam Capital applied this to software in Crunchbase News, and his sizing is the number to remember. The US spends about $0.5 trillion a year on business software and roughly $10 trillion on knowledge workers. Software today captures about 5% of the work it supports, because knowledge work has always been rationed: experts are expensive, scarce, and hard to manage. AI-powered software ends the rationing by selling the work itself, not just tools for the workers. That is a 20x larger market than the one investors just repriced.

The end of average software

Software became a monolith for one reason. Building it was expensive. When a product costs $50 million to develop, it needs a million seats to pay back, so you build for the median user. The median user does not exist.

Feature bloat was never a design failure. It was the interest paid on serving everyone at once, every niche's must-have feature shipped as everyone else's clutter. The settings maze, the admin certifications, the six-week onboarding: all of it is the cost of one product pretending to fit a million different workflows.

A 10x drop in production cost collapses the minimum viable market. Software that needed a million users now pencils at a thousand, a hundred, one. Three things follow:

• Custom software gets simpler. Fit replaces configuration. A tool built for your workflow needs no onboarding maze because it only contains what you use.
• Missing features finally get built. Every industry has needs too specific to survive a product manager's prioritization meeting at a company serving forty industries. Those features now get shipped, because the software is built for you alone.
• The market extends instead of fragmenting. The long tail of software is new spend, not redistributed spend, because it serves demand the monolith never could.

The Model T lesson

Before 1908, cars were coachbuilt. A craftsman made yours by hand, it fit you perfectly, and only the rich could afford one. That was enterprise software in the mainframe era: bespoke systems for the Fortune 500.

Then Henry Ford built the Model T. One car, any color so long as it's black, price falling from $850 to around $260 as the assembly line compounded. Mass production served everyone by serving no one in particular. That is the SaaS monolith. Per-seat platforms with four thousand settings pages are the black Model T.

Act three belongs to Alfred Sloan. At General Motors he promised "a car for every purse and purpose": segments, colors, annual model years. By the late 1920s GM had passed Ford, and Ford had to shut its factories for months just to retool. The car market did not shrink when customization arrived. US registrations nearly tripled through the decade.

Sloan could only segment to five brands. AI finishes the arc he started, moving from a car for every purse to software for every company. Bespoke fit at assembly-line prices, for the first time in the history of any industry.

Who wins the SaaS renaissance?

Not the do-it-yourself crowd. Tomasz Tunguz of Theory Ventures calls this the golden age of AI applications and makes the practical case for vendors: AI applications demand ongoing specialist work in picking models, designing feedback loops, and running evaluations, and most companies will not staff a team for every workflow. The tuning is better assigned to vendors who spread the cost across many customers.

Companies want the meal, not the cooking project. The renaissance belongs to small, deep vendors and studios shipping custom software with maintenance attached, and to the businesses that hire them. Incumbents still shipping the black Model T get repriced. That already happened.