SaaS Is Dead. Long Live SaaS! AI And The End Of The Rationing Of Knowledge Work

By now, the headline will be familiar to most SA国际传媒 News readers: SaaS is Dead.

The market believes software businesses can鈥檛 charge premiums anymore and it predicts slowing growth indefinitely.

There are two reasons. First, AI powers a 10x decrease in software production costs. Second, these AI capabilities enable a huge wave of new competitors, both VC-funded startups and in-house solutions.

Reduced software production costs and rising competition, they say, will eliminate software鈥檚 pricing power. Public software stocks traded down 20% this year through mid-May, and for the first time in history, software trades at a discount to the average S&P500 multiple on earnings. SaaS is dead.

It is true that AI has brought falling costs and rising competition. But it does not follow that SaaS is dead. Lowering the cost to produce software does not mean that software revenue will shrink. In fact, history suggests the opposite.

In certain cases, efficiency begets consumption. This is the lesson of the . It worked for coal engines, it worked for data centers, and it will also work for AI-powered software.

The Jevons Paradox

Let鈥檚 start with coal. In 1860s Britain, many worried about burning through the country鈥檚 coal resources too quickly. Conventional wisdom said that developing more-efficient coal-burning engines would make the coal last longer.

But economist William Stanley Jevons recognized that more coal-efficient engines would cause an increase in demand for coal energy with the result that Britons would burn through their coal more quickly, not less.

Jevons was right. When greater efficiency produced lower costs, it also unlocked enormous new demand. This consumed coal reserves faster, not slower.

Twenty-five years ago, I joined a buyout firm during the 2001 dot-com crash. My first investment was in a troubled datacenter company. Exodus Communication, which reached a peak market cap of $32 billion, then went through bankruptcy twice as datacenter demand continued to fall.

In 2004, I recommended that our firm acquire that datacenter business out of the second bankruptcy for $200 million and merge it into a competitor named Savvis.

At the time, the market considered datacenters a shrinking industry. Dot-com companies were pulling racks of servers out of the sites, and datacenter floors were emptying out. Industry analysts forecast that given Moore鈥檚 law about the exponential growth of chip capacity and increasing server power density, a single rack in 10 years would deliver what it took 100 racks to deliver in 2005, and that in 20 years, a rack would deliver what 10,000 racks delivered in 2005. Conventional wisdom said that more-efficient chips would require less datacenter floorspace over time.

Our thesis that demand for datacenter floorspace would grow was not a popular opinion at the time. If 10,000 racks in 2005 would be replaced by just one rack in 2025, didn鈥檛 the U.S. have plenty of datacenter floorspace already?

was running advertisements showing a room full of servers replaced by one mainframe. One skeptical investment committee member told me that this business had been through bankruptcy twice in two years, and that if it went through a third time, I would go with it.

Today, one rack can indeed deliver 20,000x the compute power of racks from 2005, and as everyone knows, far from having too much floorspace, we can鈥檛 build new datacenter capacity fast enough. Truly enormous latent demand for computing power was unlocked as rack efficiency increased. The Savvis story ended well too, sold six years later for $3.2 billion.

The Jevons Paradox was true for coal, and it was true for data centers. It will also be true for AI-supported knowledge work.

Knowledge work and market expansion

Twenty-five years ago, only the wealthy had access to personalized investment advice. In 1996, Nobel Laureate Bill Sharpe co-founded to bring personalized investment advice to anyone with a 401(k).

My firm was an investor, and I had the privilege of working closely with the company. At first it tried to sell advice about how to invest 401(k) funds, but only about 20% of employees were interested in taking advice and then managing their 401(k) positions themselves.

Financial Engines鈥� breakthrough innovation was to manage the 401(k) positions directly, not just advise. Employees could check a box: 鈥渄o it for me鈥�. The demand from people who previously had no access to this advice was beyond all expectation and did enormous good. I recall that an early customer was , whose tens of thousands of employees with an average age of 27 years had approximately 40% of their 401(k) monies in cash, 40% in stock of JCPenney (which would eventually file for bankruptcy in 2020), and 20% in everything else. Just moving them into sensible low-cost mutual funds appropriate to their age and other financial goals generated huge benefits.

Financial Engines went from zero to $169 billion in assets under management when it was acquired in 2018 for $3 billion.

The company delivered a service that is very similar to what we today would call agentic AI. The customer (an employee with retirement savings) was delegating a decision (invest my money) to a computer system, and the employee paid in a way tied to the outcome (~50 basis points on AUM).

Of course, the technology to deliver this was quite different, and this was a very narrow application. The lesson remains: Software enabled a massive efficiency in delivering knowledge work (in this case individual investment advice) and a huge latent market appeared to buy the service.

The end of the rationing of knowledge work

The increased cost efficiency of AI, like the increased cost efficiency of Financial Engines鈥� algorithms, allows demand to increase because it relaxes a supply constraint on knowledge work.

Across human history, even to today, knowledge work has always been rationed because it is supply constrained.

Knowledge workers take years of education and training, tend to want to live in high-cost places, over time want to work on only certain kinds of problems they find interesting, and require a lot of management to get along. That is why we pay them such high wages and do everything we can to make them more productive.

Business software is a tool to make knowledge workers more productive. The total business software market in the U.S. is on the order of $0.5 trillion per year, according to Gartner. The U.S. market for knowledge work, that is the amount paid to the 100 million knowledge workers in this country, is roughly $10 trillion, per numbers. Currently, we spend about 5% of the cost of knowledge workers on software tools to help them.

AI enables software companies to not just sell tools to knowledge workers, but to begin to sell the knowledge work outcomes themselves, as I have written about in prior SA国际传媒 articles.

Put those together: 90% cost compression in software development plus the ability to sell knowledge work. We know there is a huge latent demand for knowledge work, if only it were not so expensive and hard to access.

For the first time, millions of people and businesses who have never had access to a strategist, an analyst, a lawyer or a financial adviser are about to get one.

Software is far from dead. The increase in efficiency offered by AI will allow it to do much more for less, and just like a more efficient coal engine or data center, this will unlock huge latent demand for knowledge work.

Ultimately, this will increase revenue and the strength of software businesses that use AI to further improve knowledge workers鈥� productivity or deliver knowledge work outcomes directly. Software鈥檚 job today is to solve the problem of delivering this safely and reliably.

It was no small task for industry to learn how to manage knowledge workers who are human, and it will be just as big a task to learn how to manage those who are machine knowledge workers. That is the challenge.

But remember that today the market for knowledge work is 20x the size of the market for software. The scale of the prize for software companies is unlocking the latent demand for knowledge work that, if history is any guide, will dwarf today鈥檚 software market.

The market today fears that the efficiency delivered by AI will shrink the software industry. Exactly the opposite is true. AI will unlock massive latent demand for knowledge work, and the software market will explode. Long live software.

co-founded in 2014 and is managing partner. He has served on numerous private and public technology company boards, and currently is a director of , , , , and . Previously, he was a partner and member of the investment committee at . He also worked at and . Morse serves on the board of directors of and as member of the advisory board for the HMTF Center for Private Equity Finance at . He attended , graduating summa cum laude with a BSE, and , where he earned his MBA and was an Arjay Miller Scholar. Morse lives in Austin.

SA国际传媒