Much has been written about the impact of LLMs on software, specifically the cost of writing new software. Tools like Chat GPT and Github Copilot make it easier than ever to write new software and build new products from scratch. They give a single developer the power of many.
Another way to think about it is software is getting less expensive. If 12 months ago the cost of building a given product was $100, today it’s less than that. And that cost is rapidly declining. How quickly the cost of writing new software goes to zero, if it ever does, is a contentious debate, one on which I won’t opine, but it’s trending down.
A year from now, it will cost a fraction of what it does today to build a piece of software.
As the cost of something goes down, the value of it follows. If the cost of writing new software is trending down then the value of software is as well. As software builders, that needs to fundamentally change how we think about company building.
Until now, most startups were pure software companies. This attracted investors as software businesses have high gross margins and can grow incredibly quickly. In theory, they’re supposed to be the most capital efficient businesses (we can debate how true that is). But these companies’ core asset is software, the products they build and sell are software.
Customers pay for software because they calculate that it’s less expensive to buy that product than it would be to build it on their own. But when it becomes less expensive to build a product than to buy it, customers will stop buying and start building more. This is what’s starting to happen to software.
The All In broo-haha discussed this on a recent episode where they talked about taking their most expensive SaaS tools and building them in house using tools like Copilot to save hundreds of thousands of dollars a year. While I believe they’re oversimplifying the challenges, this will still be a major trend over the next few years.
If software is depleting in value, how do you build a modern software company? A company’s core assets can’t be structurally depleting in value, otherwise the business is a sinking ship.
The best AI companies won’t be pure software companies, they’ll use software as leverage on top of some other product or asset. As marginal costs go down, more software will be written by companies for internal tooling specifically crafted for their needs. Some estimates suggest 50% of software is already written for internal toolings, this number will trends towards 100%. As companies build more software for internal tools, less will be bought from vendors and the value of software companies will depreciate.
So what does it mean for early stage founders to use software as a leverage multiplier? It means building a business that sells something other than software but uses software to make itself significantly more efficient. This can be done in one of two ways: (1) investing more in internal tools that make your team doubly efficient and (2) creating easier ways for your customers to interact with your product. But the product itself can’t be software, it needs to be something with some other inherent value.
There are exceptions to this. They’re software products that have some key technical innovation that would be impossible for a company to replicate. They’ll retain their value as it will continue to be less expensive for customers to purchase their solution than to build it internally. But they aren’t exceptions as much as they’re the end of the spectrum — the more complex your technical innovation, the more expensive it is to replicate, the less likely it makes economic sense for your customers to build their own version. So while these business won’t suffer as much, the same function applies. Once their innovation becomes easy, or cheap, to replicate, their fate will be similar.
To be clear, I’m not calling an end to SaaS. There are thousands of businesses who don’t have the technical ability to duplicate software products and will continue buying them off the shelf. These will continue to be significant markets. But for SaaS companies selling to tech companies without some key technical innovation, the future is bleak.
As early stage founders still trying to find our way through the maze that is product ideation — we need to keep this framework in mind. We need to find products that have some inherent value and use software as a way to make them better, cheaper or more accessible. Software should no longer be the only thing we sell.
