Why relying on AI models could hurt software profits

This article summarizes an episode of venture capital’s video series featuring Derek Anderson, managing partner at Connor Group.

Image credit: Timmy Loen

Derek Anderson, a managing partner at Connor Group, warns that the AI boom may be degrading the financial quality of software companies. By relying heavily on external AI models, firms risk losing operational independence and weakening their high-margin business models.

Rather than blindly adopting external AI platforms and risking margin collapse, leaders need to recognize these financial risks. He notes that companies aware of the “foundational model tax” and focused on regulated administrative automation are more likely to build sustainable businesses.

The foundational model tax is here

The fundamental economics of the software industry are changing rapidly. For new AI-driven companies, the usage-based costs tied to external AI models introduce an unpredictable expense. This fundamentally destabilizes long-term contracts for both the vendor and the customer.

Peter Harris from Venture Capital Podcast says customers are moving away from five-year deals and toward shorter, usage-based contracts. “[Many] AI-first platforms have to pay that tax back to [the foundational model] platforms. And so their cost basis is even more usage-driven.”

This shift from predictable subscriptions to variable billing adds pressure to operations and revenue planning. Anderson notes that usage-based contracts also create accounting challenges, especially around revenue recognition.

Depending on one platform creates survival risks for new founders
Beyond the financial pressure, founders may also be giving up control of their core product. Building on top of an external platform can create business risks that were less common in traditional software.

Anderson explains the risk, “If you’re dependent on OpenAI or on whatever platform you’re using, then what happens when they release something that just totally disrupts what you’re doing, puts you out of business, or whatever? Do you trust that they’re using your data in the way that they should?”

Wall Street sees the writing on the wall

This fragility also affects valuations. Institutional investors are increasingly questioning whether the highly profitable software business models of the past decade can survive this new AI-driven paradigm.

The traditional software industry was built on the principle of expanding margins: as a company grew, the marginal cost to serve each new customer approached zero, making the business increasingly profitable. The compounding costs of querying external AI models threaten to completely reverse that economic model.

Anderson explains, “You’re paying a lot of money back to them [the foundational model platforms]. So, it has a potential downward compression on your margins over time. Maybe with a traditional SaaS, you expect to have upward margin [growth] over time as you service it. This could definitely be a risk of the revenue just having a lower margin over time.”

Losing predictability causes people to rethink the value of software stocks
Public market investors liked SaaS because of long-term contracts, high margins, and low churn. The rise of foundational model costs weakens those advantages.

Venture Capital Podcast’s Jon Bradshaw  says SaaS used to offer locked-in revenue, high margins, and independence from other platforms. “You remove all of those advantages, and I don’t think the revenue is anywhere near as high quality as it used to be. And I think that’s part of the reason why SaaS stocks in the public markets are down so much.”

Why it’s so hard to quit legacy software

The unseen moat of regulation

Finance teams are a tough sell

Where automation actually works