Here is a counterintuitive finding from the field: almost no customer asks for self-hosted. They do not say SaaS does not work for us. And yet self-hosted is where the clearest market pull shows up, and it deserves to be the top strategic priority.
The reason is latent demand. When push comes to shove, a large company wants full control over its infrastructure. Full provenance over everything. That desire for control is one of the primary reasons enterprises decide to build agent systems in-house in the first place. So self-hosted is not a feature request you wait to receive. It is a forward bet that removes the reason to build internally, which is your real competitor.
The economics make the case even stronger. SaaS forces you to resell tokens and compute, which is a brutal game when competitors are giving both away. Self-hosted flips this. The customer pays for their own tokens. The customer pays for their own compute. What you deliver is pure software. Cleaner margins, no race to the bottom on inference costs.
Then there is stickiness, which compounds in a way that is easy to underestimate. Consider a customer that wants a separate deployment for each of its business units. Multiple installs of the entire product, run by separate teams. The day they consider moving off, they are not facing one migration. They are facing several, across distinct org structures. That is not a vendor they switch away from casually. The cost of leaving grows with every deployment.
The trap is treating self-hosted as a niche checkbox because it is only a small slice of current logos. That math misreads the future. The deepest enterprise accounts, the regulated banks with on-prem data centers who cannot touch public cloud at all, are unreachable without it. To serve them you need a cloud-agnostic foundation you own, deployable down to bare metal.
The takeaway for any agent company chasing enterprise revenue: do not wait for self-hosted to appear in your feature requests. It will not. Build it as a deliberate forward bet, price it as the pure-software product it is, and let the stickiness compound. It is the difference between a tool teams try and a system they cannot leave.
Key takeaways
- Enterprises rarely request self-hosted directly, but they want full control of their infrastructure when it matters most.
- Self-hosted is pure software margin because the customer pays for their own tokens and compute.
- Every separate deployment multiplies stickiness, making the product nearly impossible to rip out.
FAQ
If customers don't ask for self-hosted, why prioritize it?
Because the demand is latent. When the decision gets serious, large companies want full control and provenance over their infrastructure, which is one of the main reasons they consider building in-house at all. Offering self-hosted removes the build temptation.
Why is self-hosted more profitable than SaaS for an agent company?
In SaaS, the vendor resells tokens and compute, often competing against rivals giving those away. In self-hosted, the customer pays for their own tokens and compute directly, so the deployment is pure software with far cleaner margins.
Related Essays
The Three Services That Sell an Agent Platform
The path to enterprise agent adoption runs through three distinct service engagements, and each one exists to pull the customer deeper into the product.
Follow No Goose Absolutely
Design partner programs prevent enterprise AI companies from building three products when they should have built one.
The Forward-Deployed Engineer Is the Wedge
Forward-deployed engineers are not a cost center. They are the fastest path to turning enterprise pilots into six-figure accounts.
Drowning in pull requests that need your review? Try Tembo Review, a beautiful AI-assisted PR review tool unlike anything you’ve used.