AI Pulse · Business Models & Winners

AI-native SaaS companies are climbing toward healthier gross margins, but they’re still nowhere near the economics of traditional software. According to aggregated data from Solvimon, margins have improved from around 41% in 2024 to roughly 52% in 2026, with a likely ceiling settling between 60% and 65%. Classic SaaS businesses, by comparison, routinely run at 70% to 90%.

The gap comes down to cost structure. AI-first companies spend 40% to 50% of revenue on model hosting and inference, the compute that powers every query, prediction, or generation. Traditional SaaS spends 15% to 20% on cost of goods sold, mostly hosting static databases and serving up screens. When your product burns tokens instead of serving cached pages, the unit economics look very different.

This explains why usage-based pricing has become the default in AI products. Flat subscriptions don’t cover the variance in compute costs across customers. A power user running hundreds of queries a day costs you real money in a way that a typical SaaS seat does not. So companies pass variable costs through to customers, which keeps margins predictable but also makes revenue harder to forecast.

For operators, this matters in two ways. If you’re buying AI tools, expect usage pricing and watch your bill as adoption spreads internally. If you’re building or investing in AI products, plan for a lower margin profile than you’d tolerate in traditional software. The 80% margin SaaS playbook does not apply here, and the cost curve is not collapsing fast enough to change that soon.

The companies that will win are the ones that build pricing models and cost controls into the product from day one. This is exactly the kind of operational intelligence we build into the Omni Command Centre, where usage, cost, and margin visibility sit alongside everything else you need to run the business. Treating AI spend like SaaS spend is a fast way to blow through budget without realizing it.