SaaS Budgets Flip

AI software companies are increasingly spending like mature software vendors: more on selling than building. Jason Lemkin pointed to a mix near 50% of budget for sales and marketing and about 20% for engineering and product at top performers. (saastr.com) That split lines up with what public software companies already report. HubSpot said full-year 2025 revenue grew 19% to $2.8 billion, while Klaviyo said 2025 revenue grew 32% to $1.2 billion, giving both companies more room to fund go-to-market teams at scale. (ir.hubspot.com) (investors.klaviyo.com) Klaviyo put the mix in unusually clear numbers at its September 25, 2025 investor day: 34% of first-half 2025 revenue went to sales and marketing, 18% to research and development, and 12% to general and administrative costs. The same slide showed a longer-term target of 33% for sales and marketing and 20% for research and development. (sec.gov) The pressure behind that shift is simple: AI tools are getting cheaper to build on, but customers still have to be found, persuaded, and kept. Bessemer Venture Partners said the 2025 Cloud 100 passed $1.1 trillion in combined value, with AI companies making up 42% of the list, up from 21% in 2024. (bvp.com) That has changed what “scaling” looks like for software companies over the past year. Bessemer said record 2025 valuations were shaped by an “AI-driven transformation,” while Jason Lemkin’s recent SaaStr writing has argued that AI is now changing sales and go-to-market as much as product development. (bvp.com) (saastr.com) Public filings show the old software model never stopped favoring customer acquisition once a product reached scale. HubSpot’s 2025 annual report said the company expects sales and marketing expense to rise in absolute dollars as it continues to develop its sales and marketing teams, even as that cost should decline over time as a share of revenue. (sec.gov) The difference in 2026 is that AI companies can ship product improvements faster with smaller engineering teams than earlier cloud companies could. Lemkin’s argument is that once demand appears, the bottleneck moves from writing more code to converting interest into contracts and renewals. (saastr.com) Not every company will follow the same playbook. Bessemer’s 2025 report said the sector is also in an “arms race” for AI talent, with top researchers commanding nine-figure pay packages, which means some model builders and infrastructure companies will keep engineering costs far above application-software peers. (bvp.com) For buyers and investors, the budget flip is a clue about where the AI software market is settling. The companies acting most like scaled software businesses are putting more dollars into distribution, partnerships, and customer success than into adding another block of code. (saastr.com)