Anthropic, Atlassian, VCs Agree: Skip the New AI Stack

Anthropic, Atlassian and Scale Venture Partners converged on the same message at SaaStr AI 2026: build the skills layer, not a new stack. Scale's Rory O'Driscoll put a number on the stakes. Hyperscalers have committed 688 billion dollars in AI capex against 110 billion in revenue, with no breakeven until 2031 or 2032, per his SaaStr comments.

He said foundation models need 15 to 17 percent of knowledge-worker wages, more than 25 percent for developers, just to break even. That math puts pressure on whoever sits between the model and the paycheck.

Anthropic's own numbers back the pivot. 54 percent of new enterprise logos arrived self-serve in 2026, after account executives spent 70 percent of their time on internal process, Anthropic said.

Atlassian reported more than 5 million users on its AI features. Juniors are 19 to 30 percent more likely to use AI; seniors are twice as likely to refine its output rather than accept it raw.

The tell is Anthropic's own stack

Anthropic threading Claude through Clay, LeanData, Salesforce, Gong, Slack and Intercom, rather than building competitors to any of them, is the counter-evidence to rip-and-replace. A model company that believed the stack was obsolete would be building the stack. Instead it's building the skills layer that sits on top of it, which is a bet on distribution over ownership.