The VC firm that wrote checks to Anthropic every single round since 2023 is about to print money on a scale even Sand Hill Road hasn't seen in years.
The Summary
- Menlo Ventures is preparing for its largest return ever as Anthropic files for IPO, with partner Matt Murphy confirming the firm backed every funding round since 2023
- The conversation in AI investing has shifted from model performance to compute efficiency and actual delivery
- Menlo's bet represents conviction in Anthropic's approach to AI safety and scalable infrastructure, not just raw model capabilities
The Signal
Menlo Ventures didn't just buy a lottery ticket. They doubled down every time Anthropic passed the hat. That kind of conviction investing, round after round since 2023, tells you something about what they saw that others might have missed.
Matt Murphy's comments on Bloomberg Tech signal a turning point in how serious money thinks about AI infrastructure. The narrative has moved past "who has the best model" to "who can deliver compute that actually works at scale." That's not a subtle shift. It's the difference between a research lab and a business.
"The conversation around AI is moving beyond models to delivering compute efficiently."
Anthropic's S-1 filing matters because it's the first major AI company to go public since the current wave began. The numbers will set benchmarks for how markets value:
- Training infrastructure versus inference costs
- Model performance versus deployment efficiency
- Research capabilities versus enterprise revenue
The timing is deliberate. Anthropic waited until they had proof of enterprise traction, not just impressive demos. Their Constitutional AI approach and Claude's enterprise adoption gave them a story beyond "we made a smart chatbot." They're selling infrastructure for the agent economy, the actual compute layer that runs when your AI does real work.
For Menlo, this validates a thesis about picking infrastructure over features. They didn't chase the flashiest model benchmarks. They bet on the team building the rails that other companies would need when agents moved from demos to production. Murphy's comfort discussing compute efficiency, not just model capabilities, shows where the real money thinks the value accrues.
The Implication
Watch the S-1 numbers closely. Anthropic's public filing will show the actual unit economics of running frontier AI at scale. If their compute costs per query are dropping faster than their competitors, that's your signal that infrastructure efficiency matters more than raw parameter counts.
For anyone building on AI: the majors are going public, which means the infrastructure layer is maturing. That's your green light to build serious applications. The picks-and-shovels phase is ending. The building phase is starting.