While everyone watches OpenAI and Anthropic fight over whose model is smarter, the real money is quietly flooding into the companies that help everyone else actually deploy AI without hiring a team of ML engineers.
The Summary
- Modal Labs is raising at a $15B valuation, triple its price from four months ago; Baseten is in talks at $26B
- Infrastructure for running AI models is now more valuable than most AI labs themselves
- The deployment layer, not the model layer, is where investor conviction sits right now
The Signal
The valuations tell you where the actual business value lives. Modal at $15 billion and Baseten at $26 billion means these AI infrastructure platforms are being priced higher than most frontier model companies. That's not hype. That's the market pricing in a truth most people still miss: the hard part isn't building models, it's running them in production.
Modal tripling its valuation in four months is a speed typically reserved for consumer apps that go viral. But this isn't consumer. This is infrastructure. The pace signals something specific: companies that tried to deploy AI in the last year hit a wall, and Modal became the sledgehammer.
"Infrastructure for running AI models is now more valuable than most AI labs themselves."
What Modal and Baseten actually do is turn model deployment from a six-month engineering project into an API call. You point them at your model, they handle the compute scaling, the GPU orchestration, the cost optimization, the uptime. For companies trying to ship AI features without becoming AI companies, that's the difference between shipping and not shipping.
The valuation jump also maps to a broader shift. Six months ago, every company wanted to build their own AI. Now they want to ship AI products without thinking about infrastructure. Modal and Baseten are the abstraction layer that makes that possible.
Here's what the numbers mean in practice:
- Companies can deploy models without hiring ML ops teams
- GPU costs get managed programmatically instead of manually
- Inference scales automatically instead of breaking at 10x traffic
The $26 billion Baseten valuation is especially telling. They focus on inference, the unglamorous work of actually running models at scale after training. Inference is repetitive, expensive, and invisible. It's also where 90% of AI compute spend happens once a product is live. Investors are betting that whoever controls the inference layer controls the economics of AI deployment.
The Implication
Watch for a new class of AI company to emerge: businesses that aren't AI-first but ship AI-native products because the infrastructure layer finally works. The tooling is good enough now that AI features become table stakes, not differentiators. That's when the real economic restructuring starts.
If you're building anything that touches AI, the question isn't whether to use infrastructure platforms like Modal or Baseten. It's whether you can afford not to. The companies raising at these valuations are the ones that solved the deployment problem everyone else is still wrestling with.