Baseten targets $1.5B raise to dominate AI inference infrastructure

If the reported numbers hold, Baseten’s raise would be one of the largest private rounds ever for an AI infrastructure startup, and it lands squarely in the hottest layer of the stack: inference. Training has gotten the hype, but long‑term cash flows will skew toward serving models cheaply, reliably and across many deployment scenarios. A $13 billion valuation for an inference router underscores how investors now view this layer as a tollbooth on the road to AGI-era applications.

Strategically, this kind of capital lets Baseten build out global capacity, aggressively subsidize usage and integrate deeply with both frontier and open models. That could tilt the balance away from single‑vendor stacks where model providers, cloud platforms and inference services are vertically integrated. Instead, enterprises might standardize on neutral routing layers that arbitrage between OpenAI, Anthropic, Google, Meta and open‑source models.

For the AGI race, hyperscaling inference infrastructure lowers barriers for specialized agents and domain‑specific systems that don’t own their own models. Cheaper, smarter routing effectively makes “good enough” models feel frontier in many workflows, which accelerates deployment of powerful AI across the economy even if base model progress pauses. The flip side is concentration risk: if a handful of well-funded inference platforms sit in the middle of most traffic, their design choices about safety filters, logging and model selection will quietly shape how frontier capabilities reach the real world.