Nvidia backs Baseten in $300 million round as AI ‘inference’ becomes the new battleground

Nvidia has taken a significant stake in Baseten, investing $150 million as part of a $300 million funding round that values the AI infrastructure company at $5 billion. The deal highlights a shift in investor and enterprise focus: the next competitive frontier is not only building ever-larger models, but deploying them efficiently at scale, where cost, latency, and reliability can determine whether an AI product succeeds.

Baseten positions itself as an inference platform—software and infrastructure that helps companies run large models in real-world settings. Inference is where customer-facing AI experiences live: chat systems responding instantly, copilots generating code suggestions, or enterprise tools summarizing documents on demand. As organizations move from experimentation to production, the performance and stability of inference stacks increasingly shape user satisfaction and unit economics.

The financing round was led by major investors and comes amid heightened competition among chip makers, cloud platforms, and specialized startups to own inference workloads. Industry analysts have argued that inference could become the majority of AI compute demand over time, because production usage can dwarf training once a model is widely adopted. That expectation is pushing companies to optimize the entire serving pipeline, from hardware to scheduling to model compression.

For Nvidia, the Baseten bet is strategic. Nvidia’s GPUs remain a core standard for AI development, but the company also faces pressure from alternative architectures and purpose-built inference accelerators. By backing a leading inference platform, Nvidia can strengthen its ecosystem and maintain influence over how models are deployed across cloud and hybrid environments.

The funding also signals how quickly the infrastructure layer is consolidating into winners with scale, capital, and distribution. If the “app layer” of AI is crowded, the platforms that make AI usable in production may be where durable moats form. The next year will test whether Baseten can translate fast growth into a broader default choice for enterprises—and whether Nvidia’s expanding portfolio of partnerships and investments can keep it ahead as inference demand accelerates.