Baseten funding, valuation and investors

Overview

Baseten is a San Francisco-based AI inference infrastructure company that helps teams train, deploy, and serve models in production with optimized latency, autoscaling, and cost efficiency. Its platform includes Truss (open-source model packaging), Model APIs for instant access to open-weight LLMs, and dedicated single-tenant deployments with SOC 2 and HIPAA compliance. Customers use Baseten when hyperscaler defaults cannot meet p95 latency or per-token economics at scale. In January 2026, Baseten raised $300M Series E at a $5B valuation led by IVP and CapitalG, with NVIDIA, Altimeter, Battery Ventures, Greylock, and others participating — its third round in twelve months.

Why Baseten is interesting

Baseten hit $5B on Series E ($300M, Jan 2026) by selling inference as its own layer — like CDNs split from hosting — as enterprises shift from copilots to agent workloads burning 5–50x more tokens per interaction.

Product & use cases

Training and inference platform: deploy models via Truss CLI with autoscaling GPU infrastructure, TensorRT-LLM/vLLM optimizations, OpenAI-compatible endpoints, Model APIs for hosted open models, and dedicated multi-cloud deployments.

• Production LLM serving with stable p95 latency under load
• Fine-tuned open-weight model deployment (Llama, Mistral, Qwen, DeepSeek)
• Agent workloads with high token volume and elastic autoscaling
• Multi-model routing when enterprises diversify beyond one API provider

Key facts

• $300M Series E at $5B valuation (January 23, 2026)
• Third funding round in twelve months
• Truss open-source framework for model packaging and deployment
• SOC 2 Type II and HIPAA-compliant dedicated deployments
• Customers include leading AI-native product companies

Funding history (newest first)

Investors in our directory

Funds linked from Baseten's profile — open a fund page for stage focus and related deal articles.

Competitive landscape

Edge: Purpose-built inference stack (Baseten Inference Stack) with kernel/runtime optimizations and scale-to-zero economics — differentiated from raw GPU rental or single-vendor model APIs.

Inference platforms emerged as a distinct category in 2025–2026 as agent workloads exploded token demand. Baseten's $5B valuation signals investor belief the serving layer persists even as model providers improve — but hyperscaler bundling and open-source runtime commoditization are long-term pressures.

• Together AI direct

  Inference and fine-tuning platform for open models; similar developer-focused GTM.

• Fireworks AI direct

  Fast inference for open and custom models; competes on latency benchmarks.

• AWS Bedrock incumbent

  Hyperscaler managed model API; Baseten wins when teams need custom deployments and multi-model control.

Notable stories

• Baseten's founders built the company on the thesis that inference — not bigger context windows alone — would determine which AI products survive production traffic.
• The January 2026 Series E came just months after Series D — unusual velocity reflecting investor conviction that agent-era token volumes make inference infrastructure a durable category.

Industries

Market / IPO context

Related funding articles

Venture Capital Tracker pieces that cover Baseten's financing or category context.

• Baseten's $300M Round at $5B: Why AI Inference Platforms Are the New CDN