Baseten, a startup that serves open-weight AI models for enterprise customers, is closing a roughly $1.5 billion funding round at a valuation of up to $13 billion, according to The Wall Street Journal. The round is co-led by Altimeter Capital, Conviction, Spark Capital, Sands Capital, and Wellington Management, and is structured in two tiers priced at $11 billion and $13 billion.
The valuation chart is steep. Baseten was worth approximately $2.15 billion in September 2025, $5 billion in January 2026, and now sits at up to $13 billion, a six-fold increase in nine months. Revenue has tracked the same curve: annualized run-rate has grown from roughly $200 million to about $600 million over the same window. That is the kind of line that convinces late-stage funds to pile in at what would otherwise look like a stretched multiple.
The company’s product is inference infrastructure. Baseten provides software and multi-cloud compute capacity that helps companies optimize, customize, and serve lower-cost open-weight models. Inference is the computationally expensive step of getting a trained model to produce outputs for real users, distinct from the training process itself. It is the part of AI that touches production systems every second of every day, and it has historically been underbuilt relative to training infrastructure.
The central bet is that the training-compute arms race, which consumed most of the AI investment story for the past three years, is giving way to a different infrastructure war: who can serve open models at scale, at low cost, with the reliability enterprises actually need. Baseten’s thesis is that models from Meta, Mistral, and a growing tier of specialized open-weight labs have reached a quality threshold where many enterprise workloads no longer require a proprietary frontier API. A company that does not need OpenAI’s GPT-4o or Anthropic’s Claude for its internal summarization pipeline can cut its inference bill by switching to a served open model, provided someone handles the operational complexity.
That is the market Baseten is selling into. The company abstracts away multi-cloud routing, model optimization, and serving infrastructure, letting engineering teams focus on the application rather than the deployment layer. The pitch is cost savings without the DevOps burden of self-hosting.
The round is a meaningful signal in the ongoing debate over open versus closed AI models. OpenAI and Anthropic have both built business models on the assumption that frontier proprietary models maintain enough of a quality gap to justify their API pricing. A $13 billion valuation for a company whose entire revenue thesis depends on that gap narrowing is a direct market vote that the gap is already closing for a wide swath of enterprise use cases. The Journal’s reporting does not specify which verticals or workload types are driving Baseten’s revenue, and the company has not published independent benchmarks comparing its served models against OpenAI or Anthropic APIs on the specific tasks that converted customers are running.
The structure of the round, with two price tiers, suggests some investors accepted a lower entry valuation than the headline number. That is not unusual for rounds this size, but it means the effective blended valuation is somewhere between $11 billion and $13 billion, not definitively at the top of the range.
For teams currently evaluating their AI API spend, Baseten’s growth trajectory is a concrete data point: enterprises are moving workloads off proprietary APIs in volumes large enough to generate $600 million in annualized revenue for a single infrastructure provider. Running a cost comparison between your current API calls and an equivalent open-model deployment has become a legitimate quarterly exercise, not a speculative one.
Reported by The Wall Street Journal on June 19, 2026.
