The last time SambaNova’s name was in the papers, it was the subject of an acquisition rumor with a price tag of roughly $1.6 billion. On Wednesday, the AI chip startup announced a round nearly ten times that size as a valuation: $1 billion in new funding at an $11 billion post-money valuation, led by General Atlantic. The company is staying independent, and it is paying to prove the point.
The round, a first close of a Series F, comes roughly five months after SambaNova raised more than $350 million and unveiled its next-generation SN50 chip. Intel, which had been in acquisition talks with the company late last year, participated through its venture arm, an arrangement that turned a potential takeover into a strategic collaboration.
The investor list reads like a who’s who of growth capital: Seligman Ventures, T. Rowe Price, and Capital Group joined General Atlantic, with BlackRock, the Qatar Investment Authority, Battery Ventures, Vista Equity Partners, and Volantis also participating, according to the company. The breadth of the group signals that institutional money sees AI inference as a durable market rather than a passing fashion.
SambaNova’s pitch is different from Nvidia’s. Where Nvidia sells general-purpose GPUs, SambaNova builds specialized chips and full-stack systems tuned for inference, the process of running trained models. The company argues that as AI shifts from training to deployment, and as agents and chatbots multiply, the economics of running models will matter more than the economics of building them.
The customer list supports the story. JPMorganChase has selected SambaNova as an inference infrastructure partner, deploying its SN40 and SN50 systems for on-premises AI workloads, and SoftBank is the first deployment partner for the SN50 chip when it begins shipping in the second half of the year. Banks and telecom operators, neither known for early adoption, are the kind of validation startups need.
Chief executive Rodrigo Liang said the capital is going into the ground. “We’re using that capital to secure the supply chain,” he said, describing it as essential to fulfilling orders and buying the materials the company needs over the next 12 months. The company plans to expand capacity, scale deployments worldwide, and keep investing across chips, systems, and software.
The timing reflects the state of the AI chip market. Demand for inference hardware is exploding as agentic AI moves into production, and every hyperscaler and enterprise is shopping for alternatives to Nvidia’s dominance. Venture money has flooded into challengers including Groq, Cerebras, and Tenstorrent, each betting that a specialized architecture can win a share of the fastest-growing hardware market in the industry’s history.
SambaNova’s journey has not been smooth. The company was once valued far higher in private markets, and the Intel acquisition talks at $1.6 billion, reported in December, suggested a humbler future. Wednesday’s round resets the narrative: a $1 billion raise at an $11 billion valuation is a statement of independence backed by real orders.
The competitive question is whether specialized inference chips can dent Nvidia’s position. Nvidia’s accelerators remain the default choice, and its software ecosystem gives it a grip that challengers have struggled to loosen. But inference workloads have different characteristics than training, and customers are showing willingness to test alternatives where cost per query matters.
SambaNova’s technology story goes back further than its recent headlines. The company shipped its SN40L chip in late 2023, built around a reconfigurable dataflow architecture that it says handles AI workloads more efficiently than general-purpose designs. The architecture is the work of a founding team with deep academic roots: Rodrigo Liang, the chief executive, co-founded the company with Stanford professors Kunle Olukotun and Chris Re, and the design reflects years of research into how specialized hardware can accelerate machine learning.
The market SambaNova is chasing is enormous and growing. Analysts project that AI inference spending will expand by a triple-digit percentage over the next few years as trained models move from data centers into production applications across finance, healthcare, retail, and government. Every one of those deployments needs hardware to run on, and the scale of the opportunity is why investors are funding challengers despite Nvidia’s dominance.
The round’s structure is worth noting. A first close means SambaNova can bank the $1 billion now and add more investors later, a common pattern when demand exceeds initial allocations. The company said additional investors are expected to join, and the final size of the round could exceed the announced figure.
The risks are equally clear. Nvidia’s software ecosystem remains the industry standard, and customers that standardized on its programming framework are expensive to move. SambaNova’s answer is total cost of ownership, arguing that its systems deliver more performance per watt and per dollar for inference workloads, a claim that customers like JPMorganChase are now testing in production.
For now, SambaNova has what it needs: capital, customers, and a product cycle that aligns with the market’s growth. The next test is execution, shipping the SN50 at scale and proving that the $11 billion valuation was not just a number on a term sheet.
