SambaNova Funding: AI Chip Startup Raises $1 Billion at an $11 Billion Valuation

by Zaira Varghese 09. July 2026 AI 0

The latest round of SambaNova funding lands at $1 billion on an $11 billion valuation, and it arrives just five months after the company’s last mega-raise. Announced July 8, 2026, the first close of the Series F was led by General Atlantic, with T. Rowe Price, Capital Group, BlackRock funds, Intel Capital, Battery Ventures, Vista Equity Partners and the Qatar Investment Authority all writing checks. For a company that spent years being described as the other AI chip startup, the number is a statement.

Why the SambaNova Funding Round Matters Now

The AI accelerator market has been a one-horse race for the better part of three years. Nvidia’s margins have been extraordinary precisely because there has been nowhere else to go for large-scale training and inference at speed. That is changing — not because anyone has out-engineered Nvidia on raw training throughput, but because inference is now the bigger, uglier, more expensive problem.

Inference is where the money leaks. Every chatbot reply, every agentic tool call, every RAG lookup is a forward pass someone has to pay for. Investors backing challengers like SambaNova, Groq and Cerebras are betting that the economics of serving models will diverge sharply from the economics of building them, and that the winner of one is not automatically the winner of the other. TechCrunch reported the first close of the round on the same day General Atlantic confirmed it.

The JPMorgan Chase Deal Is the Real Signal

Alongside the raise, SambaNova disclosed that JPMorgan Chase has selected it as an inference infrastructure partner, deploying SN40 and SN50 systems to run secure, on-premises AI inference inside the bank.

That single sentence carries more weight than the valuation. Regulated financial institutions do not send customer data to a public inference API. They want the model in their own rack, behind their own firewall, on hardware they control. It is a market segment Nvidia serves through partners rather than directly, and it is one where SambaNova’s full-stack pitch — chips, systems, software, deployed as a unit — is easier to sell than a pile of GPUs and a services contract.

Round size: $1 billion, first close of a Series F
Valuation: $11 billion, post-money
Lead investor: General Atlantic
Anchor customer: JPMorgan Chase, on-prem inference
Stated plan: IPO under consideration for 2027, likely in the US

Sovereign AI Is Quietly Funding the Challengers

The investor list is worth reading twice. The Qatar Investment Authority’s presence is not incidental. Sovereign buyers — national governments building domestic AI capacity — are a category that barely existed in 2023 and now represents a meaningful slice of the accelerator market. They want supply that is not contingent on a single American vendor’s allocation queue, and they are willing to fund the alternative into existence.

SambaNova says proceeds will expand capacity, accelerate product work, and scale deployments across enterprises, neoclouds, sovereign customers and service providers. Read plainly: the company has more demand than it can currently manufacture against, which is the same constraint that has defined the entire sector since 2023.

What Could Go Wrong

An $11 billion valuation on a company that is not yet a public-market comparable prices in a lot of future. Three things could puncture it. First, Nvidia has never been slow to cut prices when a challenger gets traction, and it has the gross margin to do so painfully. Second, the custom-silicon programs at OpenAI, Google and Amazon remove the largest inference workloads from the merchant market entirely. Third, the software moat is not the chip — it is CUDA, and it remains the hardest thing in this industry to dislodge.

Against that, SambaNova has something Groq and Cerebras are still assembling: a named, regulated, brand-name enterprise customer running production inference on its metal.

The Takeaway

SambaNova funding at this scale does not mean Nvidia is in trouble. It means the market has finally decided that inference is a distinct business worth capitalizing separately, and that a full-stack, on-premises alternative is worth $11 billion of optionality. The CEO is openly weighing a 2027 US listing. If the JPMorgan deployment holds up, that listing will not be a hard sell.