SambaNova Scores $1B, Hits $11B Valuation in AI Chip War

The Price of Rebellion Just Went Up

Want to build an alternative to Nvidia’s AI chips? You'll need deep pockets. A billionaire's pockets. Palo Alto's SambaNova Systems just proved it, raising a jaw-dropping $1 billion in a Series F funding round that explodes its valuation to $11 billion. This isn't just another funding announcement; it's a dramatic escalation in the war for enterprise AI. The round was led by growth equity giant General Atlantic, bringing along a heavy-hitting roster of new and existing investors like BlackRock, Intel Capital, T. Rowe Price, and the Qatar Investment Authority.

This isn't just a vote of confidence. It's a war chest. Co-founder and CEO Rodrigo Liang says the money is critical for one reason: meeting an "incredible wave of demand" by expanding capacity and shoring up the company's supply chain. And they need it now. Announced Wednesday, this is just the *first close* of the Series F round. Liang told TechCrunch more investors are lining up. The urgency is plain to see, coming just five months after SambaNova banked a $350 million Series E in February.

A Bet on Inference and On-Premise Power

Nvidia’s GPUs own the market for training giant AI models. That's a given. But SambaNova is making a calculated bet on the other side of the equation. Inference. That's the real work—running trained models to generate answers, drive apps, and make actual decisions. The computational cost is shifting hard in this direction as companies move from AI experiments to full-scale deployment. It’s a massive market, too, pegged to hit over $253 billion by 2030, according to Grand View Research. SambaNova's whole architecture, a full-stack system of custom chips and software, is built for exactly this moment.

Their secret sauce? The Reconfigurable Dataflow Unit, or RDU. Now in its fifth generation with the SN50 chip, it's nothing like a general-purpose GPU. The RDU is a specialist. It maps the data flow of a neural network straight onto the silicon, effectively creating a computational assembly line. The whole point is to cut down on data movement, which is the biggest time-and-energy suck in AI processing, and deliver pure speed and efficiency. And it's working. The approach just landed a huge customer: JPMorgan Chase. The financial titan will deploy SambaNova's SN40 and SN50 systems for its secure, on-premise AI inference. For banks, keeping sensitive data inside their own walls isn't a choice; it's a mandate. "Having JPMorgan Chase decide they're going to use SambaNova for their inference solution is a big deal," Liang said.

Navigating a Crowded and Cutthroat Field

This $11 billion valuation marks a stunning turnaround. Just seven months ago, back in December 2025, Intel was reportedly in advanced talks to acquire the company for just $1.6 billion. A bargain. But those talks stalled. Instead, Intel Capital became a strategic partner and investor—a pivot that has turned out to be incredibly lucrative for SambaNova's early backers in the middle of this blistering AI gold rush.

The field is absolutely brutal. SambaNova isn't just fighting Nvidia's market dominance and its locked-in CUDA software ecosystem. A whole cohort of well-funded hardware startups are in the ring. There's Groq, obsessed with raw speed. Cerebras, with its wild, wafer-scale chips. And the recent stock collapse of Cerebras after its IPO is a chilling reminder of the pressure to justify these valuations. [cite: /article/cerebras-stock-plummets-investor-probes-post-ipo] Let's not forget the cloud giants. Google, Amazon, and Microsoft are all building their own custom silicon, making the market even more complex.

Still, investors are betting there's room for more than one winner. A lot of room. Martín Escobari, Co-President at lead investor General Atlantic, thinks so. "SambaNova's platform is differentiated, built for a market where inference has become foundational to enterprise and industry transformation," he said. Then there's the Intel connection, which adds another layer of intrigue. Intel is trying to claw back data center dominance with its own chips, like the upcoming Nova Lake CPUs, yet it's also backing a potential rival. Interestingly, Intel CEO Lip-Bu Tan has chaired SambaNova's board since 2017, though he did step away from the latest collaboration talks.

SambaNova now has over $2 billion in total funding. That's enough cash to put up a real fight. But money isn't everything. The challenge now is pure execution. They have to scale production, grow their global customer base, and prove their specialized dataflow architecture delivers a TCO low enough to lure enterprises away from the Nvidia default. The AI chip war is far from over. It just got a whole lot more expensive.