SambaNova Raises $1B Series F at $11B Valuation
SambaNova Systems has closed a $1B Series F at an $11B valuation, just five months after its previous mega-round — a dramatic reversal from earlier rumors that Intel was eyeing an acquisition at roughly $1.6B.
Original sourceSambaNova Systems has raised $1 billion in a Series F first close at an $11 billion valuation, according to TechCrunch. The round comes roughly five months after the company's last large fundraise, signaling sustained investor appetite for purpose-built AI inference infrastructure at a time when generic GPU capacity is increasingly commoditized.
The timing is notable in part because of what didn't happen: Intel was reportedly in discussions to acquire SambaNova for approximately $1.6 billion earlier this year. That deal never materialized, and SambaNova has now raised at nearly seven times that rumored acquisition price — a testament either to dramatically improved business fundamentals or to the speculative heat still present in AI infrastructure markets, depending on your priors.
SambaNova builds custom AI chips and systems optimized for large-scale inference workloads, competing in a market that includes established players like Nvidia and Cerebras as well as a wave of well-funded startups like Groq and Etched. The company has positioned itself around full-stack delivery — chips, systems, and software — rather than selling silicon alone, which gives it a distinct go-to-market compared to pure-play chip vendors.
With two major rounds in under six months, SambaNova is clearly in capital-intensive execution mode. Whether the $11B valuation reflects genuine enterprise traction at scale or is being carried by the broader infrastructure funding wave will likely become clear as the company is pushed toward more public disclosure ahead of what many expect will be an IPO process.
Panel Takes
The Founder
Business & Market
“The Intel acquisition rumor at $1.6B followed by an $11B valuation raise in the same year is a fascinating data point — either the business fundamentally re-rated or someone blinked at the wrong time. The full-stack positioning (chips plus software plus systems) is the right moat-building move; pure silicon vendors are one Nvidia pricing shift away from margin collapse. What I'd want to see before calling this a real business: enterprise revenue cohorts, not just logos, and whether inference workloads are sticky enough to create genuine switching costs once the software layer is embedded.”
The Skeptic
Reality Check
“Two mega-rounds in five months isn't momentum — it's a company burning fast enough to need continuous capital injections, and no one is publishing the burn rate. The jump from a rumored $1.6B acquisition price to an $11B self-reported valuation in one year is the kind of delta that makes sense in exactly two scenarios: extraordinary undisclosed revenue growth, or late-stage AI infrastructure bubble dynamics where the last check sets the number. My prediction for what kills this in 18 months: Nvidia ships a software stack that makes SambaNova's full-stack pitch redundant for the top 20 enterprise accounts, and the valuation resets to something that matches actual contracted ARR.”
The Futurist
Big Picture
“The thesis SambaNova is betting on is falsifiable and specific: inference at scale will require heterogeneous silicon, not just more H100s, and enterprises will pay a premium for full-stack delivery over assembling their own inference stack from commodity parts. That bet only pays off if inference workloads continue to diversify in architecture requirements — which is plausible given the proliferation of mixture-of-experts models and long-context workloads that penalize general-purpose GPUs. The second-order effect worth watching: if SambaNova wins, it accelerates the fragmentation of the AI hardware market and erodes Nvidia's ability to use CUDA lock-in as a moat, which reshuffles power across the entire inference supply chain.”
The Builder
Developer Perspective
“The part nobody talks about with full-stack AI chip companies is the developer surface: if I can't get a clean API to SambaNova's inference layer without a 6-month enterprise procurement cycle and a custom integration contract, the hardware specs are irrelevant to 99% of builders. Cerebras at least has a public API endpoint you can hit with a curl command — SambaNova's developer story is still almost entirely sales-led, which means the ecosystem flywheel that makes Nvidia sticky simply doesn't start spinning. Until there's a documented, publicly accessible inference API with real pricing and a free tier, this is infrastructure that exists for Fortune 500 procurement teams, not the people who actually build with it.”