SambaNova Raises $1 Billion at $11B Valuation as On-Premises AI Inference Demand Soars

Amid a broader cooling in AI chip valuations that has battered public equities, private infrastructure bets are accelerating. SambaNova Systems announced on July 8 that it has closed the first tranche of a $1 billion Series F funding round at an $11 billion valuation — the latest and largest raise in a company that has quietly built one of the more credible alternatives to Nvidia’s dominance in enterprise AI infrastructure.

The round was led by General Atlantic, with participation from Intel Capital, BlackRock, T. Rowe Price, Capital Group, Seligman Ventures, Qatar Investment Authority, and others. It follows a $350 million Series E closed just five months ago in February 2026, marking an extraordinary fundraising cadence that reflects both the intensity of demand for on-premises AI systems and investors’ confidence that SambaNova can capture a meaningful slice of it.

The On-Premises Bet

SambaNova’s core thesis is deceptively simple: the organizations that need AI most urgently are also the organizations least able to use public cloud infrastructure to run it. Banks, healthcare systems, defense contractors, and sovereign governments all face regulatory requirements, data sovereignty obligations, or security policies that make sending sensitive workloads to a hyperscaler’s inference endpoints legally or strategically untenable.

SambaNova builds the alternative. The company develops custom silicon, hardware systems, and the software stack to run them — delivered as a self-contained rack that enterprises deploy in their own data centers. The SN40L chip, released in 2023, anchored the company’s current product lineup. CEO Rodrigo Liang has described its defining capability in unusually concrete terms: the system can fit multi-trillion parameter models onto a single rack, enabling them to run with low latency and high throughput without disaggregated infrastructure spread across hundreds of nodes.

That claim matters more today than it did 18 months ago. As model sizes have continued to grow — GPT-5.6, Anthropic’s Claude Sonnet 5, and Meta’s Watermelon model all represent architectures measured in hundreds of billions to trillions of parameters — the cost and complexity of on-premises inference has increased in ways that commodity hardware cannot accommodate. SambaNova’s custom silicon is designed to collapse that complexity.

The SN50 Chip and What’s Next

The company’s next-generation chip, the SN50, was unveiled in February 2026 and is scheduled to begin shipping in the second half of this year, with SoftBank as the first confirmed deployment partner. SambaNova has not disclosed detailed specifications for the SN50, but the company’s engineering roadmap suggests it targets the same performance-per-watt and model-density metrics that made the SN40L attractive to customers running the largest available models in controlled environments.

The SN50 launch will be the most critical product test in SambaNova’s history. It arrives as the competitive landscape has shifted considerably — Etched, Groq, Cerebras, and various hyperscaler custom silicon programs are all competing for the same on-premises and high-density inference market. Intel’s own Gaudi 3 accelerators, backed by Intel Capital’s participation in this very round, represent both a competitive and a complementary bet by SambaNova’s chip-sector backer.

A Customer Base That Signals Serious Intent

Perhaps the most telling indicator of SambaNova’s momentum is who is buying its systems. JPMorgan Chase publicly announced it would deploy SambaNova’s hardware for “on-prem inference in our demanding enterprise AI workloads” — a statement that carries unusual specificity for a financial institution famously cautious about AI vendor endorsements. Saudi Aramco, the world’s largest oil company and a participant in an earlier SambaNova round through Aramco Ventures, is also a deployment partner. SoftBank’s selection as the initial SN50 launch partner adds a third major institutional validator.

These are not pilot programs or proof-of-concept deployments. Financial institutions and major energy companies procure infrastructure at scale and over multi-year cycles. A JPMorgan commitment signals that SambaNova’s systems have cleared internal security review, performance benchmarking, and the kind of sustained-load testing that enterprise buyers impose before signing infrastructure contracts.

The Acquisition That Didn’t Happen

One revealing subplot in SambaNova’s recent history: Intel explored acquiring the company for approximately $1.6 billion, according to sources familiar with the discussions. Those talks stalled, and SambaNova instead raised independently at a valuation now seven times Intel’s reported offer price. The outcome speaks to two things simultaneously — the trajectory of SambaNova’s perceived value since those discussions, and Intel’s continued difficulty executing on its own AI infrastructure ambitions even as it participates in SambaNova’s funding round.

Intel’s presence in this Series F is worth reading carefully. By investing in SambaNova rather than acquiring it, Intel is betting on a partnership model that leverages its manufacturing capabilities and customer relationships without the integration risk of a full acquisition. For SambaNova, Intel Capital’s participation provides a politically important signal at a moment when U.S. AI infrastructure supply chains are under intense geopolitical scrutiny.

The IPO Question

SambaNova’s CEO has indicated that an IPO is possible in the company’s future, though no timeline has been communicated. The aggressive fundraising pace and swelling valuation suggest the company is not in a rush to access public markets — at $11 billion, SambaNova has room to grow substantially before a public offering would meaningfully change its capital access or public profile. But the path is clearly being laid. The Series F’s institutional investor roster — BlackRock, T. Rowe Price, Capital Group — reads more like an IPO roadshow than a startup funding round. These are firms that manage public equities and are positioning for liquidity.

Why It Matters Now

SambaNova’s raise lands at a peculiar moment in the AI infrastructure market. Public chip stocks have experienced significant corrections in 2026, with investors growing skeptical of the timeline for AI spending to translate into durable revenue for hyperscalers and their suppliers. Yet private infrastructure bets like SambaNova’s continue to attract capital at aggressive valuations, reflecting a divergence between Wall Street’s near-term anxiety about AI monetization and institutional investors’ long-term conviction that the demand for secure, high-performance AI compute is structurally growing.

The inference market — running models rather than training them — is where most commercial AI value creation will ultimately accrue. SambaNova’s bet is that the most valuable inference workloads will never leave enterprise data centers, and that no one yet owns that market. With $1 billion in fresh capital, the SN50 shipping imminently, and a customer list that includes JPMorgan and Saudi Aramco, it is in as strong a position as any company to test that hypothesis.