AI chip company SambaNova Systems has closed a $1 billion Series F round at an $11 billion post-money valuation, one of the largest AI infrastructure bets of the year. General Atlantic led the round, with participation from Seligman Ventures, T. Rowe Price Associates, Capital Group, BlackRock, Intel Capital, the Qatar Investment Authority, Battery Ventures, and Vista Equity Partners, among others.

The round arrives five months after SambaNova’s previous mega-round and underscores just how competitive the race for AI inference infrastructure has become. As businesses move from AI pilots to production deployments of AI agents, the question of where inference runs and at what cost has become central to every enterprise AI strategy.

What SambaNova Actually Builds

SambaNova is one of a small group of companies attempting to dent Nvidia’s dominance in AI accelerator chips. Its SN50 chip, unveiled in February 2026, is scheduled to begin shipping in the second half of this year. The company claims the SN50 delivers more than three times the throughput of Nvidia’s B200 GPU and reaches top speeds five times faster.

These are significant claims, and enterprises are taking them seriously. SoftBank signed on as the first SN50 deployment partner. JPMorgan Chase selected SambaNova as its inference infrastructure provider for on-premises AI deployments, specifically citing the need for secure, controlled inference that doesn’t route sensitive data through third-party cloud services.

Intel has also entered a multi-year partnership with SambaNova to co-develop and jointly market AI inference products built on Intel’s Xeon architecture.

The new capital will go primarily toward scaling the supply chain to meet what the company describes as surging order demand over the next 12 months.

Why Inference Infrastructure Matters Now

For most businesses deploying AI today, the model is the visible part of the stack. The infrastructure it runs on is invisible but crucial.

Inference is the process of running a trained AI model to generate outputs. Every time your AI agent answers a question, routes a call, processes a document, or executes a workflow step, inference happens. At scale, the cost and latency of inference shapes whether AI is economically viable or not.

This is why the race to build cheaper, faster inference chips is heating up. The hyperscalers (AWS, Azure, Google Cloud) each have their own silicon programs, but enterprises looking for on-premises deployments or alternatives to the dominant cloud providers are actively evaluating options like SambaNova.

The JPMorgan partnership is a telling signal. Banks and financial institutions handling sensitive data are among the most motivated buyers of on-premises AI infrastructure. If SambaNova can win that class of customer, it has proven that its hardware meets enterprise-grade requirements for security, reliability, and performance.

The Nvidia Challenger Landscape

SambaNova is not operating alone in this space. Groq, Cerebras, and Etched are each pursuing different approaches to making inference faster and cheaper. But SambaNova’s approach, a full-stack architecture that spans custom silicon, systems, software, and deployment, has attracted a notably institutional investor base.

BlackRock and T. Rowe Price are not typical deep-tech venture bettors. Their participation in this round suggests that AI infrastructure has crossed from speculative technology into something institutional capital is comfortable holding.

The $11 billion valuation, up substantially from previous rounds, also tells a story about how investors are pricing the competition for Nvidia’s market position. Nvidia reported AI data center revenues that dwarfed the entire AI software market in 2025. Even a small share of that market represents enormous value.

What This Means for Business

If you are deploying AI agents or evaluating AI infrastructure, the SambaNova round matters for a few reasons.

Competition drives costs down. Every well-funded challenger to Nvidia pushes the entire industry toward better price-performance ratios. Over the next 12 to 24 months, businesses deploying AI at scale will have more options and more negotiating leverage than they did in 2024 or 2025.

On-premises AI is viable. The JPMorgan partnership signals that high-performance on-premises inference is now within reach for enterprises that need it. This is particularly relevant for businesses in regulated industries such as healthcare, financial services, and legal, where sending data to third-party cloud inference endpoints creates compliance complexity.

Infrastructure is becoming a strategic decision. For companies building AI-native workflows, the choice of inference infrastructure increasingly shapes what is possible. Latency, throughput, and cost-per-token directly affect how many agent interactions your business can run and at what quality level.

At Enterprise DNA, we work with businesses at exactly this inflection point. Most organizations are not yet thinking about inference infrastructure as a strategic choice. They are still focused on which AI model to use. But as AI agents move from experiment to core operations, the infrastructure layer becomes critical.

The question is not just “which AI?” but “at what speed, at what cost, on what hardware, and with what data controls?”

SambaNova’s $1 billion round is a signal that the answers to those questions are becoming commercially important and that serious capital is flowing to the companies building them.

---

Enterprise DNA helps business leaders understand and deploy AI strategically. If you are evaluating AI infrastructure for your organization or want to understand how AI agents fit into your operations, book a discovery call with our team.