SambaNova Just Raised $1 Billion at an $11 Billion Valuation — and JPMorgan Is the Customer

AI chip startup SambaNova Systems has raised $1 billion at an $11 billion valuation in the first close of its Series F round, led by General Atlantic. The round comes five months after the company’s $350 million Series E — and roughly six months after Intel was reportedly trying to buy it for $1.6 billion. Now SambaNova is worth nearly seven times that abandoned acquisition price.

🔍 THE BOTTOM LINE

The valuation jump from $1.6B (Intel’s rumored acquisition offer) to $11B in eight months tells you everything about what changed: AI inference infrastructure became a category, and JPMorgan Chase just became the proof point that banks will build their own on-premises AI rather than depend on cloud providers. SambaNova is selling the picks — and the gold rush is enterprise inference.

The Deal

SambaNova raised $1 billion in a Series F first close led by General Atlantic, with TechCrunch reporting that more investors are expected to join. CEO Rodrigo Liang told TechCrunch: “In the next few weeks, a few more investors will be coming in, and the second close is likely to finish up.”

The investor list reads like a who’s-who of growth capital: T. Rowe Price Associates, Capital Group, BlackRock, Vista Equity Partners, Qatar Investment Authority, Battery Ventures, and others. That’s not just venture money — it’s sovereign wealth and institutional capital betting that inference infrastructure is a durable market, not a flash.

Bloomberg and The Information both confirmed the valuation, which represents a 7x jump from the $1.6 billion Intel was reportedly willing to pay in December.

The Customer That Matters: JPMorgan Chase

Alongside the funding, SambaNova announced that JPMorganChase has selected it as an “inference-infrastructure partner,” with SN40L and SN50 systems set to power secure, on-premises AI inference at the bank.

Liang’s framing: “Having JPMorgan Chase decide they’re going to use SambaNova for their inference solution is a big deal. It sends a message to the banking industry that it’s time to not completely depend on cloud services.”

This is the signal. Banks — and enterprises more broadly — are pulling sensitive AI workloads in-house. When the largest bank in the United States by assets decides to build its own inference infrastructure rather than route through AWS, Azure, or Google Cloud, the cloud providers’ AI revenue moat has a crack in it.

The Intel Relationship

SambaNova’s ties to Intel have deepened rather than dissolved. Intel, a backer since Series C, participated in this latest round. Five months ago the two announced a multi-year partnership to co-develop AI inference products based on Intel’s Xeon chip. They now co-develop and go to market together.

Liang told TechCrunch the relationship gives SambaNova “the scale of Intel with the technology we have” — which is a way of saying SambaNova’s chips are the product, and Intel’s sales channel and Xeon integration are the distribution. The December acquisition talks that valued SambaNova at $1.6B look, in retrospect, like Intel trying to buy cheap what it’s now partnering with at $11B.

What SambaNova Actually Makes

The company launched its SN40L chip in September 2023. Its next-generation SN50, unveiled in February 2026 alongside the Series E, is due to begin shipping in the second half of 2026. SoftBank is the first deployment partner for SN50.

SambaNova’s pitch is “premium inference” — running the largest frontier models fast. The company says it can fit multi-trillion-parameter models onto a single rack, which reduces the latency and cost of inference for the biggest models. That matters because frontier models keep getting bigger, and the inference cost problem scales with parameter count.

The Competitive Landscape

SambaNova isn’t the only AI chip startup chasing the inference market. Nvidia is partnering with competitors like d-Matrix to hedge its position. Anthropic is developing its own server chips with Samsung. OpenAI has its Jalapeño chip with Broadcom. The entire frontier-lab ecosystem is vertically integrating into silicon.

What makes SambaNova different is that it’s not a captive chip shop for one model maker — it’s selling inference infrastructure to third parties. That’s a fundamentally different business: it’s competing with Nvidia’s data-center revenue, not just making one company’s models run faster.

NZ Angle

New Zealand’s sovereign AI ambitions — to the extent they exist — face the same inference-infrastructure question SambaNova is answering for JPMorgan. Local banks and government agencies running sensitive models will eventually need on-premises inference, not cloud API calls. Whether that hardware comes from SambaNova, Nvidia, or a sovereign-cloud partner is the procurement decision that will define who controls New Zealand’s AI infrastructure.

❓ FAQ

How much did SambaNova raise and at what valuation? $1 billion in a Series F first close at an $11 billion valuation, led by General Atlantic. More investors are expected in a second close.

Why is JPMorgan Chase significant? JPMorgan is the first major bank to publicly commit to on-premises AI inference with a non-cloud-vendor partner. It signals that financial institutions want control over their AI infrastructure, not just API access.

What happened with the Intel acquisition? Intel was reportedly in talks to acquire SambaNova for roughly $1.6 billion in December. Those talks stalled. SambaNova is now worth $11B — nearly 7x the abandoned offer — and Intel is both an investor and a go-to-market partner.

What chips does SambaNova make? The SN40L (launched 2023) and the SN50 (unveiled February 2026, shipping H2 2026). The company says its systems can fit multi-trillion-parameter models on a single rack for fast inference.

How does this compare to Nvidia? SambaNova is a pure-play inference company selling to third parties. Nvidia dominates the training market and is now partnering with competitors like d-Matrix to cover the inference side. SambaNova’s pitch is that it was purpose-built for inference, while Nvidia’s architecture was designed for training and adapted.

🔍 THE BOTTOM LINE

An $11 billion valuation for an inference-chip startup eight months after it was worth $1.6 billion is not a normal valuation curve. It’s a market repricing — the same kind of repricing that happened to GPU demand when ChatGPT launched. The difference is that this time the bet isn’t on training the next model. It’s on running the ones we already have. JPMorgan’s commitment is the first proof point that banks will pay for that capability. The question is whether SambaNova’s silicon can deliver at the scale its valuation now demands.