On July 8, 2026, TechCrunch reported that AI chip maker SambaNova Systems closed the first $1 billion of its Series F round at an $11 billion valuation, led by General Atlantic with existing backers including Intel participating. Only five months have passed since its previous $350 million Series E.
A Second Mega Round in Five Months
This is a first close: CEO Rodrigo Liang says a second close with additional investors should finish within weeks. Beyond lead investor General Atlantic, participants include Intel (a backer since Series C), T. Rowe Price, BlackRock, Capital Group, Battery Ventures, Vista Equity Partners, and the Qatar Investment Authority, among others. The use of proceeds is explicit: scale the business and secure the supply chain over the next 12 months.
The pace itself is the signal. In February 2026 the company closed a $350 million Series E led by Vista and unveiled its next-generation SN50 chip at the same time. Five months later it has pulled in $1 billion at an $11 billion valuation, which says the capital race in inference compute is still accelerating — and that investors will pay a premium for any credible alternative to the GPU status quo.
SN50: A Fifth-Generation RDU Built for Agentic Inference
SambaNova’s core product is the RDU (Reconfigurable Dataflow Unit), a dataflow architecture that takes a different path from GPU clusters. The fifth-generation SN50, introduced on February 24, 2026, packs 16 chips into each SambaRack system. The company’s figures claim 5X maximum speed and 3X-plus throughput versus NVIDIA’s Blackwell B200 on agentic inference benchmarks using Llama 3.3 70B, and when scaled to 256 accelerators it can run models of up to 10 trillion parameters with context lengths up to 10 million tokens.
The SN50’s design philosophy treats inference as a data-movement problem: with a tiered memory hierarchy spanning SRAM, HBM, and large-capacity memory, models hot-swap in milliseconds, while input-token caching cuts Time to First Token. Average rack power runs about 20 kW and deploys with air cooling, easing retrofit costs in existing data centers. The SN50 ships to customers in the second half of 2026, with SoftBank as the first deployment partner.
The Customer Map: From SoftBank to JPMorgan
Alongside the round, JPMorganChase announced it has selected SambaNova as an inference-infrastructure partner and will run security-sensitive inference workloads on SN40L and SN50 systems inside its own facilities. Liang frames the deal as a signal to the banking industry: heavy AI users do not have to depend on the cloud alone.
Other disclosed customers include Saudi Aramco and several Japanese firms. The company’s target buyers are sovereign clouds, neoclouds, and large enterprises — precisely the customers with the strongest demands for data sovereignty and on-premises deployment. The earlier SN40L is already available both in the cloud and on premises, while the SN50 leads with a single-rack, frontier-model pitch.
The Complicated Intel Relationship
Intel’s role is the most delicate part of the story. It has invested since Series C and joined this round too. In December 2025, Bloomberg reported that Intel had neared a roughly $1.6 billion acquisition of SambaNova before talks stalled. In February 2026 the two sides settled on a multi-year collaboration instead, co-developing and jointly selling AI inference solutions built on Intel’s Xeon platform.
Asked whether the company will stay independent, Liang stays noncommittal, saying momentum points toward an eventual IPO. At an $11 billion valuation, any acquirer’s bar now sits far higher than it did five months ago — the acquisition window is closing fast.
The Next Phase of the Inference Market
SambaNova positions itself in premium inference: running multi-trillion-parameter frontier models at single-rack scale, in direct competition with NVIDIA’s GPU ecosystem. Cerebras and Groq race on the same track, all betting on the inference demand explosion that agentic AI is expected to bring.
For buyers, two practical implications stand out. First, the menu of inference hardware keeps growing — GPUs are no longer the only path, and the negotiating room on cost and latency widens with every credible entrant that ships. Second, the SN50’s real-world numbers will face scrutiny once shipments begin in the second half of 2026; whether the claimed multiples hold up will decide whether this $1 billion was a bubble or a ticket. Either way, the fact that a rack-scale inference specialist can now raise at an $11 billion valuation shows where the market believes the next phase of AI infrastructure spending will land.
Sources
- AI chip maker SambaNova raises $1B at $11B valuation — TechCrunch
- Introducing the SN50 RDU — SambaNova
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.