Research notes
Verified Sources Summary
Source 1: TechCrunch (2026-07-08)
- SambaNova raised $1B at $11B valuation in Series F first close, led by General Atlantic
- CEO Rodrigo Liang confirmed more investors expected to join in next few weeks for second close
- Raised $350M Series E in February 2026 alongside SN50 chip unveiling
- Intel had engaged in acquisition talks valued at ~$1.6B per Bloomberg December 2025 report
- JPMorganChase selected SambaNova as inference-infrastructure partner; SN40L and SN50 systems to power secure on-prem AI inference
- SN40L launched September 2023, available cloud and on-premises from November 2023
- SN50 unveiled February 2026, shipping to customers H2 2026 with SoftBank as first deployment partner
- Existing investors include Intel Capital; company and Intel deepened ties after Series C
- SambaNova is building heterogeneous infrastructure for enterprise inference; CEO expects trend to spread beyond banking
Source 2: General Atlantic Press Release (2026-07-08)
- Confirms $1B first close at $11B post-money valuation led by General Atlantic
- Other investors: Seligman Ventures, T. Rowe Price Associates, Capital Group, A&E Investment, Assam Ventures, Battery Ventures, BlackRock, Cambium Capital, Intel Capital, Kabila Capital, QFO Capital, Qatar Investment Authority (QIA), Vista Equity Partners, Volantis
- JPMorganChase deploying SN40 and SN50 systems for secure on-prem AI inference
- Quote from Martín Escobari (General Atlantic): "inference has become foundational to enterprise and industry transformation"
- Quote from Umesh Padval (Seligman Ventures): RDU architecture uniquely positioned to lower cost per token at scale
- SN50 unveiled earlier in 2026 with $350M+ raise and Intel collaboration
- Company will use proceeds to expand capacity, accelerate product innovation, scale deployments for enterprises, neo-clouds, sovereign AI customers
Source 3: TechStartups (2025-10-25)
- SambaNova explored sale in October 2025 after fundraising stalled; valuation dropped from $5B (2021) to ~$4B perceived
- Founded 2017 by Rodrigo Liang, Kunle Olukotun, Christopher Ré
- Reconfigurable Dataflow Unit (RDU) architecture with three-tier memory hierarchy; benchmarks showed 16 RDUs processing 198–255 tokens/sec on DeepSeek-R1 671B
- SN40L launched September 2023; supports models twice the size of advanced ChatGPT versions
- Claims sixfold faster training and 10x performance vs. Nvidia A100 on certain workloads
- Enterprise clients include U.S. DoE labs (Oak Ridge, Livermore), OTP Bank, Accenture, Aramco, SoftBank, NetApp, RIKEN
- Raised $2.17B across six rounds; investors include BlackRock, GIC, GV, Intel Capital, Temasek, Walden International, Samsung Catalyst Fund
- In March 2025 launched "deep research" agent framework with CrewAI Inc.
- Focus shifted toward inference optimization, multilingual modeling after GenAI surge
Planned Article Sections & Facts
Section 1: Funding Announcement & Valuation
- $1B Series F first close at $11B post-money valuation (TechCrunch, General Atlantic)
- Led by General Atlantic; Seligman Ventures, T. Rowe Price, Capital Group participating
- Second close expected in coming weeks; Rodrigo Liang confirms "a few more investors" joining
- Valuation spikes from $1.6B rumored acquisition price just months prior (TechCrunch, TechStartups)
Section 2: Why Valuation Skyrocketed — From Acquisition Talks to IPO Ambitions
- December 2025 rumor of Intel acquisition at ~$1.6B (TechCrunch)
- Founder noncommittal on exit path but notes doors open to acquisition; growth likely drives toward public listing
- Company fielding interest daily; CEO says momentum and growth drive strategy (TechCrunch)
Section 3: Hardware Roadmap — SN40L, SN50, and Premium Inference
- SN40L launched September 2023; cloud + on-prem availability November 2023 (TechCrunch)
- SN50 unveiled February 2026 with $350M Series E; shipping H2 2026, SoftBank first deployment (TechCrunch)
- RDU architecture handles multi-trillion-parameter models on single rack (TechCrunch, TechStartups)
- Benchmarks show 16 RDUs process 198–255 tokens/sec on DeepSeek-R1 671B, occasionally outperforming Nvidia (TechStartups)
- Company claims 6x faster training and 10x performance gains vs. A100 on specific workloads (TechStartups)
Section 4: Enterprise Wins — JPMorganChase and Sovereign Clouds
- JPMorganChase selects SambaNova as inference-infrastructure partner for secure on-prem inference (TechCrunch, General Atlantic)
- Quote from Darrin Alves, JPMorganChase CIO on performance, control, and reliability requirements (General Atlantic)
- Founder notes banks building private infrastructure for sensitive model inference; trend expected beyond finance (TechCrunch)
- Other customers: Saudi Aramco, Intel, Japanese firms (TechCrunch); DoE labs, OTP Bank, Accenture, RIKEN (TechStartups)
Section 5: Intel Partnership — From Rivals to Co-Development
- Intel backed SambaNova since Series C; co-develop products and bring to market together (TechCrunch)
- Multi-year partnership announced five months ago for AI inference on Intel Xeon platform (General Atlantic)
- CEO explains partnership gives leverage of Intel scale with SambaNova tech advantage (TechCrunch)
Section 6: Market Shift — From Training to Production Inference
- Company repositioning around premium inference, agentic AI, and enterprise deployment (General Atlantic)
- CEO says demand accelerating well ahead of supply; this round supports scaling capacity (TechCrunch, General Atlantic)
- Founder sees huge revenue still untapped beyond model makers and frontier labs (TechCrunch)
- Enterprise adoption expected to grow in banking, government, healthcare, energy sectors (TechCrunch, TechStartups)
Section 7: Funding Use — Supply Chain and Scale
- Proceeds to expand capacity, accelerate product innovation, scale deployments (General Atlantic)
- Critical use of capital: securing supply chain to fulfill orders and buy materials for next 12 months (TechCrunch)
- Will grow customer programs and invest across chips, systems, software, full-stack AI infrastructure (General Atlantic)