Market thesis / Jul 8, 2026 / 4 min
$11B for Chips Intel Walked Away From
On July 8, SambaNova closed the first $1 billion tranche of a Series F at an $11 billion valuation and named JPMorgan Chase as an on-premises inference partner — five months after a $350 million Series E and seven months after Intel's reported $1.6 billion acquisition talks collapsed — while Micron, Samsung, and SK Hynix shares kept falling through the week's memory selloff.
SambaNova raised $1 billion at an $11 billion valuation on July 8 and landed JPMorgan Chase as an inference partner — the same week Seoul hit its sixth circuit breaker, memory stocks cratered, and Wall Street closed SK Hynix's ~$29 billion Nasdaq book anyway.
The raise:
- General Atlantic led the Series F first close at $11 billion post-money, with Seligman Ventures, T. Rowe Price, Capital Group, BlackRock, Intel Capital, Qatar Investment Authority, and Vista Equity Partners among participants (BusinessWire, TechCrunch, July 8).
- CEO Rodrigo Liang told TechCrunch a second close is coming in "the next few weeks" with more investors joining.
- Proceeds will "secure the supply chain" against "an incredible wave of demand" over the next 12 months, Liang said (TechCrunch).
The customer:
- JPMorgan selected SambaNova SN40L and SN50 systems for secure, on-premises inference — not a third-party cloud lease (BusinessWire, Quartz).
- Darrin Alves, CIO of infrastructure platforms: "At JPMorganChase, AI infrastructure has to meet a very high bar for performance, control and reliability" (BusinessWire).
- Liang: "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 not to completely depend on cloud services" (TechCrunch).
The arc:
- Bloomberg reported in December 2025 that Intel was in advanced talks to acquire SambaNova for about $1.6 billion including debt; the deal never closed (Bloomberg, TechCrunch).
- In February 2026, SambaNova raised $350 million in Series E alongside a multi-year Intel partnership on heterogeneous inference — GPUs for prefill, SambaNova RDUs for decode (TechCrunch, EE Times).
- Reuters reported in April that Intel planned another $15 million investment that would lift its SambaNova stake toward 9% (Reuters).
What's shipping:
- SN50 chips unveiled in February begin customer shipments in the second half of 2026; SoftBank is the first deployment partner (TechCrunch).
- Liang told Bloomberg that SN40 and SN50 can run the decode portion of inference five to ten times faster than competing approaches, per Quartz's reporting — freeing GPU capacity for other tasks.
- Liang told CNBC the company is eyeing a U.S. IPO in 2027, according to Quartz.
Why now:
- The Bank of England modeled a 2.2% UK GDP hit from an AI equity correction on July 7 — the same day Seoul's circuit breaker fired on record Samsung earnings (Convina, July 8).
- SK Hynix closed an oversubscribed ~$29 billion Nasdaq book on July 8 while Micron, Sandisk, and memory peers kept sliding (Convina, July 8).
- SambaNova's raise lands in the opposite corner: inference hardware regulated banks want inside the firewall, not another hyperscaler token bill.
Convina's view: The chip trade this week told you Wall Street is selling memory perfection and buying whatever fits behind a compliance officer's desk. JPMorgan didn't endorse Nvidia's next training run — it signed for racks that keep fraud models and credit-risk inference inside its own walls. That is the real rotation: from scarcer HBM wafers to sovereign inference stacks General Atlantic will underwrite at eleven times what Intel wouldn't pay. Until those SN50 boxes ship and benchmark in production, the valuation is a bet that regulated enterprises would rather own the rack than rent someone else's cloud.