FULL STORY

Nvidia's $7B Poolside Deal Unpacked

Nvidia was reported to be pursuing a ~$7B deal with Poolside; follow-ups clarified it involves $1B investment, a $6B non-exclusive license, and hiring 109 staff rather than a full acquisition.

2026-08-23 ~ 2026-08-25 · 2 episodes · 16 posts

Episode 1 · Nvidia's Reported $7B Poolside Deal Targets Top Open-Weight Model (2026-08-23, 14 posts)

According to the Wall Street Journal and multiple sources, Nvidia is pursuing a $7B deal with AI coding startup Poolside: $6B to license its Model Factory technology plus a $1B investment, with 100+ engineers moving to Nvidia to work on the open-weight Nemotron family—reportedly supporting a trillion-parameter Nemotron 4. It would be one of Nvidia's largest open-weight AI moves to date, aimed at building a US open-source camp to rival DeepSeek and Kimi. The deal remains at a proposed stage without official confirmation, and valuation details are disputed.

Confirmed

  • Per WSJ reports relayed by @mrgreatheart, @beffjezos, @kimmonismus, @BenBajarin and others: $6B tech license plus $1B investment, $7B total
  • Valuation accounts differ: @beffjezos cites a $12B post-money valuation, while @AI寒武纪 reports the $1B investment at a $12B pre-money valuation
  • Per @rohanpaulai, the $6B buys a non-exclusive license to Poolside's internal Model Factory platform—training, evaluation, RL, architecture testing, synthetic data generation and cross-GPU-cluster data mixing—not the finished Laguna product; @markk confirms the deal centers on Model Factory
  • 100+ Poolside employees (from the Poolside Sara team) will join Nvidia's Nemotron effort; @markk says this supports the trillion-parameter open model Nemotron 4
  • @ollama announced a partnership with Poolside AI to release open models, with Nvidia's team advancing Nemotron and other open projects
  • Per @gnukeith and @DeryaTR, Nvidia also plans a US-based entity aimed at challenging DeepSeek and Kimi
  • Per WSJ via @SumitGup, Poolside accepted because it struggled to raise enough capital for infrastructure and saw compute potentially running out next year

Unconfirmed

  • WSJ describes the deal as planned/proposed with no official announcement from either company; @AI寒武纪 claims it is "finalized," conflicting with WSJ. Terms, timeline, valuation basis, Nemotron 4 specs and the new US entity remain unconfirmed

Why it matters

  • Per @BenBajarin, Nvidia aims to build the world's strongest open-weight model, targeting both Chinese labs and US frontier companies, potentially reshaping open-model competition
  • @rohanpaulai and @markk note Nvidia is buying Poolside's model R&D pipeline itself rather than a finished model, signaling a bid for systematic model engineering capability
  • @gnukeith calls it one of Nvidia's largest open-weight AI plays; the new entity would directly benchmark against DeepSeek and Kimi
  • @kimmonismus sees Nvidia deepening its move from hardware vendor into model ecosystem, with an unprecedented $6B-scale open-model commitment
  • @ollama's announcement shows the Nemotron open ecosystem attracting third-party platforms; future open-weight releases merit attention
  • Poolside's funding and compute constraints suggest even star AI startups cannot independently afford frontier training infrastructure—acquisition by giants may become the norm

Episode 2 · Nvidia Pours $7B into Poolside to Boost Open-Source AI (2026-08-25, 2 posts)

Nvidia reportedly invested $1B in Poolside and paid $6B for non-exclusive rights to its 'model factory' tech while hiring its 109 employees, aiming to strengthen open-source AI ecosystems and spur demand for its computing power.