Poolside

Poolside is a US startup, founded in 2023, that is building a coding-specific foundation model rather than a general-purpose one. It is led by CEO Jason Warner, former CTO of GitHub, and was valued at $12 billion as of April 2026.

Snapshot

Valuation

Strategy and positioning

Poolside positions itself as a coding-specific frontier lab, distinct from general-purpose labs such as OpenAI, Anthropic, and Google DeepMind. Its bet is that specialization in coding can compete against the scaling advantages of those larger, general-purpose developers.

In June 2025, TechCrunch reported under the headline "Poolside CEO says most companies shouldn't build foundation models" that Warner takes the public position that foundation-model building is not a plausible strategy for most organizations, with Poolside's own coding specialization presented as the exception (Source: TechCrunch, June 2025).

Nvidia licensing deal

A letter to investors reported on August 20, 2026 described a non-exclusive licensing deal between Poolside and Nvidia worth $6 billion, together with a $1 billion Nvidia investment in Poolside at a $12 billion pre-money valuation. What Nvidia licensed, the term, and the payment structure are not established: the remainder of the article sits behind a subscription wall and neither company had confirmed the arrangement (Source: newcomer.co).

The purpose of the licence was reported two days later. According to people familiar with the matter in a Wall Street Journal account of August 22, 2026, Nvidia plans to use the $6 billion licensing deal to build one of the world's most powerful open-weight AI models, positioned against Chinese open-weight models including DeepSeek's and Kimi K3, and against OpenAI and Anthropic on the grounds that open-weight models are generally far cheaper to operate and allow easier customization. The remainder of that article also sits behind a subscription wall, so the licence terms, the model's timeline and Poolside's continuing role remain unestablished (Source: wsj.com). See Open-Weight Frontier Models.

Models

On April 28, 2026, Poolside released its flagship Laguna M.1 model alongside Laguna XS.2, its first open-weight model (Source: poolside.ai). Warner positioned the release toward high-security government and public-sector deployments, framing open weights as a procurement-trust requirement for sensitive deployments rather than as a community-development gesture. The combination retains a closed flagship while adding an open-weight model, placing Poolside closer to the open-weights approach associated with DeepSeek, Qwen, and Mistral on the open versus closed question.

By August 2, 2026 the line had reached a 2.1 generation. An open-model review by Florian Brand and Nathan Lambert described Laguna-S-2.1 as a 118B-parameter mixture-of-experts model with 8B active per token, sized to fit on an NVIDIA DGX Spark and released under the OpenMDW licence with published evaluation trajectories, alongside a smaller Laguna-XS-2.1 at 33B total and 3B active (Source: interconnects.ai). See Open-Weight Frontier Models.

The government and public-sector framing positions Poolside as a potential contender in US federal AI procurement (compare GSA OneGov Program and USAi Platform (August 2025)), particularly with buyers seeking weight-level inspection rather than API-only access.

The release coincided with other coding-stack repositioning the same day: Anthropic launched Claude for Creative Work plus an Opus opt-in layer, and OpenAI open-sourced its Codex orchestration tool Symphony.

Relationships

• related: Cursor (Anysphere), Cognition AI, Replit, AI Coding Agents, Open-Source AI / Open-Weight Models, GSA OneGov Program and USAi Platform (August 2025), METR — Measuring Impact of Early-2025 AI on Experienced Open-Source Developer Productivity.

Sources

• TechCrunch / Poolside CEO says most companies shouldn't build foundation models (June 2025)
• Poolside / Introducing Laguna XS.2 and M.1 (April 28, 2026): poolside.ai