On August 24, 2026, NVIDIA shook the artificial intelligence world with the most expensive deal in language model history: a $6 billion agreement to license the 'Model Factory' system from AI startup Poolside, plus an additional $1 billion investment valuing Poolside at $12 billion pre-money. In total, NVIDIA is committing $7 billion to build the world's most powerful open-weight AI model under its Nemotron project, while recruiting more than 100 Poolside engineers in the process. This is not a conventional acquisition — Poolside continues operating independently with its leadership intact — but Poolside's core technology now powers NVIDIA's most ambitious bet to dominate not just AI hardware, but the models running on top of it.
What Did NVIDIA Announce with the Poolside Deal?
NVIDIA licensed Poolside's 'Model Factory' — the internal system the startup relied upon to develop and train its AI models — and 109 of its specialized engineers are joining NVIDIA to work exclusively on the Nemotron project. The stated goal: build the most capable open-weight model available, capable of outperforming closed models from OpenAI and Anthropic, as well as Chinese open models like DeepSeek V4-Pro, Kimi K3, and Qwen3.8-Max (which currently leads at 2.4 trillion parameters). Nemotron will be free to download and use, meaning zero licensing cost for any company including SMBs. Poolside described the arrangement as 'not an acquisition and not an acquihire' — yet in practice, NVIDIA gains frontier model-building capacity without formally purchasing the company.
"When NVIDIA — the company that makes 80% of the chips that train every AI model in the world — decides to build its own open models, the equation changes for everyone: costs drop, access is democratized, and SMBs gain real options."
Davarion Group & LabsReal Impact for SMBs
- 01Free access to frontier models: Nemotron will be open-weight, meaning any business can download and run it without paying per-token licensing fees — a dramatic cost reduction compared to GPT-5.6 or Claude Opus 5.
- 02More competition means lower prices industry-wide: with NVIDIA entering as a model provider, OpenAI and Anthropic will need to cut prices or add value, benefiting businesses already using AI APIs.
- 03Key consideration: open-weight models require your own infrastructure for optimal deployment — it's not plug-and-play like an API; SMBs without a technical team will need an implementation partner.
- 04Recommended immediate action: evaluate whether your current AI workloads (customer support, data analysis, content generation) could migrate to an open model and reduce costs by up to 70%, with the help of a specialized partner.
NVIDIA's move is profoundly strategic and extends far beyond the dollars invested. By controlling both the hardware (H100, H200, Blackwell GPUs) and the world's most powerful AI model, NVIDIA positions itself as the full-stack AI provider: selling the picks and shovels, and also the map to where the gold is. For Latin American and U.S. SMBs, this represents the most concrete opportunity to date to access enterprise-grade AI models without the subscription costs of closed labs. The arrival of Nemotron could accelerate the adoption of autonomous AI agents in sectors such as healthcare, logistics, manufacturing, and professional services — areas where cost-per-token has been the main barrier to entry.
At Davarion Group & Labs, we understand that news like this reshapes the technological options for your business — but also adds complexity: when to migrate, to which model, how to integrate open models into your current processes without disrupting operations? Our autonomous AI agents are designed to be model-agnostic, meaning that when Nemotron becomes publicly available, we can migrate your automation flows to leverage lower costs without rewriting your infrastructure. If you run a business in Houston TX or across Latin America and want to prepare for the next open AI cycle, let's talk at davarion.com.
