Nvidia is deploying seven billion dollars to enter the open-weight model race, directly targeting DeepSeek and OpenAI.
Nvidia is accelerating its push into the open-weight artificial intelligence model arena through a two-pronged transaction valued at $7 billion. The AI chip giant reached an agreement with startup Poolside this week, securing a $6 billion technology licensing deal and making a $1 billion investment at a $12 billion pre-money valuation. As part of the arrangement, more than 100 Poolside engineers will join Nvidia’s Nemotron open-weight model program to help develop the largest and most technically complex version of the series to date. Poolside’s management team will not join Nvidia and will continue operating independently. The move directly challenges China’s DeepSeek and Kimi K3 while intensifying competition with U.S. AI leaders OpenAI and Anthropic, further complicating Nvidia’s co-opetition dynamic with key chip customers.
According to The Wall Street Journal, Poolside disclosed the terms in a letter to shareholders. The $6 billion licensing fee grants Nvidia access to Poolside’s model development technology. Founders Eiso Kant and former GitHub CTO Jason Warner wrote that a primary goal of the transaction is to ensure future artificial general intelligence does not become controlled by a handful of companies behind closed doors, but rather can be built openly by a broader community. Notably, executives including Kant, Warner, and Chief Operating Officer Margarida Garcia will remain independent to pursue undisclosed research work.
Open-weight models allow developers to download them freely and modify weights for specific needs, typically offering lower usage costs and greater customization flexibility. By contrast, the U.S. AI industry has historically devoted more resources to closed models. Major players such as OpenAI, Anthropic, and Google do not make full model weights or source code available to developers. This divergence has become a critical fault line in the U.S.-China AI rivalry. Because U.S. firms have not prioritized open models, global enterprises and nations are increasingly turning to Chinese-developed open-weight alternatives.
Nvidia CEO Jensen Huang has long championed open-weight architectures, visibly accelerating related initiatives over the past six months. In March, Nvidia established the Nemotron Coalition, uniting major open-model developers including Mistral, Thinking Machines Lab, and Perplexity to share data, expertise, and computing resources. Last month, Huang posted his first-ever message on X, publishing an open letter titled “Open Weights and American AI Leadership.” He emphasized that whether the United States wins the AI race depends not on a single most-advanced model, but on building a robust open ecosystem and extending AI technology across all industries.
The rapid ascent of Chinese AI models forms a critical backdrop to Nvidia’s strategy. In January 2025, DeepSeek released its first low-cost open-weight model, triggering sharp volatility in U.S. tech stocks. Since then, the U.S.-China AI race has intensified, with China producing a steady stream of new entrants like Moonshot’s Kimi and Z.AI’s GLM series, narrowing the technological gap with advanced U.S. systems to just a few months.
This competitive landscape places Nvidia in a delicate position. OpenAI, Anthropic, and other major labs are among Nvidia’s most important chip customers, yet they are also the primary developers of closed models. As Nvidia strengthens its own open-weight offerings, cooperative ties increasingly incorporate competitive friction. These labs are already developing custom silicon to reduce reliance on Nvidia. Anthropic launched a custom chip design division this month, recruiting Amir Salek, who previously led Google’s TPU development. OpenAI partnered with Broadcom in June to co-develop its own inference chip, “Halapeño,” marking direct competition with Nvidia’s hardware business. Additionally, other Nvidia-backed open-model developers could rival Nemotron. In March, Reflection AI, another Nemotron Coalition participant described by investors as the “DeepSeek of the West,” entered large-scale fundraising talks at a $25 billion valuation to raise $2.5 billion.
While Poolside and Nvidia maintained a working relationship, the transaction was finalized rapidly over recent weeks. Nvidia’s proposal coincided with Poolside’s release of its latest model, “Laguna S,” which has quickly become one of the most popular open-weight models in Western markets. Poolside’s shareholder letter revealed that a failed 2025 fundraising round was a decisive catalyst. “Late last year, we had six weeks to raise $2 billion to pay for a computing cluster equipped with 40,000 GB300 chips that was set to go live in January,” the letter stated. Poolside missed the deadline, lost access to the cluster, and reassessed its trajectory. The management team noted the company “picked itself back up and returned to work,” but soon realized that without adequate funding and data center resources, its computing capacity would be exhausted as early as next year. Nvidia emerged as an ideal partner precisely because it faces relatively fewer infrastructure and cash constraints.
Nvidia’s strategic pivot arrives as the global AI industry enters a “full-stack competition” phase. Model performance and cost are no longer dictated by isolated chips or algorithms, but by the integrated efficiency of silicon, models, software ecosystems, and data centers. On the capital markets front, Alibaba announced on August 23 the launch of its first new share placement since its 2019 listing, raising HK$80 billion (approximately $10.2 billion). All proceeds are earmarked for full-stack AI capabilities and infrastructure development. Market sources indicated the placement was quickly oversubscribed, drawing particularly strong demand from high-quality long-term investors such as sovereign wealth funds.
In the United States, AI lab valuations and public market ambitions continue to surge. Anthropic confidentially filed for an IPO with the U.S. Securities and Exchange Commission in June, targeting a potential October listing and a $2 trillion valuation. OpenAI is similarly aiming for a fourth-quarter listing, carrying a current private market valuation of approximately $850 billion.
Nvidia reports earnings on Wednesday, August 26. While the Poolside transaction is unlikely to immediately alter its profit structure, it signals Nvidia’s intent to leverage the cash and computing advantages accumulated from its GPU business to expand into the AI model layer. The central question for the market remains: Can Nemotron achieve sufficiently broad adoption to grant Nvidia meaningful influence over another tier of the AI stack, without creating excessive friction with its largest chip customers? Analyst sentiment remains highly bullish, with 26 analysts issuing buy ratings over the past three months and an averag