Nvidia discusses boosting Reflection investment, computing deal
Reflection AI debuted in October its first open-weight model, a market dominated by Chinese startups.
Nvidia is reportedly in early-stage discussions to either increase its investment in Reflection AI or provide the startup with enhanced computing resources, according to a source familiar with the matter. The talks have not yet reached a conclusion, and no definitive terms have been agreed upon, the source emphasized. Nvidia has not yet addressed the inquiry, whereas Reflection chose not to comment on the matter.
Earlier reports from The Financial Times suggested these discussions could potentially culminate in an acquisition proposal. Potential deal structures may encompass an "acqui-hire," wherein Nvidia would hire Reflection's personnel and license its technology, an agreement to furnish the startup with additional chips, or an elevation in Nvidia's equity stake in the company.
Reflection introduced its inaugural open-weight model, Beam, in early October, entering a sector where Chinese startups hold considerable influence. Beam is engineered to excel for business clients and developers tackling coding and AI agent tasks, while simultaneously being more cost-effective than many competing offerings, Reflection disclosed on October 5.
The startup, founded by two ex-DeepMind researchers, has been engaged in talks to raise US$2.5 billion (S$3.2 billion) at a valuation of US$25 billion, as previously reported by the Wall Street Journal. Reflection has also finalized multimillion-dollar agreements with SpaceX and Nebius Group NV to secure computing capabilities for model development.
This potential collaboration with Nvidia would contribute to a surge of transactions witnessed by the chipmaker over recent months. Nvidia, presently the world's most valuable company, has leveraged its financial resources to foster the broader AI ecosystem, including a US$13 billion acquisition of AI model firm Hugging Face in September and a roughly US$20 billion licensing pact with Groq, a chip startup.
Nvidia's Chief Executive Officer, Jensen Huang, is a staunch advocate for open-weight models. The company's aim is to encourage US development of such technology, potentially curbing the adoption of Chinese open-weight models among startups. These models are significantly cheaper than those from American entities like OpenAI and Anthropic PBC, yet closely match in intelligence.
The US government is also promoting the development of open-weight models, viewing them as potential global standards in certain sectors of business and academic research. President Donald Trump's July 2025 AI action plan underscores this commitment, stating, "We need to ensure America has leading open models founded on American values."
Written by urgent.news from Straits Times Business's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
