Nvidia, which has poured tens of billions of dollars into AI hardware companies such as chip manufacturers and data centers, is turning its attention to AI software. The chipmaker has recently been challenging a U.S. market dominated by closed models with Nemotron, its own open-source model, and the moves are seen as a decision to accelerate that push.
The Information, a U.S. technology news outlet, reported on the 23rd, citing sources, that Nvidia is in talks to join a new equity investment round in AI startup Perplexity. The round values Perplexity at more than $30 billion (about 41 trillion won), with the fundraising expected to reach several billion dollars.
Nvidia and Perplexity already have close business ties. In March, Perplexity joined the "Nemotron alliance" that supports the development of Nvidia's open-source models. Nvidia has already invested in Perplexity several times, and Nvidia CEO Jensen Huang has said in an interview that Perplexity is his favorite chatbot.
Nvidia had previously considered an investment that would license some of Perplexity's technology and absorb its employees, according to reports. Last year, Nvidia signed similar licensing and hiring deals with Groq, an AI inference chip design firm, and Enfabrica, a networking hardware design firm. It is not clear why those discussions shifted to an equity investment, The Information said. If the deal is completed, Perplexity is expected to gain a stable footing to manage its business through a planned initial public offering targeted for 2028.
Nvidia also drew attention the previous day with a large partnership deal with an AI software startup. The Wall Street Journal reported on the 22nd, citing sources, that Nvidia had struck a $6 billion (about 8.3 trillion won) partnership with AI startup Poolside. Nvidia will pay $6 billion to license Poolside's technology and hire about 100 engineers to work on developing Nemotron, its open-source project. Separately, it invested $1 billion (about 1.4 trillion won) to take an equity stake.
Poolside, founded in 2023 by software developer Eiso Kant and former GitHub Chief Technology Officer Jason Warner, has drawn attention for its coding-automation model built for government and defense use. Poolside reportedly weighed seeking outside investment after facing a financial crunch late last year, when it failed to raise the $2 billion needed to secure a large graphics processing unit (GPU) cluster within six weeks.
Nvidia, Which Built an AI Kingdom, Moves to Secure the Open-Weight Camp
Nvidia's back-to-back decisions to invest in AI software startups are seen as part of its recent push to expand the market for open AI models. Nvidia has long invested in AI software companies such as OpenAI, but the main goal there was to sell its own hardware, including GPUs, by supporting those firms. Its recent investments, by contrast, are seen as focused on improving and expanding Nemotron, Nvidia's open-source model.
The U.S. AI market is currently centered on closed models such as those from OpenAI and Anthropic. Open-weight models, in contrast to closed models, release their weights after training is complete, allowing anyone to download, run or tune them, which makes them well suited to developing customized AI programs. That is why models from Chinese AI startups, which have focused on open-weight models, have also gained attention in the United States.
Against this backdrop, 35 organizations, including Nvidia, Meta, Microsoft, Hugging Face, Mistral and Palantir, along with other companies and investors, last month issued a statement titled "Open Weights and American AI Leadership," arguing that for the United States to maintain its lead in the AI era, it must actively foster an open-weight AI ecosystem as well as closed AI. Huang stressed at the time that open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty, and that the world needs both frontier closed models and frontier open models. He also said Nvidia is pursuing a strategy of not merely supporting outside open models but developing open models itself and releasing them to the world. In particular, he explained that the company is releasing open models and related data across a range of fields, including autonomous driving, robotics and digital biology.
This could effectively put Nvidia in competition with its largest customers, major AI developers such as OpenAI, as well as with open-weight startups it has invested in, such as Reflection AI and Thinking Machines Lab.
The Information assessed that "Nvidia has concluded that it has no choice but to pursue this as a counterweight to the rise of open-weight AI models led by Chinese developers." Foreign media also noted the calculation that if Nvidia offers open-weight AI models made in the United States rather than in China, it could ultimately drive even greater demand for Nvidia's hardware.
