Nvidia commits $26B to open AI models, expands beyond chipmaking
Strategic pivot creates 'full-stack AI laboratory' to challenge OpenAI and Chinese competitors
Nvidia is embarking on a $26 billion strategic transformation that will extend its dominance from artificial intelligence hardware into software development, as the company announced plans to build its own open-source AI models over the next five years.
The investment, confirmed in regulatory filings and by company executives, marks Nvidia's evolution from being primarily an AI chip supplier to becoming what analysts describe as a "full-stack top-tier AI laboratory." The company expects to release its first self-developed open-source models by late 2026 or early 2027, with significant capital deployment expected over the next 18 to 24 months.
"This is NVIDIA's most ambitious bet yet on owning the entire AI stack," said analysts covering the company. "By developing its own models, NVIDIA is positioning itself to compete directly with OpenAI, Anthropic, and Chinese developers like DeepSeek."
Nvidia's approach focuses on "open-weight" models, which means the core parameters that dictate AI behavior will be disclosed and available for free download. This allows enterprises and developers to run or fine-tune the models on their own infrastructure or private clouds, addressing corporate demands for data privacy, customization, and cost control.
The company has already completed pre-training of a 550 billion-parameter model, providing technical validation for its open-source development program. Key focus areas include multimodal applications spanning language, coding, scientific computing, and intelligent agents.
This move comes as Nvidia continues an aggressive AI investment spree. Separately, the company is reportedly finalizing a $30 billion equity investment in OpenAI, though that figure represents a reduction from a previously discussed $100 billion stake. Chief Executive Jensen Huang has indicated this may be one of Nvidia's last significant equity investments before OpenAI's anticipated initial public offering in late 2026.
The open-source strategy has garnered support from technology leaders, including Block Chief Executive Jack Dorsey, who praised Nvidia's $26 billion bet on open AI models, calling the approach "excellent." Dorsey, a vocal advocate for open technologies, has long argued that broader access to large language models enables global developers to build upon the technology.
Nvidia's move contrasts with competitors' strategies. Meta Platforms initially released its Llama model as open-source in 2023 but has since indicated future models may not be fully open. OpenAI offers an open-weight model called GPT-OSS, though it remains less advanced than the company's proprietary systems and offers limited modification flexibility.
The announcement arrives as Nvidia's stock shows signs of being oversold, with a 14-day Relative Strength Index at 27. Shares rose 0.7% in Wednesday trading to $186, giving the company a market capitalization of approximately $4.44 trillion. The stock remains well below its 52-week high of $212.18, despite impressive financial metrics that include a 95.6% year-over-year increase in quarterly earnings and a 73.2% surge in quarterly revenue.
Analysts remain overwhelmingly bullish on Nvidia's prospects. Of 61 analysts covering the stock, 58 rate it as a strong buy or buy, while just two recommend hold and one suggests sell. The average price target stands at $266.35, implying roughly 43% upside from current levels.
Nvidia is expected to provide additional details about its open-source strategy at the company's GPU Technology Conference (GTC) running March 16-19 in San Jose, California. Huang's keynote address is anticipated to outline the company's "full-stack" approach, encompassing chips, software, models, and applications.
However, some analysts have raised concerns about potential "circular financing" risks given Nvidia's investments in major customers like OpenAI. These analysts suggest the close ties between Nvidia's hardware business and its model development could invite regulatory scrutiny, particularly as the company's influence across the AI ecosystem continues to expand.
The $26 billion commitment surpasses the $3 billion that OpenAI reportedly spent training GPT-4, underscoring the scale of Nvidia's ambition. By developing models optimized for its own GPUs, the company aims to create a self-reinforcing ecosystem where software and hardware work in concert, potentially solidifying its competitive moat against rival chipmakers and Chinese developers alike.