据知情人士透露，英伟达计划向OpenAI投资高达1000亿美元，以帮助其训练和运行最新的人工智能模型，但由于这家芯片巨头内部一些人士对这笔交易表示怀疑，该计划目前已陷入停滞。
去年9月，两家公司在英伟达位于加州圣克拉拉总部公布了这项巨额协议。他们宣布了一份谅解备忘录，英伟达将为OpenAI构建至少10吉瓦的计算能力，这家芯片制造商还同意投资高达1000亿美元，以帮助OpenAI支付相关费用。作为协议的一部分，OpenAI同意从英伟达租赁芯片。
当时，据知情人士透露，这家ChatGPT的开发商预计交易谈判将在未来几周内完成。但一些知情人士表示，谈判至今仍停留在早期阶段。
现在，一些知情人士表示，双方正在重新考虑未来的合作关系。他们表示，最新的讨论包括英伟达对OpenAI进行数百亿美元的股权投资，作为其当前融资轮的一部分。
知情人士透露，英伟达首席执行官黄仁勋近几个月来私下向业内人士强调，最初达成的1000亿美元协议不具约束力，尚未最终敲定。一些知情人士还表示，黄仁勋私下批评了OpenAI在商业模式上缺乏纪律性，并对OpenAI面临来自谷歌和Anthropic等公司的竞争表示担忧。
OpenAI发言人表示：“我们的团队正在积极敲定合作细节。英伟达的技术从一开始就支撑着我们的突破，如今也为我们的系统提供动力，并将继续在我们未来扩展业务的过程中发挥核心作用。”
英伟达发言人表示，英伟达在过去十年一直是OpenAI的首选合作伙伴，并期待继续与该公司合作。
OpenAI正为2026年底上市奠定基础，并在过去一年中竭力获取大量计算能力，以支持其未来的产品开发和业务增长。与英伟达的合作协议陷入僵局，这对OpenAI的努力无疑是一次打击，也暴露出其首席执行官萨姆·奥特曼热衷于宣布高调的巨额交易，但如果条款尚未最终敲定，则可能适得其反。
在与奥特曼和OpenAI总裁格雷格·布罗克曼共同发布的9月份合作协议公告中，黄仁勋称该协议为“史上最大的计算项目”。消息公布后，英伟达股价上涨近4%，市值接近4.5万亿美元。据《华尔街日报》此前报道，作为协议的一部分，英伟达曾讨论为OpenAI计划用于建设自有数据中心的部分贷款提供担保。
此后，OpenAI又与多家芯片和云服务公司签署了一系列协议，推动了全球股市的上涨。但投资者随后开始担忧这家初创公司是否有能力支付这些交易，导致一些与OpenAI相关的科技股遭到抛售。奥特曼曾表示，这些交易使这家初创公司背负了1.4万亿美元的计算承诺——是其去年预期收入的100多倍。
OpenAI高管表示，考虑到部分交易存在重叠，实际承诺总额会更低，而且这些协议将在较长时间内逐步生效。
近期对OpenAI的担忧主要源于谷歌Gemini应用的成功，该应用减缓了ChatGPT的增长，并导致OpenAI发出红色警报。Anthropic凭借其广受欢迎的AI编码代理Claude Code也给OpenAI带来了压力。英伟达在11月宣布，将向Anthropic投资高达100亿美元。
英伟达在11月提交的文件中表示，无法保证“就OpenAI的合作机会或其他潜在投资达成最终协议，也无法保证任何投资能够按预期条款完成（如果最终能够完成的话）”。
次月，在亚利桑那州斯科茨代尔举行的瑞银会议上，英伟达首席财务官科莱特·克雷斯表示，公司尚未与OpenAI达成最终协议。
知情人士透露，黄仁勋曾向同事表示，他仍然认为以某种形式向OpenAI提供资金支持至关重要，部分原因是OpenAI是这家芯片设计公司最大的客户之一。如果OpenAI落后于其他人工智能开发商，可能会影响英伟达的销售额。
Anthropic公司严重依赖亚马逊网络服务公司（AWS）设计的名为Trainium的芯片以及谷歌自主研发的TPU处理器来训练其人工智能模型。谷歌则主要使用其TPU来训练Gemini模型。这两款芯片都对英伟达最畅销的产品——图形处理器（GPU）——构成了重大竞争威胁。
Nvidia’s plan to invest up to $100 billion in OpenAI to help it train and run its latest artificial-intelligence models has stalled after some inside the chip giant expressed doubts about the deal, people familiar with the matter said.
The companies unveiled the giant agreement last September at Nvidia’s Santa Clara, Calif., headquarters. They announced a memorandum of understanding for Nvidia to build at least 10 gigawatts of computing power for OpenAI, and the chip maker also agreed to invest up to $100 billion to help OpenAI pay for it. As part of the deal, OpenAI agreed to lease the chips from Nvidia.
At the time, the ChatGPT-maker expected the deal negotiations to be completed in the coming weeks, people familiar with the plans said. But the talks haven’t progressed beyond the early stages, some of the people said.
Now, the two sides are rethinking the future of their partnership, some of the people said. The latest discussions, they said, include an equity investment of tens of billions of dollars as part of OpenAI’s current funding round.
Nvidia Chief Executive Jensen Huang has privately emphasized to industry associates in recent months that the original $100 billion agreement was nonbinding and not finalized, people familiar with the matter said. He has also privately criticized what he has described as a lack of discipline in OpenAI’s business approach and expressed concern about the competition it faces from the likes of Google and Anthropic, some of the people said.
“Our teams are actively working through details of our partnership. NVIDIA technology has underpinned our breakthroughs from the start, powers our systems today, and will remain central as we scale what comes next,” an OpenAI spokesman said.
An Nvidia spokeswoman said that Nvidia has been OpenAI’s preferred partner for the past 10 years and it looked forward to continuing working with the company.
OpenAI is laying the foundationto go public by the end of 2026, and has spent much of the past year racing to secure large amounts of computing capacity to help power OpenAI’s future products and growth. The stalled Nvidia pact is a blow to this effort and shows how CEO Sam Altman’s penchant for announcing flashy big-ticket deals carries the potential to backfire if the terms have yet to be finalized.
In a joint announcement unveiling the September deal with Altman and OpenAI President Greg Brockman, Huang called the deal “the largest computing project in history.” Nvidia’s stock rose by nearly 4% on the news, pushing its market value to almost $4.5 trillion. As part of the deal, Nvidia discussed guaranteeing some of the loans OpenAI planned to take out to build its own data centers, The Wall Street Journal previously reported.
OpenAI went on to sign a string of other agreements with chip and cloud companies that helped fuel a global stock-market rally. But investors have since grown jittery about the startup’s ability to pay for these deals, leading to a selloff in some tech stocks tied to OpenAI. Altman has said that the deals put the startup on the hook for $1.4 trillion in computing commitments—more than 100 times the revenue it was on pace to generate last year.
OpenAI executives say the total commitments are lower after accounting for overlap in some of the deals, and that the agreements will take place over a long period.
Much of the recent concern about OpenAI has come from the success of Google’s Gemini app, which slowed ChatGPT’s growth and led OpenAI to declare a code red. Anthropic is also putting pressure on OpenAI thanks to its popular AI coding agent, called Claude Code. Nvidia said in November that it was committing to invest up to $10 billion into Anthropic.
In a November filing, Nvidia said there was no assurance that it would “enter into definitive agreements with respect to the OpenAI opportunity or other potential investments, or that any investment will be completed on expected terms, if at all.”
At a UBS conference in Scottsdale, Ariz., the following month, Nvidia CFO Colette Kress said that the company hadn’t completed a definitive agreement with OpenAI.
Huang has indicated to associates that he still believes it is crucially important to provide OpenAI with financial support in one form or another, in part because OpenAI is one of the chip designer’s largest customers, people familiar with the matter said. If OpenAI were to fall behind other AI developers, it could dent Nvidia’s sales.
Anthropic relies heavily on a combination of chips designed by Amazon Web Services known as Trainium, as well as Google’s in-house designed TPU processors, to train its AI models. Google largely uses its TPUs to train Gemini. Both chips represent major competitive threats to Nvidia’s bestselling products, known as graphics processing units, or GPUs.