Nvidia has quietly disclosed plans to invest $26 billion in building open-weight AI models, according to regulatory filings recently uncovered. The move represents the most significant strategic shift for the chip maker since it pivoted to AI four years ago, transforming it from a supplier of computing infrastructure into a direct competitor with the very customers that made it dominant.
The disclosed spending will fund model development, research talent, computing infrastructure, and what the filings describe as "ecosystem development" likely including partnerships and acquisitions. To put the scale in perspective: OpenAI reportedly spent around $3 billion training GPT-4, meaning Nvidia could feasibly develop multiple frontier models with budget left over.
This is not a casual venture. The move puts the GPU king in direct competition with OpenAI, Anthropic, and upstart DeepSeek, companies that have until now been Nvidia's biggest customers. The company already holds stakes in both OpenAI and Anthropic. It also ships the hardware on which virtually every major AI lab depends.
The strategic calculation
Nvidia's reasoning appears straightforward on the surface. If Nvidia captures even 10% of the foundation model market while maintaining its hardware dominance, it could add $50 billion annually to revenue within three years. The company controls the compute infrastructure everyone else needs, giving it structural advantages that few competitors possess.
But the risks are substantial. Microsoft, Amazon, and Google—Nvidia's three largest cloud customers—are all invested in competing model providers. They're also developing their own AI chips to reduce dependence on Nvidia hardware. A direct move into model development could accelerate those efforts, particularly if customers feel locked out of Nvidia's forthcoming models or worry about sharing training data with a direct competitor.
The timing also matters. The filings indicate spending will ramp over 18-24 months, suggesting first releases could arrive late 2026 or early 2027. That window is compressed in an industry where progress accelerates monthly. DeepSeek and Qwen have gone from 1% combined global AI market share in January 2025 to roughly 15% by January 2026—the fastest adoption curve in AI history.
The competitive pressure
The decision signals Nvidia's concern about being rendered structurally redundant. Chinese open-source models are now outcompeting proprietary Western alternatives on cost and efficiency, while competing on capability. V4 was optimized for Huawei Ascend and Cambricon chips, with DeepSeek withholding early access from Nvidia and AMD. For self-hosting customers, it means V4 runs efficiently on a broader range of hardware than most Western models that assume NVIDIA CUDA throughout the stack.
That matters because it breaks the dependency chain that Nvidia built. If frontier models can run efficiently on diverse hardware, customers are less locked into Nvidia's pricing and availability. Once the software moat erodes, the hardware business becomes commoditised.
Conflicting interests
There is also the awkward reality of Nvidia's existing stakes in OpenAI and Anthropic. The investment Nvidia finalized just last week as part of OpenAI's $110 billion round came in at $30 billion—well short of that earlier pledge. CEO Jensen Huang indicated recently that those investments are likely Nvidia's last, citing IPO timelines.
But the deeper issue is structural: Nvidia invests in companies that buy Nvidia chips, then competes with those same companies in the software layer. Nvidia sits in a position few companies have ever occupied: both supplier and shareholder to the firms building the software atop its hardware. That arrangement, once mutually reinforcing, now appears increasingly tangled.
A pragmatic but risky pivot
The decision reflects a sober assessment of Nvidia's long-term position. Staying purely as a hardware supplier leaves it vulnerable to commoditisation if open-source models eliminate the performance rationale for premium chips. But competing directly with customers risks accelerating their abandonment of Nvidia hardware.
What's clear is that Nvidia cannot remain passive. The open-source AI landscape is shifting rapidly, and a company that controls both cutting-edge chips and competitive models would occupy extraordinary market power. The question is whether the financial return justifies the customer relationships at risk. For Nvidia investors, the next 18 months will reveal whether this is a visionary move or a costly miscalculation.