Key Takeaways
- Nvidia’s equity investments in AI companies reached an estimated $99 billion by July 26, 2026, significantly up from $7 billion a year prior.
- The company is pursuing vertical integration across the entire AI stack, exemplified by the $12.93 billion acquisition of Hugging Face and a $30 billion commitment to OpenAI.
- Nvidia’s strategy includes deploying nearly $50 billion into frontier AI labs, extending the CUDA ecosystem’s financial and technical dominance. Nvidia’s equity investments in AI companies hit an estimated $99 billion as of July 26, 2026, according to the company’s latest financial filings, up from roughly $7 billion a year earlier and about $2.2 billion two years prior. The company committed more than $40 billion to AI investments in 2026 alone, spanning frontier model labs, cloud infrastructure, optical networking and developer platforms. The speed and scale of that capital deployment is rewriting how the AI industry is funded and who controls its direction.
Beyond the GPU Business
For decades, Nvidia built its dominance on selling GPUs, first to gamers, then to AI researchers who needed massive parallel compute for model training. The current investment surge represents a different kind of play: vertical integration across the entire AI stack, from foundational models down to the networking layer that moves data between chips.
The logic is defensive as much as it is offensive. AI development is maturing, and the bottlenecks are shifting up the stack. By taking equity stakes in the companies building on top of its hardware, Nvidia creates financial alignment that reinforces its technical position. Even if a credible rival chip emerges, a substantial portion of the AI software and services layer remains tied to Nvidia’s balance sheet. That alignment, in turn, drives continued GPU demand, the hardware that started the cycle.
The CUDA Ecosystem’s Financial Reach
Nvidia’s CUDA platform is the de facto standard for AI training and inference workloads, and the $99 billion in equity is extending that technical moat into a financial one. CFO Colette Kress confirmed Nvidia has invested “nearly $50 billion in the frontier AI labs,” noting that these labs routinely outspend their own balance sheets on compute. The investment provides the capital for them to keep buying GPUs at scale.
The clearest example is Nvidia’s roughly $30 billion commitment to OpenAI in February 2026, part of a larger $110 billion funding round. That investment effectively guarantees OpenAI’s continued reliance on Nvidia’s compute infrastructure. Nvidia also confirmed participation in Anthropic‘s Series G round in February 2026, though the amount was not disclosed. Both moves lock the leading AI labs into the CUDA ecosystem at the capital level, before a single chip is purchased.
Where the Money Is Going
The investment portfolio spans every layer of the AI stack. The largest tranche, close to $50 billion, has gone into frontier AI labs. Beyond OpenAI, Nvidia invested in xAI in January 2026 and is reportedly in discussions to deploy roughly $2.5 billion into Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati. Thinking Machines Lab has already committed to running Nvidia’s Vera Rubin computing platform at 1GW scale.
Cloud infrastructure is the second major target. Nvidia provided $2 billion to CoreWeave in January 2026 and a further $2 billion to Nebius two months later. These neocloud providers buy Nvidia GPUs in bulk and lease compute capacity to enterprises, Nvidia’s capital strengthens their balance sheets so they can keep purchasing at scale. Alongside direct investment, Nvidia announced in August 2026 a plan to mobilise more than $500 billion in third-party capital for AI infrastructure, through partnerships with Apollo, BlackRock and Goldman Sachs.
Data transfer is becoming a genuine bottleneck as model sizes and cluster counts grow, and Nvidia has committed to photonics and optical networking firms since March 2026.
On the software and developer tooling side, the $12.93 billion acquisition of Hugging Face on September 3, 2026 gives Nvidia direct control over the largest open-source model distribution platform in the industry. Nvidia’s $3.5 billion in convertible bonds into MediaTek in August 2026 deepens collaboration on custom chips and AI personal computing. The company is also reportedly in discussions to join a funding round for Perplexity at a post-money valuation above $30 billion.
Valuation Pressure Across the Market
A direct investment from Nvidia carries validation weight that moves valuations. Companies that receive it attract follow-on investors faster and at higher prices than they might otherwise command. Nvidia’s $30 billion commitment to OpenAI effectively reset what counts as a “strategic” investment in AI, and smaller corporate VCs are finding it harder to compete for stakes in deals where Nvidia has moved first.
Nvidia’s equity portfolio growth is also worth comparing with its peers: Alphabet and Amazon each hold significant equity investments across their businesses, so Nvidia has not yet passed either in total holdings, but the rate of growth within the AI sector specifically is faster and more targeted than both.
The Hugging Face Problem
Of all the moves in Nvidia’s portfolio, the Hugging Face acquisition carries the sharpest edge for critics. Hugging Face built its reputation on hardware neutrality, a platform where developers could share, fine-tune and deploy models regardless of what silicon they ran on. Placing that platform under a dominant GPU vendor changes the incentive structure, even if Nvidia maintains the neutrality commitment publicly.
The concern is architectural, not just competitive. Hugging Face sits at the model distribution layer: it is where developers discover models, where organisations pull weights for fine-tuning, and where a significant share of open-source AI tooling lives. If that layer tilts toward Nvidia-centric frameworks or hardware assumptions, the effect ripples through every project built on top of it. Antitrust scrutiny is a plausible consequence as Nvidia’s footprint extends from chips to cloud to the open-source developer layer simultaneously.
The Technical Imperative Behind the Capital
The investment strategy has a hardware rationale that goes beyond market control. Frontier AI labs are the first customers for each new GPU generation, and their workloads define what the next generation needs to do. Thinking Machines Lab’s 1GW Vera Rubin deployment, if it proceeds, is effectively a real-world validation run for Nvidia’s next-generation platform at a scale no internal test environment could replicate.
Co-development access matters too. Equity relationships open engineering channels, Nvidia’s hardware teams get earlier visibility into model architecture choices, memory bandwidth requirements and inference patterns than they would as a purely arms-length supplier. That feedback tightens the loop between GPU design and the workloads those GPUs will actually run. The companies Nvidia funds also have a financial incentive to report what is and is not working on the hardware, which is more candid signal than a standard procurement relationship produces. For context on how GPU supply constraints are shaping AI deployment decisions across the industry, that dynamic is running in parallel with Nvidia’s investment activity.
Concentration Concerns Surface
The scale of Nvidia’s equity position, $99 billion across the AI stack, is drawing scrutiny that goes beyond the Hugging Face deal. Startups that take Nvidia capital gain funding and validation, but they also take on a financial relationship with the company whose hardware they are expected to run. The incentive to explore alternative silicon, whether AMD, custom ASICs or other architectures, weakens when your largest investor sells the incumbents. For context on how enterprises are already weighing non-Nvidia chip options that pressure is real and growing.
The pattern Nvidia is building, capital into labs, capital into clouds, capital into the open-source layer, capital into the networking that ties it all together, means that any company trying to compete at the hardware level now faces a financial ecosystem that is structurally aligned against them. Whether regulators in the US or EU treat that as a competition problem depends on what they find when they look at the terms attached to these investments. Those terms have not been publicly disclosed.
Originally published at https://autonainews.com/nvidias-99-billion-ai-investments-drive-full-stack-control/
Top comments (0)