Nvidia Increased Stakes in Key Artificial Intelligence Infrastructure Companies in 2026

Nvidia's investment portfolio hit roughly $99 billion by the end of its second fiscal quarter (ended July 26, 2026), split almost evenly between public stocks, private company stakes, and equity-method investments. That number, disclosed in Nvidia's own filings and reported by Yahoo Finance and The Motley Fool on September 13, 2026, is larger than most coverage of "Nvidia's $40 billion AI push" earlier this year suggested, as enterprise demand for modern AI tech stacks continues to expand across industries.

Nvidia's $99 billion portfolio breaks into three parts, according to its own Q2 FY2026 filing: $47.7 billion in publicly held stocks, $47.9 billion in non-marketable securities (private company stakes), and $3.3 billion in equity-method investments.

The public stock side is led by Intel, SpaceX (which converted from an xAI-linked position after that company's merger), and CoreWeave. The private side is where the bigger frontier-lab bets sit — where specialized data engineering service providers and custom platform pipelines support high-volume model training. Nvidia CFO Colette Kress said on the Q2 earnings call that the company has invested nearly $50 billion in frontier AI labs specifically, broken down as roughly $30 billion in OpenAI, up to $10 billion in Anthropic, and $5 billion in Safe Superintelligence.

One number worth flagging: Nvidia's $12.9 billion acquisition of Hugging Face isn't counted in the $99 billion figure at all, because the deal was announced after the fiscal quarter closed. The real total, as of September 2026, is higher than the headline number.

Three increases account for most of the stake growth reported through 2026:

Two newer commitments haven't shown up as completed stakes yet but were announced in 2026: up to $3.2 billion in optical networking supplier Corning, and up to $2.1 billion in data center operator IREN, tied to a 5-gigawatt AI infrastructure partnership.

Nvidia's money clusters into four groups, and each one maps to a different part of the AI supply chain it depends on:

The logic is straightforward: a company that can't afford to build a data center or buy enough GPUs doesn't generate GPU revenue. Nvidia's equity stakes give its customers the capital to keep buying, which is exactly why some analysts call the pattern circular.

No, and this is the point most coverage blurs together. Nvidia's $99 billion portfolio is its own direct equity, money it owns as an investor. Separately, in August 2026, Nvidia announced a partnership with six major financial institutions (Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR) to mobilize up to $500 billion in third-party capital for AI infrastructure, mostly structured as debt rather than equity. Some later reporting put the figure as high as $640 billion as additional commitments were added.

The two programs work differently. Nvidia's equity stakes are direct ownership positions the company holds on its own balance sheet. The $500 billion Wall Street initiative is a financing facility that lets outside investors lend money to Nvidia's customers, with Nvidia agreeing to backstop up to $125 billion of it, about 25% of the total. CEO Jensen Huang framed the distinction this way on CNBC: the equity stakes are Nvidia picking winners, while the financing coalition is meant to treat AI compute itself as an investable asset class for institutions that don't want to buy chip company stock directly.

That depends on who you ask, and it's a genuinely open debate rather than a settled one. Wedbush Securities analyst Matthew Bryson told CNBC the dealmaking fits "squarely into the circular investment theme," pointing to the pattern where Nvidia funds a company that then spends a meaningful share of that capital back on Nvidia hardware.

The counterargument, made by Huang and echoed by KKR co-CEOs Joe Bae and Scott Nuttall, is that AI compute has become genuine infrastructure, comparable to toll roads or power plants, and that Nvidia is simply providing the capital needed to build it at the pace demand requires. Huang has put a number on that pace: building one gigawatt of AI data center capacity costs roughly $50 to $60 billion — leading many engineering executives to audit their existing tech stack to optimize compute efficiency.

Both things can be true. Nvidia's equity stakes do reduce its customers' cost of capital and secure long-term chip demand. They also mean a slowdown in AI spending would hit Nvidia twice, once as a hardware seller and once as an equity holder in the companies buying that hardware.

Some of them, yes, though this isn't investment advice and any decision should factor in your own research and risk tolerance. Of the companies in Nvidia's portfolio, the publicly tradable names are Intel, CoreWeave, Coherent, Nebius, Lumentum, and Marvell Technology. OpenAI, Anthropic, Safe Superintelligence, and SpaceX remain private, so retail investors cannot buy shares directly in those.

A structural detail worth knowing: Nvidia's $2 billion investment in Coherent went into common stock, so it shows up directly on Nvidia's public 13F filings. The Lumentum and Marvell investments, also $2 billion each, went into convertible preferred shares instead, which means they won't appear as common stock holdings on 13F filings until they convert.

Nvidia's stake increases in 2026, CoreWeave, Nebius, and Intel most visibly, aren't isolated bets. They're part of a nearly $99 billion portfolio built around securing the customers, capacity, and physical infrastructure Nvidia's chip business depends on, and they sit apart from the separate $500 billion Wall Street financing coalition announced the same year. If you're tracking this story going forward, the number to watch isn't the total dollar figure, it's whether the companies receiving Nvidia's capital start generating revenue independent of Nvidia hardware purchases, since that's what will settle the circular investment debate one way or the other.

To evaluate your own organization's software infrastructure, explore how CloudMotiv structures custom data systems, request a guided SaaS Stack Audit, or review our guide to software outsourcing companies.