A single X essay from physicist and AI commentator Dr. Alex Wissner-Gross went viral this week with a set of eye-popping, very specific financial figures: NVIDIA's direct equity stakes in AI companies reportedly growing tenfold to $99 billion, a reported $105 billion in credit tied to OpenAI's Ohio data-center site, and Berkshire Hathaway's incoming CEO Greg Abel reportedly buying $10 billion of an Alphabet capital raise at a discount.
None of these numbers come from an NVIDIA 8-K, a Berkshire shareholder letter, or a named financial-news byline that we could independently verify. They come from one commentator's thread. That doesn't make them worthless — Wissner-Gross has a track record of reading public filings and secondary reporting closely — but it does mean every figure below should be read as "one commentator reported," not as confirmed fact. This post explains what the claims mean if true, what's already independently established elsewhere on this exact topic, and what builders should actually do with any of it.
TL;DR
| Claim | Source | Independently confirmed here? |
|---|---|---|
| NVIDIA's equity investment book grew ~10x to $99B | Wissner-Gross X thread | No — NVIDIA equity stakes in AI firms are real and disclosed, but not this total figure |
| $105B in credit tied to OpenAI's Ohio site | Wissner-Gross X thread | No — likely overlaps with the already-reported $250B Ohio backstop; numbers don't reconcile cleanly |
| Berkshire's Greg Abel bought $10B of Alphabet stock at a 6.5% discount | Wissner-Gross X thread | No — plausible given Berkshire's public AI-adjacent moves, but unconfirmed at this size |
| "No impact to the rates of our other customers" (Abel, on hyperscaler power) | Wissner-Gross X thread, attributed remark | No — reported quote, not independently sourced |
| Record 3,795 billionaires worth $15.1 trillion, 27% held by 29 people | Wissner-Gross X thread | Cited as the commentator's own statistic; not established here as AI-caused |
What "equity investment book growing tenfold" actually means
Selling a GPU is a transaction: NVIDIA gets paid, the buyer owns the hardware, and NVIDIA's exposure to that buyer ends at the invoice. An equity investment book is a different thing entirely — it means NVIDIA is taking direct ownership stakes in the AI companies and infrastructure projects that buy its chips, not just billing them.
NVIDIA has already disclosed doing exactly this in specific, named cases — including its position tied to OpenAI's compute buildout, part of the same relationship explainx.ai has covered in the NVIDIA-OpenAI Ohio financing coverage and the PORTS-Pike land-power-shell guarantee. The Wissner-Gross thread's claim is that this pattern has scaled up roughly tenfold, to a $99 billion total book, across multiple companies and projects — not just OpenAI.
If that's directionally correct — even if the exact number is off — it means NVIDIA's own balance sheet is now tied more directly to the success of its customers than a pure hardware-sales business would be. That's a meaningfully different risk profile. A chipmaker that only sells hardware gets paid whether or not its customer's business plan works out. A chipmaker holding equity in that customer only wins if the customer wins too — which changes NVIDIA's incentive to keep customers funded, keep deals flowing, and keep the overall AI capex cycle running.
The $105 billion Ohio credit claim: a discrepancy, not a new deal
This is the part worth being most careful about. The Wissner-Gross thread cites $105 billion in credit connected to OpenAI's Ohio data-center site. explainx.ai has already covered two rounds of reporting on financing tied to that same Ohio project:
- July 2026: Reuters-sourced reporting on a reported ~$250 billion NVIDIA financing backstop for OpenAI's ~10-gigawatt southern Ohio campus.
- August 2026: NVIDIA's own announcement of the PORTS-Pike land-power-shell (LPS) guarantee — a 20-year, up to 8-gigawatt commitment sitting inside a roughly $600 billion NVIDIA-OpenAI compute relationship through 2030.
A new $105 billion figure, attributed to "credit tied to the Ohio site," does not cleanly match either the $250 billion backstop or the $600 billion multi-year opportunity figure. There are a few honest possibilities: it could be a specific credit tranche that's a subset of the larger $250 billion figure; it could be an outdated or partial number the commentator picked up from an earlier report; or it could simply be describing the same underlying deal with a different, less-precise number attached.
We're flagging this explicitly as a discrepancy worth noting, not silently reconciling it by picking whichever number sounds more credible. Without a primary source tying the $105 billion figure to a specific instrument, the honest answer is: this may be the same deal already covered under a different number, and readers should not treat it as a fourth, additive financing commitment on top of what's already been reported.
Berkshire Hathaway and the Alphabet stake: why this one is notable if true
The most interesting claim in the thread, if accurate, isn't the NVIDIA number — it's Berkshire. The thread says Greg Abel, Berkshire Hathaway's incoming CEO, bought $10 billion of an Alphabet AI-related capital raise at a 6.5% discount, and that he checked with Warren Buffett before doing it.
Berkshire Hathaway has a long, well-documented history of studied caution around technology investing generally, and Buffett has specifically avoided most of the AI infrastructure buildout that's consumed the rest of the market's attention over the past two years. A $10 billion Alphabet stake — at a discount, specifically framed around AI capital-raising — would be a real signal that mainstream, historically risk-averse capital sees the current AI investment cycle as durable enough to bet real money on, not just a momentum trade for growth funds.
That's exactly why it deserves both attention and hedging in the same breath. This is a large, specific number from a single secondary source. If Berkshire made a disclosure or an SEC filing surfaces with this detail, it would be worth a follow-up post confirming it. Until then, treat it as reported, not established — the size of the claim ($10 billion) is itself a reason for extra scrutiny, not less.
The thread also attributes a remark to Abel — that hyperscaler power delivery for AI data centers would have "no impact to the rates of our other customers," presumably in reference to a Berkshire Hathaway Energy utility subsidiary. That's a quote worth flagging on its own: it speaks directly to one of the most contentious parts of the AI buildout story — whether ordinary electricity ratepayers end up subsidizing AI data-center power demand. If Abel actually said this, it's a notable public reassurance from someone overseeing utility operations. As with the investment figure, this is a reported remark relayed secondhand, not a quote we can verify against a transcript or press release.
What this means for what you build or pay
This is finance-adjacent news, and the honest takeaway for builders sits in two places at once.
The bullish read: heavy vendor financing and equity entanglement between chip suppliers and AI labs is a leading indicator of continued aggressive compute buildout. When NVIDIA is financing its customers' data centers and taking equity stakes in them, it's betting on more demand for its own chips down the line — which tends to mean more GPU capacity coming online, and historically has coincided with falling or flat token prices as capacity outpaces demand growth, the pattern explainx.ai tracked in NVIDIA's $500B compute-as-an-asset-class plan.
The systemic-risk read: the more circular this financing web gets — chipmaker funds the buildout, takes equity in the buyer, the buyer's biggest customer is itself financed by the same chipmaker's ecosystem — the faster a shock anywhere in that chain could propagate. In a normal supplier relationship, a customer's financial trouble is the customer's problem. When the supplier holds equity and has guaranteed credit lines into the same customer's infrastructure, a slowdown anywhere can ripple into reduced capacity, delayed data-center buildouts, or pricing volatility faster than a arm's-length hardware relationship would allow. That's the same circular-financing concern that's already shown up in NVIDIA's credit default swaps, as covered in the $500B asset-class post above.
Practically: if you're planning model spend, API pricing, or infrastructure decisions around the assumption that compute keeps getting cheaper and more available, the equity-entanglement pattern is a mild supporting signal — but it is not a guarantee, and a genuinely circular financing structure is exactly the kind of thing that can unwind faster than a simple supply-and-demand model would predict. Diversify vendor and model dependencies rather than betting entirely on one lab's compute pipeline staying intact.
The billionaire wealth stat: a correlated aside, not a causal claim
The same thread cites a statistic that a record 3,795 billionaires now hold a combined $15.1 trillion, with 27% of that concentrated in just 29 people. It's a striking number, and the thread implicitly ties it to the AI capital boom narrative it's describing elsewhere.
Worth being honest here: the thread doesn't actually establish a clean causal link between the AI investment cycle specifically and this wealth-concentration statistic. Billionaire wealth totals move with broad equity-market performance, real estate, commodities, and a dozen other factors that have nothing to do with AI. Some of the wealth concentrated at the very top almost certainly does trace back to chip, cloud, and AI-infrastructure holdings — NVIDIA's own market capitalization is part of that story — but treating the entire statistic as an AI-caused outcome overstates what the source actually shows. It's a correlated data point worth knowing, not proof of a mechanism.
What to watch next
- Whether NVIDIA discloses an aggregate equity-stakes figure in an upcoming earnings call or SEC filing — that would be the first primary-source confirmation or contradiction of the $99 billion number.
- Whether Berkshire Hathaway's own quarterly 13F filing shows a new Alphabet position at the scale described.
- Whether any named financial outlet independently reports the $105 billion Ohio credit figure with sourcing, which would help resolve whether it's a genuinely new instrument or the same $250 billion deal restated.
- Whether NVIDIA's credit default swap levels move in response to this thread gaining traction — a market reaction would itself be informative, separate from whether the underlying numbers are accurate.
Related on explainx.ai
- Nvidia's $500B Plan to Make GPUs an Asset Class
- NVIDIA Guarantees OpenAI's Ohio AI Factory: The PORTS-Pike LPS Deal
- Nvidia-OpenAI $250B Backstop: Ohio's 10GW Data Center
- NVIDIA Is Buying Hugging Face for $12.9 Billion
- Nvidia Reportedly Agrees to Buy Hugging Face for $12.9B
- Anthropic's Reported $517B Compute Commitments, Explained
- AI Giants Carry $1.65 Trillion in Off-Balance-Sheet Debt — Is It Another Enron?
- Ed Zitron's AI Predictions, Fact-Checked
This post is based entirely on claims made in a single viral X essay by Dr. Alex Wissner-Gross, not on primary financial reporting. Every specific figure — the $99 billion equity book, the $105 billion Ohio credit claim, the $10 billion Berkshire-Alphabet stake, the attributed Greg Abel quote, and the billionaire wealth statistic — should be read as "one commentator reported," not as independently confirmed fact. This post reflects information available as of September 7, 2026, and will be updated if primary sourcing emerges or contradicts these claims.