Last updated: July 28, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
Nvidia's $750B in deals reignite circular AI fears for a simple reason: the money keeps flowing back to Nvidia. The two biggest pieces are a $500 billion-plus pact with SK Group and a proposed $250 billion backstop for OpenAI's Ohio data center. Both send new spending back toward Nvidia's own chip sales. Markets noticed fast. Nvidia shares fell almost 5%, and its credit default swaps posted their sharpest single-day jump on record the same day the news broke.
Short answer: Nvidia's $750B in deals reignite circular AI fears because the total combines a $500B+ pact with SK Group and a proposed $250B backstop so OpenAI can lease a SoftBank-built data center — both set up so Nvidia's money flows back into demand for Nvidia's own GPUs. In my testing, AI research tools surface the numbers fast, but only primary reporting confirms what's actually signed.
I test AI tools for a living. Part of that job is asking AI assistants to summarize breaking financial news, and this story shows why that's risky without a second check. The $750 billion headline sounds like one check Nvidia is writing. It's actually two different kinds of deals, announced days apart, with different levels of proof behind them. Below is what's actually confirmed, plus the process I used — and that you can copy — to sort the real numbers from the speculative ones.
What's actually happening
Bloomberg reported on July 27, 2026 that Nvidia is working on a fresh round of AI deals worth more than $750 billion. The setup is reviving fears of "circular" AI financing: money that flows from Nvidia to a partner, then back to Nvidia as chip sales, without much new cash entering the system. Two deals make up most of the total.
The first is a partnership with South Korean company SK Group. It was unveiled the Friday before. CEO Jensen Huang said it represents more than $500 billion in business between the two firms. That figure mixes Nvidia's memory purchases from SK Hynix with SK Group's purchases of Nvidia GPUs and supercomputers. It's the value of long-term trade, not a cash check either side is writing today. As part of the deal, Nvidia and SK Group will build more than 2 gigawatts of AI data centers on the Korean Peninsula. The first site opens next year through SK Telecom.
The second is a proposed backstop for OpenAI. Nvidia is in talks to guarantee up to $250 billion so OpenAI can lease power from a 10-gigawatt data center. SB Energy, SoftBank's power arm, is building it on a former uranium-enrichment site about 50 miles south of Columbus, Ohio. Nvidia is also weighing up to $350 billion in financing to help OpenAI buy Nvidia GPUs for that site. That figure sits outside the $750 billion headline and hasn't been settled yet. Unlike the SK Group pact, the OpenAI backstop was still early-stage and unsigned as of the report.
Both deals draw the same circular-financing pushback. Nvidia's cash, or its credit guarantee, helps fund the customer that then buys Nvidia's chips. Bloomberg summed up the risk this way: the setup works fine "unless one of us has a problem, in which case both of us have a problem." If OpenAI's or SK Group's AI bet falls short, Nvidia is exposed on the supplier side and the lending side at once. Gary Tan of Allspring Global Investments told Bloomberg that while the deals support confidence in long-term AI buildouts, investors "remain concerned about circular financing."
Markets moved within a day. Nvidia stock dropped nearly 5%. Five-year credit default swaps on Nvidia's own debt jumped about 0.14 percentage points, to roughly 0.82 percentage points. That's the largest single-day move since Nvidia's swaps started trading, back in November 2025, per AI Weekly’s July 27, 2026 report. A credit-market move like that isn't hype. It's bond buyers pricing in more default risk, tied directly to how these deals are built.
What you'll need
You don't need a Bloomberg terminal to check a story like this. You need the habit of splitting "announced" from "signed," plus a couple of free tools. Open an AI research assistant with live web access — Perplexity, ChatGPT with browsing, or Claude with search — and pull up the original wire story instead of a summary of a summary. Ten minutes is enough to confirm the headline number, the deal structure, and whether the market reaction is real.
Step-by-step: how to check a mega-deal story like this yourself
1. Find the original reporter, not the copycat
Dozens of outlets ran versions of this story within hours, most citing Bloomberg's July 27, 2026 report. In my testing, copycat versions often dropped the line between the confirmed $500 billion SK Group pact and the still-in-talks $250 billion OpenAI backstop. They flattened both into one "$750 billion deal" headline. Always trace a number back to the outlet that broke it.
2. Split "signed" from "in talks"
This is the single most useful check on any mega-deal headline. Here, the SK Group figure came from Jensen Huang describing an unveiled partnership. The OpenAI backstop was reported as talks that hadn't closed. Treating both as equally final is how this kind of story gets blown up on social media.
3. Ask an AI tool to show its source, then open it
When I asked an AI assistant to summarize the deal, it correctly split the $500 billion and $250 billion figures — but only after I asked it to name where each number came from. Don't take a fluent summary at face value. Ask: "which article gave you that number, and on what date?" Then open that article yourself.
4. Check the market move against a real data source
A stock move or credit-spread number is either real and checkable, or it's a guess from a chatbot's training data. I confirmed the roughly 5% Nvidia share drop and the CDS spread move against AI Weekly's same-day market report, rather than trusting an AI summary of "the stock fell" with no number attached.
5. Read one skeptic and one company-side take
Circular financing stories tend to run from the critic's angle (systemic risk) or the company's angle (long-term demand). Reading one of each — here, the credit-market coverage versus Huang's framing of the SK Group deal as ordinary trade volume — gives a fuller picture than either one alone.
Example prompts you can copy
- Force a source check: "Summarize the Nvidia $750 billion deal news, and for every dollar figure, tell me the exact outlet and date it came from."
- Split confirmed from speculative: "Which parts of this Nvidia deal are signed, and which are still in talks or unconfirmed?"
- Explain the mechanism: "Explain in plain terms why analysts call this deal structure 'circular financing,' and what would have to go wrong for it to matter."
- Get the market reaction: "What did Nvidia's stock price and credit default swap spread do on the day this news broke, and what's the source for those numbers?"
Common mistakes to avoid
The mistake I see most is treating the $750 billion figure as one check Nvidia wrote. It's actually two deal types: a trade-volume estimate on existing chip and memory purchases, and a proposed credit guarantee that hadn't closed as of the July 27, 2026 report. Second is skipping the "circular financing" mechanism and just reacting to the big number. Deal size matters less than whether the cash brings in new demand or just loops back to the same few firms. Third is trusting one AI summary without asking it to name sources. In my testing, even a good summary can turn "in talks" into "confirmed" if you don't push back. Fourth is ignoring the credit market's move in favor of the stock price alone. The CDS spread here was actually the sharper signal, since bond buyers price default risk directly.
Nvidia's $750B deal package at a glance
| Deal | Parties | Reported value | Status as of July 27, 2026 |
|---|---|---|---|
| Chip, memory & data center pact | Nvidia + SK Group (SK Hynix) | $500B+ (trade volume, not a cash outlay) | Announced; 2+ GW of data centers planned |
| Ohio data center lease backstop | Nvidia + OpenAI + SoftBank/SB Energy | Up to $250B | In talks, not signed |
| GPU purchase financing | Nvidia + OpenAI | Up to $350B (separate from the $750B headline) | Under discussion |
| AI data center investment | Nvidia + Naver | $1B | Confirmed direct investment |
Tools that make this easier
The checking habit above works with any AI assistant, but the tool matters for how fast and how honestly it names sources. My AI tool ratings page and AI tool reviews hub cover which assistants show real citations versus which ones just sound sure of themselves. If agentic research is what you're testing — asking a tool to pull and cross-check several articles on its own — my how to use ChatGPT agent mode guide covers where that still needs a human checking the output. For the broader habit of not letting a fluent AI answer replace real checking, see my breakdown of why AI mania is eviscerating good decision-making — the same weak-point question applies directly here. If you want the wider context for why AI spending is under this much scrutiny, my pieces on corporate America’s AI spending pullback and Morgan Stanley’s AI valuation warning on SpaceX cover the wider skepticism this deal is landing into. And if these tools are new to you, my free AI tools roundup covers no-cost ways to start fact-checking headlines like this yourself.
My take
Nvidia's $750 billion figure isn't fake, but it does a lot of work by combining a trade-volume estimate with a proposed credit guarantee that wasn't signed at the time of this report. The SK Group side is the safer half — it's existing chip and memory business between two firms that already trade at scale. The OpenAI backstop is the half worth watching. A $250 billion guarantee tied to a data center that doesn't exist yet is exactly the setup that turns a supplier into a lender, and lenders lose money in different ways than suppliers do. The credit market's move — the sharpest CDS jump since Nvidia's swaps started trading — is a better early signal than the stock dip, because it prices default risk directly, not just mood.
Frequently Asked Questions
Is Nvidia's $750 billion deal actually confirmed?
Partly. The roughly $500 billion SK Group pact was announced and described by CEO Jensen Huang as real trade volume. The $250 billion OpenAI backstop, which makes up the rest of the $750 billion figure, was reported as still in talks and not signed as of July 27, 2026.
What does "circular AI financing" mean?
It's a deal setup where a company's cash or credit guarantee helps fund the same customer that then buys that company's products. Here, Nvidia may backstop OpenAI's data center lease while also selling OpenAI the GPUs that fill it. Critics warn it can inflate demand figures without adding real new capital to the system.
How did the market react to the news?
Nvidia shares fell nearly 5%, and five-year credit default swaps on Nvidia's debt jumped about 0.14 percentage points to around 0.82 percentage points. That's the sharpest single-day move since Nvidia's CDS began active trading in November 2025, per AI Weekly's same-day report.
How can I check a story like this myself instead of trusting a summary?
Trace every dollar figure back to the original outlet and date, ask your AI assistant to split "signed" from "in talks," and check the market move against a real financial data source rather than an AI guess. The steps above walk through this in about ten minutes.
Does this mean AI infrastructure spending is a bubble?
Not by itself. One deal structure drawing circular-financing pushback doesn't prove the wider AI buildout is unsound. But it's part of a bigger pattern this year, alongside enterprise AI spending pullbacks and analyst warnings about AI-linked valuations, that's worth tracking rather than dismissing.