Frankfurt. Jensen Huang is at the peak of his power. Simply because the CEO has made his chip manufacturer Nvidia the most valuable company in the world. The market values the company, which has a monopoly-like key position in the sale of AI chips, at $5.4 trillion. The sum is higher than the market valuations of the major US banks Bank of America, Goldman Sachs and Citigroup combined.
And yet Jensen Huang's power extends further than this number suggests. Huang has not only built the most valuable company in the world, but also a unique network of connections in the tech and financial sectors. And he, Jensen Huang, sits like the spider in the middle of this web.
On the one hand, there are the hyperscalers such as Microsoft, Amazon, Meta and Google, which account for around half of Nvidia's sales. There are the data center operators like Coreweave, the large AI providers Anthropic and OpenAI, as well as dozens of start-ups in which Nvidia has a stake. And on the other hand, there are the large financial institutions on Wall Street, such as Blackrock, Blackstone and Apollo, who together want to invest $500 billion in the AI industry and have Nvidia secure part of their investments.
Other attributions are not quite as euphoric. Nvidia has long been in an unusually powerful position: the company is the central player in the AI boom and is in the process of becoming closely linked to Wall Street. It is the most important chip manufacturer and central financier of its own sales market. The accusation: This creates a kind of circular economy. The AI companies' business is doing well because Nvidia also finances risky things. And Nvidia's business is doing well because its customers' business is doing well.
“It is precisely this dual role that makes Nvidia systemically relevant,” says capital market expert Mohamed El-Erian, pointing to two reasons. Does this mean that Nvidia is “Too Big To Fail”, too big to fail because the effects would be devastating? “From my point of view, that is the case,” says El-Erian.
“Too Big To Fail”. The term dominated public discussion around the 2008 financial crisis. The concept that a bank is simply too big, too important and too connected to fail caused a rethink among regulators worldwide. Since then, major players like JP Morgan Chase and Bank of America have been even more strictly regulated than the rest of the industry. There are even “Living Wills” that outline the best way to handle them in an emergency.
How fragile Nvidia's business model is became clear at the weekend: The heads of the large AI companies Anthropic and OpenAI, Dario Amodei and Sam Altman, raised concerns about a loss of control over the AI. Faced with growing risks, they urged slowing down the development of particularly advanced AI models. On Monday, this pushed down the stock prices of the tech group Samsung and the semiconductor manufacturer SK Hynix in Asia by four and seven percent. Softbank, an investor in OpenAI, lost 11 percent.
OpenAI and Anthropic “have a significant impact on the entire technology industry,” Mark Mahaney, an analyst at Evercore ISI, told Bloomberg. “If both companies were to significantly reduce their spending on research and development, their hiring and their investments, this could have a significant impact on the financial markets.”
This changes the requirements for regulatory authorities. The Financial Stability Board (FSB), an international organization that makes recommendations to governments, articulated the risks weeks ago. In a letter to the G20 finance ministers, FSB chief Andrew Bailey wrote that he was concerned about the mix of rising debt levels, a higher concentration of tech stocks in the stock market and the "increasing cross-investment between artificial intelligence companies and hyperscalers". “This could exacerbate a future market correction,” said Bailey, who is also governor of the Bank of England.
El-Erian emphasizes that this is not a likely scenario, but rather a so-called tail risk. This refers to risks that have a very low probability of occurring. But if they occur, the damage is extremely high. “In a world where AI has become such a fundamental innovation, we need to increasingly understand this risk,” says El-Erian.
Nvidia has repeatedly broken records throughout its history. Sales almost doubled in the past fiscal year to $216 billion. Analysts believe it is possible that Nvidia could reach annual sales of $1 trillion in the 2029 fiscal year, becoming the first company ever to do so.
Bank of America analyst Vivek Arya highlights the large amounts of cash Nvidia will generate. “In a year, Nvidia will generate a billion dollars of free cash flow every single business day,” he said on CNBC at the end of August. “No other company on the planet has ever done this.” Nvidia also has around $99 billion in cash and liquid securities.
The group is driving its expansion from a position of exceptional strength. And this is precisely what is dividing the financial world: supporters believe that Nvidia could use its resources even more decisively to accelerate the expansion of the AI infrastructure. “We are still at the beginning of the AI boom,” says Dan Morgan, portfolio manager at Synovus.
Frank Lee, Nvidia analyst at HSBC, is among the most optimistic in the industry. Lee primarily points to the increasing diversification of the business beyond the large cloud providers, which could drive profits in the future. Finally, the group is expanding its business with corporate customers, states and AI laboratories.
Others are concerned about how Huang is using this financial power. According to El-Erian, it is the second argument why Nvidia has become systemically relevant. The group is increasingly using its balance sheet to support customers in purchasing their own chips. The group invests in start-ups, helps finance data centers and guarantees minimum sales and the residual value of its hardware. Nvidia wants to provide guarantees of up to $105 billion for a planned data center in Ohio that will use OpenAI. In return, the project will use 1.5 million Nvidia processors.
Morgan Stanley analysts Lindsay Tyler and Nishant Satyam call this strategy “balance sheet as a service.” The term is an allusion to the “Software as a Service” business model, in which customers do not buy software and operate it themselves, but rather use it continuously as a service and pay for it regularly.
Analysts are divided. On the one hand, in an analysis from the end of August, they confirmed that Nvidia's balance sheet was extremely robust and had a low level of debt. But the fact that the chip manufacturer is “turning its balance sheet strength into a strategic financing instrument for artificial intelligence also creates new risks,” they made clear. They point out that “conventional debt ratios” do not reflect the entirety of credit risks. Because Nvidia's support of the AI ecosystem would only be partially visible.
Circular deals and risks that grow in secret remind El-Erian of the so-called collateralized debt obligations, or CDOs for short. They became an important accelerator of the financial crisis. “In the years before 2008, massive circularity emerged within the financial system,” says El-Erian. Loans or mortgage-related securities were initially bundled into CDOs and divided into different risk tranches. Tranches of several CDOs were then bundled again into new CDOs - and these again into CDOs. A bundle of dependencies arose that hardly anyone understood.
A new $500 billion collaboration with Wall Street is intended to reduce the risk of circular deals, Huang explained in a blog post in August. The chip manufacturer should no longer decide which projects are financed. But financial institutions like Blackrock, Blackstone, Goldman Sachs and Apollo. These would collect money from customers and set up independent financing.
From Huang's perspective, this is a good deal, says a tech investor. Ultimately, the Wall Street houses would make the investment decisions and provide Huang with a steady influx of chip buyers. This project will not work without the financial power of Nvidia. In the letter of intent, the group promised guarantees totaling $125 billion. The fundamental problem does not become smaller, but larger: Even if Nvidia reduces circular deals, Nvidia would continue to intertwine itself with the financial world and increase its own system relevance. It's no longer about the Nvidia company alone, but about the Nvidia system.
"What about Nvidia's customers? They have to be doing well for Nvidia to be doing well," El-Erian points out. "We have not only created a company that is in danger of becoming too big to fail. We are currently developing an entire industry that is systemically important."
Capital market expert Ed Yardeni also considers OpenAI to be “Too Big To Fail,” as he explained in an interview with Handelsblatt in August. “There is so much cross-funding and business relationships between OpenAI and other companies that OpenAI could become the weakest link in the AI chain if the company does not generate enough revenue to meet its obligations.” Nvidia announced a strategic partnership with OpenAI last year.
