The family influence surrounding NVIDIA (NVDA) is coming into sharper focus.
A little over a week ago, Jensen Huang's daughter, Madison Huang, held her wedding in Hawaii. Her husband, Nico Caprez, was once a professional skier, and Madison introduced him to NVIDIA (NVDA) two and a half years ago. He is no ordinary employee.
Now 35, Caprez quickly grew into one of NVIDIA's (NVDA) core executives and a primary point of contact for many customers and partners. Executives at several related companies say he manages NVIDIA's (NVDA) multibillion-dollar partnerships with emerging cloud providers and other data center-related companies, providing them with funding support and sales leads to help them continue building computing facilities equipped with NVIDIA (NVDA) chips.
Madison Huang, also 35, and Caprez form a powerful team. Some employees believe that under the leadership of 63-year-old CEO Jensen Huang, the two may become the backbone of NVIDIA's (NVDA) future leadership. One NVIDIA (NVDA) employee who works with Caprez said, "I think Jensen is laying the groundwork for Madison and Nico to take over the next phase of work." Another employee said NVIDIA's (NVDA) current operating model is far more like a family business than outsiders believe, which is very rare in the tech industry. (Jensen Huang's other child, 36-year-old Spencer, joined in 2022 as senior director of product management for robotics. NVIDIA (NVDA) also develops its own chip systems and supporting software for the robotics business.)
Jensen Huang wants to turn NVIDIA (NVDA) into something like a "central bank," providing funding support to customers and working with financial institutions to enable borrowing secured by expensive chips and supporting hardware, thereby sustaining the AI boom. Caprez is the core executor of this strategy.
In August, Caprez helped advance a plan jointly announced by NVIDIA (NVDA), Wall Street, and private equity firms such as Blackstone to raise $500 billion from investors to finance data center purchases of NVIDIA (NVDA) chips, though many details of the plan remain unsettled. As borrowing costs for data center projects rise and U.S. local political factors slow construction on some projects, Caprez's task has become increasingly critical.
For example, this summer Caprez led NVIDIA's (NVDA) push to take stakes in two leading U.S. companies that specialize in securing land and power resources for data centers. The investment was intended to lock in short-term power access for computing facilities and ensure they can continue running multiple generations of NVIDIA (NVDA) chips for years to come. The investment could also help avoid the chip glut risk feared by data center developers and SpaceX CEO Elon Musk and others: lagging grid construction combined with local opposition could leave newly purchased AI hardware without power and sitting idle.
Sikander Rashid, global head of AI infrastructure at Brookfield Asset Management, said, "Caprez's vision is to help NVIDIA (NVDA) control the AI infrastructure ecosystem." Brookfield owns several data center companies that use NVIDIA (NVDA) chips, and Rashid communicates with Caprez weekly. Vladimir Troy, a longtime NVIDIA (NVDA) senior vice president who has worked with Caprez for years, put it succinctly: "Nico matters." A cloud executive who once met Caprez called him "Jensen Huang's trusted strategist."
Rashid and other partners say Caprez is skilled at negotiation, especially at designing financial solutions such as credit backstops. Several people familiar with the matter said Caprez led NVIDIA's (NVDA) credit support agreement with Australian emerging cloud provider Firmus in June. Firmus used it to obtain loans to buy 170,000 NVIDIA (NVDA) Grace Blackwell and Vera Rubin chips, with a total purchase value potentially exceeding $15 billion. In exchange, NVIDIA (NVDA) can share in the project's future cloud business revenue once Firmus begins leasing out those chips early next year. The deal is part of a new financing initiative led by Caprez aimed at unlocking more capital for AI data centers, especially in regions where such credit business is still immature.
Caprez has kept a low profile and declined interviews. But that may be changing: over the past two months, he has repeatedly represented NVIDIA (NVDA) at industry forum roundtables, posting on LinkedIn that "AI has become core infrastructure" and that NVIDIA (NVDA) chips have "financing-friendly attributes"; his views were also quoted this summer in several press releases related to NVIDIA (NVDA) investments in data center companies.
If There's Something to Do, Find Nico
Before joining NVIDIA (NVDA), Caprez came from the Engadin Valley in eastern Switzerland and was a professional skier. In the early 2010s he represented Switzerland in alpine skiing events, competing in super-G, slalom, and downhill. After ending his skiing career in 2013, Caprez shifted from physical risk to decision-making risk, eventually becoming a management consultant at Boston Consulting Group in Zurich. Several years later, he enrolled in a master's program at London Business School, where he became classmates with Madison Huang, and the two graduated in 2021.
After business school, Madison Huang joined NVIDIA (NVDA), while Caprez went to Zurich-based PSquared Asset Management as an investment analyst; the small firm is said to manage about $750 million in assets. In early 2024, Caprez moved to Silicon Valley and joined NVIDIA (NVDA), where Madison Huang worked, as a corporate development manager. Soon after he joined, NVIDIA's (NVDA) market value became the highest of any company in the world.
Some NVIDIA (NVDA) partners say that even as a newcomer, he immediately took part in multiple core meetings with customers and partners. Jas Khehra, head of AI at private equity giant Blackstone, recalled meeting Caprez around that time to discuss expanding a landmark cooperation reached the previous autumn: a $2.3 billion credit line for cloud provider CoreWeave secured by NVIDIA (NVDA) graphics cards. That early deal was essentially the beginning of NVIDIA's (NVDA) push to make GPUs a class of valu able, easily traded financial assets, making it easier for chip buyers to borrow against them. Khehra said Caprez helped Blackstone and CoreWeave finalize a larger $7.5 billion credit agreement.
Kelly Littlepage, CEO of OneChronos, is building an auction marketplace for AI servers. He said he discussed with Caprez in early 2025 how the trading platform could assess the full-lifecycle value of server chips. Littlepage said Caprez quickly understood the logic of GPU residual value assessment, which is the basis for lenders to finance chip projects. "It was clear he was one of NVIDIA's (NVDA) top analysts," Littlepage said. "He is extremely good at cutting through clutter and getting to the core."
The market for GPU-backed debt is at least tens of billions of dollars, and NVIDIA (NVDA) wants it to expand further. GPUs are still at an early stage in becoming a widely recognized, tradable asset class like aircraft or shipping equipment, and more support is needed from debt investors and regulators, including approval of futures hedging tools related to GPU debt. Brookfield's Rashid said that in 2025, when Jensen Huang first met with him to discuss selling AI servers to governments and advancing the sovereign AI concept, he strongly praised Caprez. "Jensen basically said: if something needs to get done, find Nico."
Several partners said that as emerging cloud providers such as CoreWeave and Firmus become more important to NVIDIA (NVDA), Caprez's authority has continued to expand. Less than two years after joining, in January of this year, Caprez was promoted directly from director to vice president of global AI infrastructure growth, skipping the senior director level. According to people familiar with the organizational structure, Caprez currently reports to NVIDIA's (NVDA) new executive vice president of global business, Nick Parker, who in turn reports to Jensen Huang. One NVIDIA (NVDA) executive said that compared with his organizational rank, Caprez's close access to Jensen Huang is more important.
Months later, in March of this year, Madison Huang was also promoted, becoming senior director of product marketing for physical AI (mainly robotics). Her direct supervisor is a vice president, who reports further up to Jensen Huang. Brookfield's Rashid repeated Jensen Huang's words: "If something needs to get done, find Nico."
Over the past two years, Madison Huang's responsibilities have continued to expand, covering marketing for NVIDIA's (NVDA) 3D world models, robotics models, simulation models, and various open-source projects, in order to attract more application developers to use NVIDIA (NVDA) hardware. This summer, NVIDIA (NVDA) held a media event related to the Nemotron open-source AI model, and Madison Huang attended. NVIDIA (NVDA) spends billions of dollars a year developing the model to compete with closed-source large models from Anthropic and OpenAI. For the same strategic reasons, NVIDIA (NVDA) recently acquired open-source model hosting platform Hugging Face for about $20 billion and obtained technology licensing from open-source AI company Poolside.
Madison Huang also often accompanies her father to international conferences, helping Jensen Huang finalize cooperation with supply chain and manufacturing partners.
Connecting Resources Across Parties
Caprez also travels around the world, connecting with customers, investors, and partners, sometimes attending meetings together with Jensen Huang and Madison Huang. Several emerging cloud executives say Caprez is their primary contact at NVIDIA (NVDA), communicating weekly or even more frequently. He mainly helps emerging cloud providers connect with customers and financing channels. One emerging cloud executive said that earlier this year Caprez called to say an AI developer urgently needed server chips and that the cloud provider could quickly take the order. The cloud provider ultimately completed delivery, and the AI developer became one of its largest customers. Another emerging cloud executive said Caprez pushes potential customer leads to executives through WhatsApp groups. Other executives said Caprez helps companies find scarce available land and power resources.
Currently, Caprez co-leads NVIDIA's (NVDA) cloud partner program. The program includes dozens of cloud providers, the vast majority of them emerging cloud companies. These providers build data centers according to NVIDIA (NVDA) standards and use NVIDIA (NVDA) hardware. The other co-leader of the program is Raj Mithal, a senior vice president who has been with the company for more than 20 years. The program existed before Caprez joined, and traditional cloud providers such as Amazon and Google are not included. One NVIDIA (NVDA) employee tracking the cloud partner program said that in the future such emerging cloud providers could contribute as much as half of NVIDIA's (NVDA) total revenue, helping it reduce dependence on traditional cloud providers such as Amazon and Google. Although traditional cloud providers buy large amounts of NVIDIA (NVDA) chips, they also develop their own AI chips and purchase hardware from NVIDIA's (NVDA) competitors.
Since the rise of the AI wave, supporting emerging cloud providers has been a key strategy for Jensen Huang, though such companies currently account for only in the low teens percentage of NVIDIA's (NVDA) total revenue. In August, Caprez helped formulate NVIDIA's (NVDA) financing plan with Wall Street and private equity institutions to raise $500 billion for purchases of NVIDIA (NVDA) AI hardware. Revenue generated by AI businesses is currently far from enough to cover the huge sums already invested in computing facilities, and whether this financing plan succeeds will determine how long the AI data center construction boom can last. When announcing the plan, NVIDIA (NVDA) said it could provide up to 25% credit support for the data center projects of partner projects, but specific agreements have not yet been finalized.
Khehra said he had dinner with Caprez last week to finalize more details of the cooperation. Khehra said Caprez never speaks carelessly. But when Jensen Huang assigns follow-up tasks after meetings, it is obvious that he entrusts a large amount of important work to Caprez.
Jensen Huang's Trust
Executives who have negotiated with Caprez say that although he is tough in negotiations, he usually remains rational in meetings and communicates constructively. Rashid said, "Sometimes when the negotiation atmosphere gets tense, he suddenly makes a joke to ease the mood."
Although emerging cloud providers rely heavily on the chip design company, tensions still exist between NVIDIA (NVDA) and them. One emerging cloud executive said NVIDIA (NVDA) in effect largely dictates how they operate. For example, if NVIDIA (NVDA) chip deliveries are delayed, it discourages cloud providers from buying hardware from other vendors. (NVIDIA (NVDA) disagrees with this account. Dion Harris, senior director of high-performance computing at NVIDIA (NVDA), said the company allows customers to freely combine hardware, "use whatever you want.")
Jensen Huang often points out to emerging cloud providers how dependent they are on NVIDIA (NVDA). On the Dwarkesh podcast earlier this year, he said, "For cloud providers, if we hadn't supported CoreWeave in the beginning, these emerging cloud and AI cloud companies wouldn't exist at all." Currently only a few emerging cloud providers have annual revenue above $200 million, and none has become profitable. Jensen Huang most likely wants to avoid a situation in which large emerging cloud companies acquire weaker peers. That means Caprez carries a heavy burden and needs to push these companies to grow steadily.
NVIDIA's (NVDA) AI computing cooperation program is one such measure, for example reaching a credit guarantee with Firmus in exchange for future revenue sharing. But Caprez and other executives are reevaluating the program's structure. Two people familiar with the matter said larger, more mature emerging cloud providers such as Nebius have refused to join, unwilling to become more financially dependent on the chipmaker or to have their margins squeezed by revenue sharing. Other sources said NVIDIA (NVDA) is also worried that cloud providers willing to join the program could become overly financially dependent on NVIDIA (NVDA). Two people familiar with the matter said NVIDIA (NVDA) is adjusting investment term sheets and contracts to avoid these risks.
At the same time, data center lenders are reassessing their risk tolerance and becoming more cautious about funding new projects. Some investors are reassessing the valuations of GPU cloud providers and supporting infrastructure companies, which could affect the ability of some NVIDIA (NVDA) customers and partners to go public, raise funds, and obtain low-cost debt financing. Jensen Huang is also closely watching the market's reaction to NVIDIA's (NVDA) increasing use of its balance sheet to support the AI industry's expansion. People familiar with the matter said he often asks subordinates about changes in the spread on NVIDIA's (NVDA) credit default swaps. Over the past two months, that spread has roughly doubled, indicating investors see a slightly higher risk of NVIDIA (NVDA) defaulting on its debt. Objectively speaking, of course, the risk remains very low, and Moody's credit rating for NVIDIA (NVDA) is only one notch below the highest level.
Jensen Huang's intense, around-the-clock work style seems to have influenced Caprez as well. Business partners say Caprez is one of the hardest-working people they have ever met, always online and quick to respond no matter the time zone. His wedding to Madison Huang coincided with NVIDIA's (NVDA) quarterly holiday, when all company employees had two days off on September 24 and 25. One business partner said that even while vacationing in Hawaii, Caprez continued working and returned to his post on Monday, September 28. Dinner participants said that the next evening, the two attended NVIDIA's (NVDA) investor dinner in New York wearing wedding rings.
Comments