Nvidia introduced the resignation of Jensen Huang after practically per week of rolling revelations alleging romantic, extramarital relationships with multiple women, together with one imagined to be an agent of the Chinese government.
The departure of Nvidia’s showman of a CEO will inevitably shed extra mild on his closest collaborators on the chip firm. One government poised to exert affect within the post-Huang period is Ian Buck, Nvidia’s vp of hyperscale and high-performance computing.
The 48-year-old’s contribution to Nvidia is hard-coded into the corporate’s historical past. He is been working alongside Huang since 2004, when he began constructing the corporate’s breakthrough software program, referred to as Compute Unified Gadget Structure.
Again then, the crew had two individuals and two libraries. Right this moment, CUDA helps greater than 900 libraries and synthetic intelligence fashions. Every library corresponds to an trade utilizing Nvidia know-how.
Particularly in mild of Huang’s departure, Buck’s affect is simply starting. CUDA is the platform from which Nvidia jumped to, at one level, 90% market share in AI computing. CUDA is how the corporate defends its moat.
It is the heft of Buck’s contribution that has led trade analysts and onlookers to repeatedly cite Buck in discussions of a doable successor to Huang, even lengthy earlier than one was wanted.
One structure to rule all of them
Dion Harris, Nvidia’s senior director of high-performance computing and AI manufacturing unit options, generally forgets that he is within the room with the Dr. Ian Buck. Then it hits him that his boss, and good friend, is a computing legend.
Since Buck’s undergrad days at Princeton within the late Nineties, he had been targeted on graphics — a very punishing subject inside pc science with no apparent connection to AI on the time.
“Laptop graphics was such a dweebie subject,” mentioned Stephen Witt, the creator of “The Thinking Machine,” which particulars Nvidia’s rise from obscurity to the world’s most dear firm.
“There was a stigma to working in pc graphics — you had been perhaps some type of man-child if this was your focus,” Witt mentioned.
Whereas pursuing his Ph.D. at Stanford, Buck linked a number of graphics processing items to push them to their limits. He had interned at Nvidia earlier than pursuing his Ph.D., so he was acquainted with the GPU. Initially, he used it for graphics like everybody else.
Buck has mentioned that he and his cohort would use the chips to play video video games resembling “Quake” and “Doom,” however finally, he began asking himself what else his gaming setup may do.
He turned fixated on proving that you possibly can use GPUs for something and all the things. He acquired funding from Nvidia and the Protection Superior Analysis Tasks Company, amongst others, to develop instruments to show a GPU right into a general-purpose supercomputing machine. When the corporate noticed Brook—Buck’s try at a programming language that utilized the facility of GPUs past graphics, Nvidia employed him.
He wasn’t alone. John Nickolls, a {hardware} professional after which director of structure for GPU computing, was additionally instrumental in constructing CUDA. Buck might have been perpetually paired with Nickolls if the latter had not died from most cancers in 2011.
“Each Nickolls and Buck had this obsession with making computer systems go quicker in the best way {that a} Method 1 mechanic would have an obsession with making the race automotive go quicker,” Witt informed BI. (The creator mentioned Huang expressed frustration that Nickolls does not get the popularity he deserves since his passing.)
Buck, Nickolls, and a small crew of consultants constructed a framework that allowed builders to make use of an present coding language, C, to harness the GPU’s skill to run immense calculations concurrently slightly than one after the other and apply it to any subject. The consequence was CUDA, a automobile to convey parallel computing to the lots.
The rise of CUDA as an important factor on the planet of AI wasn’t inevitable. Huang insisted on making each chip suitable with the software program, despite the fact that hardly anybody was utilizing it, regardless of it being free. In actual fact, Nvidia misplaced hundreds of thousands of {dollars} for greater than a decade due to CUDA.
The remainder is lore. When ChatGPT launched, Nvidia was already powering the AI computing revolution that’s now the main focus of trillions in infrastructure spending, a lot of which finally goes to Nvidia.
King of the nerds
Buck’s smarts do have limits. He joked to an keen viewers at GTC in 2025 that quantum chromodynamics, a subject of particle physics, simply will not stick in his mind. However he thrives in Nvidia’s notoriously rigorous surroundings.
Underneath Huang, the Santa Clara, California, firm has maintained an intense tradition that eschews one-on-one conferences and makes missteps and disagreements public. It’d sound terrifying, however for these with the brains to maintain up, the directness and the rigor are perfect.
For this report, Enterprise Insider spoke with 4 individuals who have both labored immediately with Buck at Stanford, Nvidia, or each. Those that know Buck personally describe him as gregarious and easygoing however able to depth when targets are on the road. He is targeted on outcomes slightly than concept.
In his public remarks, at panels and in interviews on behalf of Nvidia, Buck volleys from fast, technical strains of thought to slower, easy descriptions in layman’s phrases.
At a GTC press convention, he detailed the most recent improvement in convolutional neural networks after which described proteins as “difficult 3D squigglies in your physique.” He describes the tiny, delicate interconnects between elements of an Nvidia chipset just like the Geek Squad explaining the again of a TV from reminiscence — it is all in his head.
Harris mentioned storytelling skill is especially essential within the higher ranks at Nvidia. For the reason that firm primarily had a promising know-how ready for a marketplace for years, Huang noticed being early as a enterprise technique. He branded it “going after zero billion-dollar markets.” The potential of AI, “AI factories,” and the infrastructure spending that goes with them is a narrative Nvidia cannot cease telling.
Buck’s shilling abilities have improved through the years. However even in 15-year-old footage, he is most animated when explaining the inside workings of Nvidia’s know-how.
“Quite a lot of builders are superb, however they are saying, ‘Go away me alone. I will write my code within the mountains someplace,” Paul Bloch, president of Nvidia partner DDN, informed BI. Nvidia’s leaders aren’t like that, he mentioned. A lot of Nvidia’s higher echelon might have the talents to match the reclusive set, however they do not select between showmanship and code.
Ian Buck’s subsequent act
Ian Buck’s work at Nvidia started with a easy mandate: make the GPU work for each trade. That mission could be very practically achieved. There are lots of of CUDA libraries concentrating on industries from climate forecasting to medical imaging.
“The libraries are actually there to attach the dots so that each enterprise does not need to study CUDA,” Harris mentioned.
CUDA attracts its power from hundreds of thousands of builders, amassed over a long time, who consistently innovate and enhance on the platform. To this point, nobody has caught Nvidia’s heels, however the competitors is coming quicker than ever.
Whilst Buck spoke at GTC in 2025, builders internationally had been attempting to interrupt by way of CUDA’s dominance. The primary night time of the convention, a forged of rivals convened by TensorWave, an AI cloud firm solely utilizing chips from Nvidia’s solely US rival, AMD, held an occasion entitled “Past CUDA.”
TensorWave cofounder Jeff Tatarchuck mentioned it included greater than “24 presenters speaking about what they’re doing to beat the CUDA moat.” New startups looking to crack CUDA are nonetheless fundraising, too.
AMD, which additionally introduced on the occasion, is making an express effort to make amends for the software side of AI computing, however trade analysts say they are not there but. And regardless of the rise of coding brokers able to writing CUDA kernels, demand for human CUDA engineers continues to develop. There have been extra jobs requiring CUDA abilities posted in 2026 than in 2025, in accordance with the labor analytics platform Lightcast.
Harris informed BI that Buck’s crew spends a variety of time talking with researchers to remain on prime. That is at all times been true, however the nature of the work has modified. A decade in the past, Buck was convincing researchers to use accelerated computing to their issues; now the tables have turned.
“Probably the most difficult elements of my job typically is to attempt to predict the longer term, however AI is at all times stunning us,” Buck mentioned at a Financial institution of America convention in 2025. Understanding what the neatest minds in AI want from Nvidia is paramount.
Many noticed DeepSeek, the corporate that spooked markets final 12 months with its ostensibly low cost reasoning AI mannequin, as a menace to Nvidia because it achieved aggressive outcomes with much less compute, not realizing that the crew went deep into CUDA to squeeze out these positive aspects. Buck described the Chinese language crew as “top-of-the-line CUDA builders on the market.”
AI is coming into a brand new section as extra corporations commercialize their tools. Even with Nvidia’s monumental head begin, partly constructed by Buck, the tempo is intense.
Expertise can be an enormous a part of how Nvidia goes to attempt to keep its dominance, and one other place the place, Witt mentioned, Buck has worth past his technical abilities. He is not precisely a family title, even at Stanford. However for a sure sort of developer, those who can play in Buck’s extraordinarily complicated sandbox, he’s a draw.
“Everybody’s attempting to rent these guys, particularly after DeepSeek,” mentioned Witt. “This was not an attractive area in pc science for a very long time. Now it’s white sizzling.” Huang’s departure will make that facet of his function much more essential as a number of trade analysts have expressed concern that Nvidians could also be extra prone to depart the corporate now. Traditionally, Nvidia has had an especially low fee of staff turnover.
Nvidia declined to make Ian Buck obtainable for an interview with Enterprise Insider and declined to touch upon this report.
Who’s Nvidia’s subsequent CEO?
Buck is greater than a decade youthful than Huang and has been a part of the Huang succession dialogue lengthy earlier than the present scandal.
Lip-Bu Tan, a semiconductor trade legend who turned Intel’s CEO in 2025, informed BI that Buck was one in all a handful of true collaborators for Huang, who has greater than 60 direct stories.
“Jensen has three right-hand males,” Tan informed BI earlier than he took over at Intel. Buck is one. Vice President of GPU Engineering Jonah Alben is one other. And CFO Colette Kress, although clearly not a person, is the third, Tan mentioned. Kress was appointed interim CEO by Nvidia’s board Tuesday.
“Prediction markets have recognized Rene Haas, former Nvidia VP and present CEO of Arm, as a possible exterior substitute, along with inside candidates Colette Kress and Ian Buck, head of Nvidia’s AI efforts,” mentioned Brian Colello, Senior Fairness Analyst at Morningstar, in an electronic mail Thursday.
“I do not suppose Ian spends a variety of time doing enterprise technique. He is extra just like the world’s greatest mechanic,” Witt mentioned in 2025.
