Head over to our on-demand library to view sessions from VB Transform 2023. Register Here
1.25 million square feet across two massive steel and glass structures. Hundreds of conference rooms named after Star Trek places, alien races and starships, as well as astronomical objects — planets, constellations and galaxies. Acres of greenery and elevated “birds nests” where people can work and meet. A bar called ‘Shannon’s’ with a panoramic view and plenty of table space for board games.
This is the nearly $1 billion headquarters of Nvidia in Santa Clara, California — located on a patch of prime Silicon Valley land where the technology company has spent the past three decades growing from a hardware provider for video game acceleration to a full-stack hardware and software company currently powering the generative AI revolution.
But amid the lavish architecture and the fun perks, it can be difficult to discern the hard work and intense pressure that supported Nvidia’s entrance into the $1 trillion valuation club last month, alongside fellow tech giants Alphabet, Amazon, Apple and Microsoft. As I walked the equivalent of a winding Yellow Brick Road to the main entrance, with a view of the towering curves and lines of the two buildings rising over the San Tomas Expressway, I wondered whether I’d get a peek behind the PR curtain at Nvidia’s true nature.
“Where’s Jensen?” I asked Anna Kiachian, the Nvidia PR manager who had arranged my campus visit.
VB Transform 2023 On-Demand
Did you miss a session from VB Transform 2023? Register to access the on-demand library for all of our featured sessions.
The truth is, I hadn’t expected to get an audience with Nvidia CEO and co-founder Jensen Huang. For all I knew, Huang had been relaxing in the Maldives ever since Nvidia became a Wall Street darling this spring in the wake of the generative AI boom — ten years after helping to power the deep learning “revolution” of a decade ago. Industry analysts estimate that Nvidia’s dominance extends to over 80% of the graphical processing unit (GPU) market, which is a must-have for every company running AI models, from OpenAI down to the smallest startup.
Still, I figured perhaps a random sighting of Huang’s ubiquitous black leather jacket — from afar — was possible.
“I’m not sure,” Kiachian replied with a conspiratorial smile as we strolled through an immense atrium with hundreds of triangular-shaped skylights gleaming overhead. But she emphasized that Jensen came into the office every day when he was in town: “So you never know!”
Luckily, the sight lines were excellent for Jensen-watching, especially since the headquarters’ two buildings — Endeavor, which opened in 2017 and Voyager, which debuted in 2022 — were hardly filled to capacity. There were obviously plenty of Nvidia employees still working at home or on summer vacation, leaving plenty of white space to spot one black leather jacket.
But if any space could lure people back to the office, this is it: Endeavor and Voyager cost a whopping $920 million to build — a small price to pay, apparently, to meet Huang’s vision of giving every employee a view while boosting collaboration and random connections. Designed by architecture firm Gensler, which built the largest skyscraper in China, they are anything but a claustrophobic maze of hallways, cubicles and data centers. Instead, I felt like I could spot Jensen from a half-mile away across the sprawling, soaring, angular expanse.
There was no time for approaching him, however. I was on a strict schedule of meetings, beginning with a campus tour led by Jack Dahlgren, who heads up Nvidia Omniverse developer relations but also served as project and design manager for the buildings. As I racked up steps on my Fitbit, Dahlgren interjected with fun facts, like how people kept getting lost searching for conference rooms in Endeavor because their order was understood only by the most devoted sci-fi nerds and there was little signage (Dahlgren said Jensen felt a large map would clutter the landscape). The newer Voyager, he explained, has them in alphabetical order.
The triangular design of the two buildings, he continued, are repeated in the triangles throughout the roof and floor plans, which were computationally designed with an algorithm. “Triangles represent the building blocks of all 3D graphics,” he said.
There are also hidden metaphors: For example, Endeavor’s core can be seen as a tree trunk, with branches spread out from the center. It’s very noisy and busy in the middle, while around the outside are relaxed and quiet common spaces. Voyager, on the other hand, with its many noisy, whirring labs in the center, called “The Mountain,” with public spaces spread over the top (with ‘Shannon’s’ bar at the pinnacle), featuring views facing Silicon Valley and the mountains beyond it.
Jensen Huang’s presence looms large at Nvidia
Huang, a native of Taiwan whose family emigrated to the U.S. when he was just four years old, co-founded Nvidia in 1993 with the goal of building graphics chips for accelerated computing — first for gaming, and then, it turned out, for AI. These days, Nvidia is as much, if not more, of a software company as a hardware company, with a full-stack ecosystem that began nearly two decades ago by building CUDA [compute unified device architecture], which put general purpose acceleration into the hands of millions of developers. Today, experts see little chance of anyone catching Nvidia when it comes to AI compute dominance, with the largest companies with the deepest pockets battling for access to Nvidia latest H100 GPUs.
Whether he is in the office or not, it’s clear that Huang’s presence looms large around every corner. He seems to serve as founder, fatherly figure and as a sort of revered Star Trek captain. The phrase “Jensen says” is commonly uttered, whether it is quotes from his many inspirational speeches around strategy and culture, or his emphasis on a “first principles” approach — kind of a mission statement for each project.
“Jensen says the mission is the boss,” said Dalhgren. For example, the mission was to build the headquarters, he explained. But no one was the boss of the project. Groups came together, he explained, and the project itself was the boss.
That seemed a bit hard to believe — Huang certainly seemed like the boss. In a previous piece I wrote about Nvidia, an analyst told me that Huang is seen as demanding. There were graphics engineers at other tech companies that were ‘renegades’ from Nvidia, he said — who left because they couldn’t handle the pressure.
Still, Nvidia prides itself on its lack of hierarchy — other than Huang at the helm. One of the most important in the “everyone else besides Jensen” camp is Chris Malachowsky, one of Huang’s two co-founders who now serves as SVP for engineering and operations. In one of those “random connections” moments, Kiachian gave an excited little leap when she realized he was walking towards us, and gave me a warm introduction.
When I asked him what he thought of the new campus, Malachowsky said it “boggled his imagination” and went on to quote one of Huang’s oft-repeated themes: “I know it seems absurd, but we think of ourselves as a startup,” he said. “Jensen used to say we were always 30 days from going out of business, so to actually be confronted with what not going out of business means is flattering and nice, I can honestly just say ‘wow.’”
Nvidia’s hardworking AI chips
Malachowsky’s mellow vibe did not extend, however, to the windowless lab that concluded my campus tour — a cold, noisy, claustrophobic space where Nvidia’s AI chips were being tested.
Dahlgren pointed out that the basic principles for the chip designs were also used in the building’s designs. “Before we send the chip off to the fab to get built, we do pre-silicon emulation — we test it with a supercomputer which emulates how the silicon and the wires will work when it’s put together,” he said. “We did the same thing when we built the model of the building — we simulated how light would flow, we measured that, we came to an understanding of how it would perform before we built it.”
I thought of that when I saw examples of the chips in a museum-like demo room, from a $69 graphics card to the $40,000 H100 cluster — 1000 of which built OpenAI’s ChatGPT. The glossy, glimmering metal squares, rectangles and boxes were truly beautiful, disguising the massive workloads they take on to power today’s LLMs.
They reminded me of Nvidia HQ’s shimmering skylights, uplifting views and bold, geometric design — which belie the late nights, drudgery and frustration that, I felt, must also be part of the company’s success algorithm.
Beneath Nvidia’s glossy surface
The Nvidia cafeteria was filled with hungry staffers by early afternoon. Kialchia pointed out that Jensen had decided to close the Endeavor cafeteria so everyone had to come to the one in Voyager — creating even more random connections for employees. So there were actual lines at the salad bar.
Kialchia also pointed to a sign which said today was Popcorn Thursday, which she noted with a laugh was a surprisingly big deal at Nvidia. Highly-paid developers, apparently, can still love a freshly-popped bag of popcorn.
As I munched my popcorn, I couldn’t help but wonder if that’s where I’d have to look to see beneath the surface of Nvidia: At the people. No matter how beautiful the campus, how positive the culture, and how passionate the founder, doesn’t it still take people who work hard and set high standards and don’t always get along to get ahead?
But that was hard to suss out on my tour: During my walk around Endeavor and Voyager, for example, Kiachian had decreed that what I thought was a funny anecdote from Dahlgren was off the record. It was something totally silly, just a memory of how Nvidia didn’t always have such a cushy campus. It was nixed, I suppose, because it didn’t fit Nvidia’s happy-go-lucky narrative. Dahlgren, for his part, brushed it off, saying that everyone at Nvidia seemed to have a sense of humor, even if it occasionally veered towards the dark side.
“Some of it is dark humor, because work is hard,” he said. “But it’s rewarding.”
As I ended my day at Nvidia, I realized that I never got my Jensen sighting. I wasn’t disappointed — I thoroughly enjoyed my brief landing on Planet Nvidia. But I wish I could have gotten more of a sense of the blood, sweat and tears that is undoubtedly required to continue designing and building AI’s most famous picks and shovels. Still, the company’s dreamy culture of inspiration, illuminated by Endeavor and Voyager’s dramatic architecture, is hard to resist. I’d show up in person, for sure. At least for Popcorn Thursday.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.