Right places, right times: Jensen Huang steers Nvidia to trillion-dollar status

By PENG Xin

Nvidia, a name once familiar only to nerdy video game players has, through a quirk in technological development, become one of the hottest companies on the planet.

The value of Nvidia has gone from a very respectable US$150 billion three years ago to US$1 trillion (7.12 trillion yuan) today. In perspective, Intel, the name on a billion laptop stickers, is currently worth about US$125 billion.

Revolutionary history

The graphics-processing unit (GPU) market was already packed in 1999 when Nvidia released the revolutionary GeForce card, far ahead in graphics and lighting than other manufacturers.

In 2004, the company began work on CUDA (Compute Unified Device Architecture), a language similar to C++, allowing programmers to access the GPU directly. Courses in CUDA can be found in more than 200 universities worldwide.

In 2008, Nvidia introduced the Tegra system-on-a-chip (SoC) primarily for carmakers. However, in 2017, Nintendo put the Tegra in its Switch console.

In 2016, crypto miners discovered that GPUs were especially efficient at mining cryptocurrencies. Large systems were built, consuming the entire world's supply of GPU cards. The shortage of GPU cards only worsened when Covid-19 caused supply problems.

In 2019, the company bought networking outfit Mellanox, specializing in data processing units (DPUs). AI requires accessing massive data sets. The idea is to take over the processing of the networking data, while the CPU (central processing unit) continues its main job of processing actual data.

At the conclusion of COMPUTEX on June 2, Huang attended a press conference alongside Rick Tsai, CEO of MediaTek, to announce a collaboration between the two companies in automotive cockpits.

Nvidia is now for AI what TSMC is to chip wafers. According to Trendforce, as much as three-quarters of the 1.2 million AI servers shipped last year were equipped with Nvidia GPUs.

Path of the trailblazer

The company was founded in 1993 by Jen-Hsun "Jensen" Huang, Curtis Priem and Chris Malachowsky and is headquartered in Santa Clara, Calif.

Born in Taiwan in 1963, Huang’s father was a chemical engineer and his mother a teacher. While still young, Jensen moved to Thailand with his family. At nine, he was sent to the US to live with his aunt and uncle.

In 1984, he obtained a degree in electrical engineering from Oregon State University. He later earned a master’s degree from Stanford University, before becoming a successful engineer specializing in microprocessor and graphics processing.

In the 1980s, PCs were gradually transitioning from offices to homes, but they were still very much “computers” with only the most rudimentary, geeky entertainment. Determined to change that, Jensen formed a company in GPU development, naming it Nvidia, which stands for “envy” plus “idea.”

In an interview, Huang recollected that when he called experts in the GPU market, most advised him against entering the chaotic free-for-all, with at least 20 to 30 similar companies already in the fray. But Huang was more determined than ever.

Flirting with disaster

Huang went to Sand Hill Road in Silicon Valley where entrepreneurs go every day seeking funding and eventually secured US$2.2 million, the beginning of Nvidia’s chip story.

During the next four years, Nvidia released the NV1 and NV2. Unfortunately, NV1 backed the wrong technological horse. Huang was forced to go from over 100 employees to just 30.

Thanks to Sega’s offer of US$7 million for the NV2 that deviated from the current leading technology, Nvidia narrowly survived.

“By recognizing our mistakes, cutting our losses in a timely manner, and humbly seeking help, we managed to avoid bankruptcy,” Huang said.

Nvidia didn’t gain proper going until its third-generation RIVA128 appeared in 1997, the first high-performance 128-bit GPU to support Microsoft’s Direct3D graphics interface and thus garnered an endorsement from Microsoft.

That year, Huang received a long-awaited phone call from Morris Chang, the founder of TSMC. TSMC agreed to be Nvidia’s chip foundry in a win-win collaboration.

Make or break

Nvidia entrusted manufacturing to TSMC to concentrate solely on chip design. With the success of Nvidia and other clients, TSMC’s wafer foundry model was quickly accepted and GPUs became familiar to everyone in the industry and became an integral part of almost every computer.

At the end of 2000, Nvidia acquired rival 3dfx for US$110 million, emerging as the ultimate victor in the GPU war.

Nvidia immersed itself in graphics processing. GPUs provided sharper displays and stunning game effects while dealing with complex 3D computation. The graphics market meant constant demand and constant innovation. Nvidia ruled wherever displays and pixels twinkled.

Accelerated computing

At the end of 2006, CUDA was released, opening up Nvidia’s power for all kinds of purposes beyond graphics.

Huang told analysts that Nvidia would pioneer a new field called accelerated computing. All of Nvidia’s GPUs supported CUDA, which meant that anyone with a laptop equipped could develop software.

Nvidia began to establish a foundation of software, besides selling chips, and unleash the full potential of GPUs through independent innovators everywhere.

However, accelerated computing in high-performance computing like climate modeling, was a non-profitable niche market. Nvidia’s investment in CUDA left shareholders scratching their heads for about six years.

Trillion-dollar business

But developers eventually realized that GPUs excelled in the massive computations needed for AI, dwarfing traditional CPUs.

This surprised Huang, who nonetheless began emphasizing the importance of AI in his public speeches. Today, GPUs have unparalleled value in AI model training. Any mainstream AI application is benchmarked on GPU training and CUDA programming.

Throughout Nvidia’s history, its leaps have landed on the cusp of opportunity. As ChatGPT takes the world by storm, Huang finds himself at the head of a trillion-dollar business.

During this year’s GTC, Huang revealed that Nvidia’s DGX is the engine behind large-scale models. Now, half of the Fortune Global 100 companies have installed the product.

Ahead in the cloud

In March, Nvidia launched DGX Cloud with monthly services for training AI models and other applications.

“With just a web browser, Nvidia’s AI computing power can be accessed by any company. We are delighted that our business can speedily expand in the cloud,” said Huang. “We want people to be able to access our services through any public cloud.”