Jensen Huang's Big Bet On Artificial Intelligence Is Paying Off With The ChatGPT Service He Developed At Nvidia

Nvidia has been at the forefront of the computer graphics revolution for over 25 years, building a devoted following among gamers in the process.

‍

The industry for graphics processing units (GPU), which Nvidia joined in 1999 with the GeForce 256, is currently dominated by the company. Despite a recent slowdown, gaming pulled in almost $9 billion for Nvidia last year.

‍

Nevertheless, Nvidia's most recent earnings surprise suggests a brand-new trend in the GPU industry. The artificial intelligence revolution is increasingly centered on technology.

‍

CEO Jensen Huang told Trade Algo in an interview last month, "We had the good wisdom to go put the whole organization behind it. "We saw early on that this method of developing software may revolutionize society roughly ten years ago. And we made major, lateral, and bottom-up changes to the business. Artificial intelligence was the primary emphasis of every chip we produced.

‍

As the powerhouse behind LLMs like ChatGPT, Nvidia is now beginning to see the benefits of its early investment in AI. This has assisted in reducing the impact of broader semiconductor industry difficulties linked to trade tensions between the United States and China and a global chip shortage. 

‍

Not that global worries don't affect Nvidia. The United States enacted extensive new regulations in October that forbade the transfer of cutting-edge AI chips to China. Nvidia depends on sales of its well-known AI chip, the A100, in China for around a quarter of its total income.

‍

In order to reengineer all of our goods to be consistent with the rule and still be able to serve the business customers we have in China, it was a trying month or so, according to Huang. "We are able to supply and assist our clients in China with the controlled parts."

‍

Nvidia's annual GTC developer conference, which runs from March 20–23, will have a strong AI theme. In order to understand the company's place at the center of the explosion in generative AI, Trade Algo met down with Huang at Nvidia's offices in Santa Clara, California, prior to the conference.

‍

"We just assumed that someday anything new would occur, and the rest of it needs some serendipity," Huang remarked when asked whether Nvidia's fortunes are the consequence of chance or prescience. It wasn't foresight, either. Forward-thinking involved rapid computers.

‍

Around 80% of Nvidia's income comes from its principal business, GPUs. They are typically sold as plug-in cards that boost processing power to the central processing units (CPUs) created by businesses like AMD and Intel.

‍

Now, tech businesses scrambling to rival ChatGPT are publicly bragging about how many of Nvidia's roughly $10,000 A100s they have. Microsoft said that 10,000 of them were used in the supercomputer built for OpenAI.

‍

Vivek Arya, a semiconductor analyst at Bank of America Securities, noted that using their products to increase computer capacity is fairly simple. "Right now, computing power is pretty much the valley's currency."

‍

Huang demonstrated the H100, the organization's newest system, which has already begun to ship. The H denotes Hopper.

‍

Huang explained while carrying a 50-pound server board, "What makes Hopper so spectacular is this special variant of processing called transformers engine. "The generative pre-trained transformer (T of GPT) is the transformer engine. This is the first computer ever created that is capable of processing transformers on a massive scale. Hence, huge language models will be much, much quicker, and far more economical.

‍

Huang claimed to have "hand-delivered" the world's first AI supercomputer to ChatGPT creator OpenAI.

‍

Unafraid To Stake Everything

With a market price of about $600 billion, Nvidia is currently among the top ten most expensive tech companies in the world. It boasts 26,000 employees and a freshly built polygon-themed headquarters. It's additionally one of the few Silicon Valley behemoths still led by its original creator after 30 years.

‍

Huang, 60, is an immigrant who was born in Taiwan and attended Oregon State University and Stanford University to study engineering. Early in the 1990s, Chris Malachowsky, Curtis Priem, and Huang would get together at a Denny's and discuss their hopes for enabling 3D graphics on PCs.

‍

In 1993, the trio started Nvidia out of a Fremont, California, townhouse. The words "next version" (NV) and "invidia," which is Latin for "envy," served as the foundation for the name. They adopted the envious green eye as their company emblem in the hopes that they would accelerate computing to the point where everyone would turn green with envy.

‍

At the time, there were tens of GPU manufacturers, according to Arya. Due to Nvidia's excellent collaboration with the software industry and developers, they are the only ones—along with AMD—who have truly survived.

‍

Huang's ambitions and fondness for seemingly unachievable endeavors have brought the business several times dangerously close to going bankrupt.

‍

The 2021 Time magazine's Most Influential Person, Huang, remarked, "Every company is making mistakes and I make many of them. "Some of them put the company in danger, especially in the beginning because we were little, up against extremely large companies, and attempting to build this new technology,"

‍

For instance, Nvidia attempted to enter the smartphone market in the early 2010s with its Tegra processor family. The business then left the room.

‍

After firing the majority of its employees, Nvidia released the GeForce 256 in 1999, which it claims to be the first official GPU ever made. It was the first customizable graphics card that support personalized lighting and shading. Nvidia was Microsoft's initial Xbox's only graphics supplier by 2000. The business placed yet another enormous wager in 2006 when it unveiled the CUDA software package.

‍

Wall Street questioned Nvidia for ten years, asking, "Why are you taking this investment? They rated it at $0 in our market cap because 'no one is using it,' according to Bryan Catanzaro, vice president of integrated deep learning research at Nvidia. When he joined Nvidia in 2008, he was one of the few personnel working on AI. The business now has hundreds of staffers working in the space.

‍

"People didn't suddenly realize that this is a dramatically different approach of building computer programs till around 2016, 10 years after CUDA came out," Catanzaro added. It has revolutionary speedups that lead to ground-breaking outcomes in artificial intelligence.

‍

Although AI is increasing rapidly, gaming remains Nvidia's major business. The company's next significant advancement in graphics was made in 2018 using its AI skills. Based on their knowledge of AI, the business unveiled GeForce RTX.

‍

"We had to reinvent and upset ourselves, essentially modify what we invented totally, to push computer animation and computer games to the next level," Huang added. "We created this brand-new method for creating computer graphics called ray tracing, which essentially simulates the movement of light. Hence, we calculate one pixel and use artificial intelligence to imagine the other seven.

‍

In a "boom-or-bust cycle,"

Huang was determined to turn Nvidia into a "fabless" chip manufacturer, meaning that it would design the product but outsource chip production to other companies with fabs. By outsourcing the astronomical cost of manufacturing the chips to Taiwan Semiconductor Manufacturing Company, Nvidia reduces capital expenditure.

‍

Investors have every right to worry about such a high reliance on a Taiwanese business. The CHIPS Act, passed by the United States last summer, allots $52 billion as incentives for semiconductor companies to produce on American territory.

‍

"Really, U.S.-China relations and the possible effects of TSMC provide the greatest concern. Only one thing that truly gets me up at night is if I own stock in Nvidia, said C.J. analyst at Evercore named Muse. This issue affects AMD, Qualcomm, and even Intel in addition to Nvidia.

‍

According to TSMC, it will cost $40 billion to construct two new chip manufacturing facilities in Arizona. Nvidia will "definitely" employ TSMC's Arizona fabs to produce its chips, Huang said Trade Algo.

‍

Then there are concerns about supply and how many of the novel GPU use cases will continue to expand. When cryptocurrency mining took off, Nvidia experienced a rise in demand since GPUs became essential for successfully competing in that industry. The business even developed a streamlined GPU specifically for cryptocurrency. Nvidia, however, had a mismatch in demand and supply due to the collapse of cryptocurrency.

‍

Because cryptocurrency mining has a boom-or-bust cycle, this has caused issues, according to Arya. "Gaming cards run out of stock, prices spike, and there is a significant drop on the gaming side as the cryptocurrency mining boom implodes."

‍

Last year, Nvidia's new 40-series GPUs were significantly more expensive than those of the previous generation, which created serious sticker shock among certain gamers. There is currently an excess of supply, and revenue from gaming decreased by 46% from the previous quarter.

‍

Also, as more IT behemoths create their own special-purpose chips, competition is growing. Apple and Tesla both use it. Amazon and Google are also.

‍

"How can they maintain their lead is their main concern.

‍

Arya remarked. "Their clients could also be rival businesses. Microsoft may attempt to create these items on its own. These items are already being internally designed by Google and Amazon.

‍

Huang claims that such competitiveness is beneficial.

‍

According to Huang, the quantity of power required by the world in data centers will increase. "That's a serious problem for the entire world. The first thing we should do is: accelerate whatever we can in every data center in the globe, however, you choose to do it, for the cause of sustainable computing.

‍

Nvidia creates autonomous driving technologies for Mercedes-Benz as well as other automakers. Additionally, its systems are utilized to run simulations to improve the daily flow of millions of shipments and to power robotics in Amazon warehouses.

‍

It is referred to as the "omniverse" by Huang.

‍

More than 700 customers, including those in the wind turbine business, the logistics industry, and the auto industry, are currently testing it, according to Huang. It combines computer animation, artificial intelligence, robotics, and physics modeling into what is arguably the single greatest technological container ever created by Nvidia. And I have high expectations for it.