Nvidia’s transformation from a video game accelerator to an enabler of artificial intelligence (AI) and the industrial metaverse didn’t happen overnight, but its stock price soared above $1 trillion. I was.
When Nvidia reported strong results for the three months ending April 30, 2023 and predicted a 50% increase in sales in the next fiscal quarter, the company’s stock market valuation was It skyrocketed, catapulting the company into the exclusive trillion-dollar club. Famous tech giants Alphabet, Amazon, Apple and Microsoft. Once a niche chipmaker, now a Wall Street darling, the company was becoming a household name.
Investor frenzy waned later in the week, and just as former members Mehta and Tesla dropped out before that, so did the chip designer, who dropped out of the club after a short period of time, but quickly joined the club. I returned and the investment ended in mid-June. Morgan Stanley Bank forecast that Nvidia’s value could continue to rise another 15% this year.
Unlike many of the multi-trillion-dollar tech companies, Nvidia’s lack of consumer brand awareness makes its leap from Wall Street to Main Street all the more mysterious. How Nvidia got here and where it goes next reveals how the company achieved that reputation. This is largely due to the growing importance of specialty chips in business and the accelerating interest in the potential of generative AI.
graphic driver
Nvidia started in 1993 as a fabless semiconductor company designing graphics accelerator chips for PCs. Its founders believed that generating 3D graphics in video games, a rapidly growing market at the time, would impose highly repetitive and mathematical demands on a PC’s central processing unit (CPU). I noticed. They realized that these calculations could be done more quickly in parallel by a dedicated chip than by a CPU in series, and this insight led to the creation of his first Nvidia GeForce graphics card.
For years, graphics have driven Nvidia’s business. Thirty years later, gaming graphics cards, including the GeForce series, still account for over a third of his revenue, making him the world’s largest vendor of discrete graphics cards. (However, Intel makes more graphics chips, as most CPUs have their own integrated graphics silicon.)
Along the way, another use for the parallel processing capabilities of Nvidia’s graphics processing units (GPUs) emerged, solving problems of matrix math structures similar to 3D graphics modeling.
Still, software developers looking to leverage graphics chips for non-graphics applications assemble their computations into a form that can be sent to the GPU as a series of instructions in either Microsoft’s DirectX graphics API or the open-source Open Graphics Library (OpenGL). I had to. ).
Then in 2006, Nvidia introduced a new GPU architecture, CUDA. CUDA was programmed directly in C to speed up mathematical operations and simplify its use in parallel computing. One of the first applications of CUDA was in oil and gas exploration, processing mountains of data from geological surveys.
The market for using GPUs as general-purpose processors (GPGPUs) really opened up in 2009 when OpenGL publisher Khronos Group released Open Computing Language (OpenCL).
Hyperscalers such as Amazon Web Services were soon adding GPUs to some of their compute instances, making scalable GPGPU capacity available on demand, making compute-intensive work a reality for businesses around the world. Barriers to entry into roads have been lowered.
One of the biggest drivers of demand for Nvidia chips in recent years is AI, and more specifically the need to perform trillions of iterations to train machine learning models. Some of these models are really huge. OpenAI’s GPT-4 is said to have over a trillion parameters. Nvidia was an early proponent of OpenAI and also built a special compute module based on the H100 processor to accelerate the training of the Large Language Models (LLMs) the company was developing.
Another unexpected source of demand for the company’s chips is cryptocurrency mining, the computation of which can be performed faster and more energy-efficiently on GPUs than on CPUs. Demand for GPUs for cryptocurrency mining has led to a shortage of graphics cards for years, with GPU makers like Nvidia akin to pickaxe retailers during the California Gold Rush.
Nvidia’s first chips were used to power 3D games, but the manufacturing industry’s interest in 3D simulation is even deeper. Beyond the underlying OpenGL and OpenCL rendering and acceleration code libraries, Nvidia has developed a software platform called Omniverse. Omniverse is an industrial metaverse used to create and view digital twins of products and entire production lines in real time. The resulting images can be used for marketing and collaboration of new designs and manufacturing processes.
Struggling to stay in the trillion dollar club
Nvidia is moving forward in many areas. On the hardware side, we continue to sell GPUs for PCs and some game consoles. We supply compute accelerators to server makers, hyperscalers and supercomputer makers. And we manufacture chips for self-driving cars. It also has a service business and operates its own cloud infrastructure for pharmaceutical companies and manufacturing industries. Additionally, the company is also a software vendor, with a general-purpose library of code that anyone can use to speed up computations on his Nvidia hardware, as well as more specific offerings such as the cuLitho package, which optimizes the lithography stage of semiconductor manufacturing. I am developing a tool.
But interest in modern AI tools such as ChatGPT (developed on Nvidia hardware) is sparking a new wave of demand for Nvidia hardware, helping companies develop and train generative LLMs. We are urging the company to develop new software for AI is based.
Nvidia also offers cloud-based LLMs as a one-stop shop for businesses lacking the resources to build, tune, and run custom LLMs trained on their own data to perform industry-specific tasks. It pitches AI Foundations, a generative AI service. The move, announced in March, may be a smart move given the company’s growing interest in generative AI, but it puts the company in direct competition with hyperscalers who also use Nvidia’s chips. It will be.
The Nvidia AI Foundations model includes NeMo, a cloud-native enterprise framework. AI that can generate Picasso, images, videos and 3D applications. Because BioNemo deals with molecular structures, generative AI is of particular interest to accelerate drug development, which can take up to 15 years to bring a new drug to market. Nvidia says its hardware, software and services can cut early-stage drug discovery from months to weeks. Amgen and AstraZeneca are among the pharma companies testing this space, and U.S. pharma companies alone spend more than $100 billion a year on research and development, more than triple Nvidia’s revenue. and the potential upside is clear.
Drug development is progressing rapidly, but the path to adoption in another of NVIDIA’s target markets is less clear. Self-driving cars have been “just around the corner” for years, but they are proving to be tested and approved for use on open roads. It’s even more complicated than getting approval for a new drug.
Nvidia gets two hits in this market. One is to build and run virtual worlds in which self-driving algorithms are tested without endangering anyone. Another is the car itself. Once the algorithm gets out of the virtual world and onto the road, he’ll need a chip like Nvidia to process real-time images and perform the myriad of calculations needed to keep the car on course. prize. This is the smallest market segment Nvidia has captured in its quarterly results, with just $300 million in the three months ending April 30, 2023, or just 4% of his total revenue. . However, this segment is more than doubling every year.
When reporting these results, Nvidia made an ambitious forecast of a more than 50% increase in revenue for the next fiscal quarter ending July 31st. We’ll have to wait until August 23rd to see if it lived up to expectations.
