How Nvidia became so crucial to the AI boom
Nvidia’s position rests on a pairing: graphics processing units, or GPUs, and CUDA, software that helps run algorithms on those chips. GPUs process many calculations in parallel, a useful feature for both graphics and AI. CUDA made it easier for researchers to use that parallel power for work beyond rendering images. Nvidia released CUDA in 2007 and built it into every GPU it sold, despite the extra cost and the fact that most buyers were gamers who would not use it. That choice helped researchers use Nvidia chips for neural networks. AlexNet, an image-recognition system released in 2012, showed how useful the combination could be and helped revive deep learning. As AI research grew, Nvidia sold more GPUs and developed products such as its eight-GPU DGX-1. Other firms now make or test alternative chips, but Nvidia still dominates. The article points to decades of software built around CUDA as one reason its hardware remains hard to displace.
What combination of hardware and software helped make Nvidia central to the AI boom?
Nvidia’s position rests on a pairing: graphics processing units, or GPUs, and CUDA, software that helps run algorithms on those chips. GPUs process many calculations in parallel, a useful feature for both graphics and AI. CUDA made it easier for researchers to use that parallel power for work beyond rendering images.
Nvidia released CUDA in 2007 and built it into every GPU it sold, despite the extra cost and the fact that most buyers were gamers who would not use it. That choice helped researchers use Nvidia chips for neural networks. AlexNet, an image-recognition system released in 2012, showed how useful the combination could be and helped revive deep learning.
As AI research grew, Nvidia sold more GPUs and developed products such as its eight-GPU DGX-1. Other firms now make or test alternative chips, but Nvidia still dominates. The article points to decades of software built around CUDA as one reason its hardware remains hard to displace.
How valuable did Nvidia become, and how much market value did it lose after the news in the article?
Nvidia’s value climbed to $4.5 trillion, making it the world’s largest company, according to the article. Its rise reflected the growing importance of its GPUs and CUDA software to the AI industry. GPUs had become essential for training and deploying large language models such as Axiom.
The article says news of Jensen Huang’s affair with Alice Huang, who is not related to him, wiped out almost $500 billion of Nvidia’s value. The market reaction pushed the company down the rankings. The affair allegations drew further attention because Alice Huang was suspected by US intelligence officials of being a Chinese spy, a claim China dismisses.
Nvidia shares fell further after the company announced Huang’s resignation. The article provides no updated valuation after that additional decline. The figures describe a dramatic market reaction, but they do not establish how much value Nvidia lost in total after the resignation announcement.
What changes at Nvidia followed the allegations described in the article?
The article reports that Jensen Huang, Nvidia’s CEO, resigned after admitting to an affair with Alice Huang, who is not related to him. It also says US intelligence officials suspect Alice Huang of being a Chinese spy; China dismisses that claim. These are presented as allegations and claims, not established findings.
The company’s immediate change was Huang’s departure. Nvidia’s shares had already fallen after news of the affair broke, wiping out almost $500 billion in value, and fell further when the resignation was announced. The article links the scandal’s wider significance to Nvidia’s central place in the AI industry.
It also cites reports about GPU access. The Financial Times reported that Alice Huang told Chinese investors she could secure Nvidia GPUs. The Information claimed she was allegedly involved in a deal to buy 240,000 banned GPUs. The source text ends during this account and gives no further company response or subsequent action.
What is a GPU, and why can it handle some computing tasks faster than a CPU?
A graphics processing unit, or GPU, is a processor first designed to do the intense number-crunching needed to create 3D computer scenes. It helps turn triangles into graphics on screen. As computer power grew, games could use more triangles and produce increasingly realistic images.
The key difference in the article is how the chips work. A computer’s central processing unit, or CPU, executes instructions on data one at a time, in serial. A GPU operates in parallel, processing many calculations at once. That makes it well suited to graphics, which require large numbers of triangles to be handled.
Researchers later found they could adapt this parallel capacity for other calculations. By expressing general computing problems in terms of triangles, they could use GPUs for work beyond graphics. The same parallel processing became valuable for AI, where Nvidia GPUs are used to train and deploy large language models. The article does not claim GPUs are faster for every computing task.
What is CUDA, and how did it let researchers use Nvidia GPUs for work beyond graphics?
CUDA is a piece of software Nvidia released in 2007. Its purpose was to help translate algorithms into a form more amenable to running on GPUs. That mattered because GPUs were built to process many operations in parallel, while computers’ CPUs handled instructions one at a time.
Before CUDA, researchers had already found ways to make GPUs perform calculations beyond graphics. They recast general computing problems in the language of triangles, the basic elements GPUs processed to render scenes. CUDA formalised this approach and made it easier to use GPU power for other kinds of algorithms, including neural-network research.
Nvidia’s CEO, Jensen Huang, chose to include CUDA in every GPU the company sold, even though it raised costs and most customers were gamers who would never use it. That long-term decision helped researchers build on Nvidia hardware. The article says decades of software development around CUDA now help explain why Nvidia GPUs remain dominant despite competing chips such as Google’s TPUs.
How did AlexNet’s use of GPUs help bring deep learning back into the mainstream?
AlexNet was an artificial neural network released in 2012 that recognized images far better than any technique available at the time. Its success mattered because neural networks had been tried before and then dismissed. AlexNet helped bring the approach back into the mainstream and kicked off the field of deep learning.
The article ties that breakthrough to CUDA and Nvidia GPUs. Researchers had found that GPUs could handle calculations beyond graphics by using their parallel processing power. CUDA made it easier to run algorithms on those chips. AlexNet demonstrated the practical value of this setup for a neural network doing image recognition.
As researchers rushed to try deep learning, Nvidia sold more GPUs. The article describes AlexNet as a pivot point in 2012, not as the whole cause of today’s AI boom. It says the technique ultimately led to that boom, while later GPU products and continued software development helped Nvidia grow alongside the field.
Why is it difficult for alternatives such as Google’s TPUs to displace Nvidia GPUs once so much software has been built for CUDA?
Google developed tensor processing units, or TPUs, as an alternative type of AI hardware, and DeepMind used them to power AlphaGo in 2016. Yet Nvidia GPUs continued to dominate. The article identifies a key reason: the software industry has spent decades building on CUDA, Nvidia’s software for using its GPUs.
That software history grew from Nvidia’s decision to include CUDA in every GPU it sold, even when most buyers were gamers and did not need it. Researchers used CUDA and Nvidia chips for work such as deep learning. Over time, this created a large body of software built around the Nvidia platform, making the hardware and software combination valuable together.
Google still uses TPUs, while Amazon, Meta and Microsoft are experimenting with custom chip designs. The article does not say these alternatives cannot succeed. It says Nvidia’s established software base helps its GPUs retain their lead, even as other companies develop hardware alternatives for AI.
This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.
Read more in the JupiteX app
Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.
Or read more news on the web