Skip to main content
  1. Home
  2. Computing
  3. News

Nvidia is leveraging GPUs to build 2nm (and smaller) chips

Add as a preferred source on Google

As part of Nvidia’s GPU Technology Conference (GTC), the company announced a new software library that will enable 2nm and smaller transistors. The library is called cuLitho, and Nvidia has already partnered with the world’s largest semiconductor companies to use the tech, including TSMC and Synopsys.

The software stack is built for computational lithography, which is the process of etching the components of a processor into a silicon wafer. Lithography has gotten more complex as the demand for smaller manufacturing processes has increased, but the current technology has reached an inflection point.

Nvidia's cuLitho technology on a silicon wafer.
Nvidia

Nvidia says cuLitho goes beyond these limits, allowing semiconductor brands to build 2nm transistors and even smaller. Intel announced in 2021 its Extreme Ultraviolet (EUV) lithography to push these limits as well, and the company says its upcoming Intel 4 node will be the first to take advantage of EUV lithography.

Recommended Videos

The difference with cuLitho is that it runs on GPUs. Nvidia says this approach is 40 times faster than the current lithography technique, which the company claims accounts for tens of billions of CPU hours every year. With GPUs at the helm, Nvidia says cuLitho can reduce power consumption by nine times over current techniques. In practice, that’s moving from 40,000 CPU systems to 500 Nvidia DGX H100 with the same results, according to Nvidia.

It seems the industry is on board with the advancements as well. Taiwan-based TSMC is the world’s largest semiconductor manufacturer, supplying chips for Nvidia and AMD GPUs, Apple products, and most chips from Qualcomm, MediaTek, and Broadcom. C.C. Wei, CEO of TSMC, said that the introduction of cuLitho is “making important contributions to the continuation of semiconductor scaling.”

That’s at odds with Nvidia’s stance just a few months ago, when Nvidia CEO Jensen Huang boldly claimed Moore’s Law dead in a justification for higher GPU prices. According to Nvidia, cuLitho not only bypasses these bottlenecks, but also does so in less time. Aart de Geus, CEO of Synopsys, says that tasks that used to take weeks can now be accomplished in days with cuLitho.

Nvidia’s announcement could accelerate timelines at some of the world’s largest semiconductor brands. According to recent reports, Apple is looking into 3nm chips to power the next generation of MacBooks, after reportedly sidelining plans to move to a smaller node in 2022.

Jacob Roach
Former Lead Reporter, PC Hardware
Jacob Roach is the lead reporter for PC hardware at Digital Trends. In addition to covering the latest PC components, from…
How Claude helped my 65-year-old dad finally ditch his handwritten ledgers
AI has a lot to answer for, but this one small win is hard to argue with, at least for me.
Claude app on iPhone

My dad has owned a small business for as long as I can remember, and for just as long, he's kept his books the old-fashioned way. Every sale gets written down by hand so he can file his taxes later. The problem is that his accountant needs this data in Excel, and my dad, who didn’t grow up around computers, has never learned how to use it.

For years, his workaround was paying someone to manually type his handwritten entries into a spreadsheet. It worked, but it was adding additional cost to his business, which he wanted to avoid, but couldn't.

Read more
AI’s energy tax was already concerning. Research says AI agents are over hundred times worse
AI agents could consume 136 times more energy than today's AI, study finds
AI agents

The AI industry's soaring electricity demand has already become a growing concern for governments, utilities, and technology companies. But a new study suggests the next generation of artificial intelligence could make that problem significantly worse.

Researchers from the Korea Advanced Institute of Science and Technology (KAIST) have published what they describe as the first comprehensive analysis of the energy cost of AI agents - AI systems capable of reasoning, planning, and completing tasks autonomously. Their findings show that these systems can consume up to 136.5 times as much energy per query as conventional generative AI models, raising fresh questions about whether the infrastructure supporting tomorrow's AI is ready for what's coming.

Read more
I hope Apple keeps the MacBook Neo away from the AI hype and preserves its true identity
The cheapest MacBook beats the cheapest AI MacBook.
Computer, Electronics, Laptop

If there's one thing that has disrupted consumer tech economics over the last year while changing how we understand and recommend products, it's the ever-rising cost of memory and chips. 

The desperate need to scale up AI infrastructure has pushed major manufacturers to prioritize enterprise demand, leaving everyday consumers with far fewer choices. Those available cost significantly more than they did a year ago.

Read more