Nvidia Corp. founder and Chief Executive Jensen Huang says artificial intelligence, enabled by the company’s most powerful hardware, is being used to accelerate computer chip manufacturing and create even more advanced microprocessors.

Speaking to an audience of leaders in the semiconductor, technology and communications industries at Interuniversity Microelectronics Centre’s flagship annual event, ITF World, Huang (pictured) said the exponential performance increases in central processing units were the governing dynamic in the tech industry for almost four decades. However, the rate at which CPUs become more efficient and powerful is slowing as it becomes exponentially more difficult to squeeze more transistors onto silicon wafers. Meanwhile, demand for more advanced computing hardware is soaring, he said.

“As a result, global demand for cloud computing is causing data center power consumption to skyrocket,” Huang said.

To meet this demand, Nvidia has emerged at the forefront of a new approach that involves coupling the parallel processing capabilities of graphics processing units with CPUs. It’s this new approach that sparked the AI revolution, Huang said, and Nvidia has responded by reinventing its computing stack for deep learning, creating new opportunities in robotics, autonomous vehicles and manufacturing.

According to Huang, advanced chip manufacturing involves more than 1,000 steps to produce features the size of a biomolecule. What’s more, each step must be handled almost perfectly to yield functional output.

“Sophisticated computational sciences are performed at every stage to compute the features to be patterned and to do defect detection for in-line process control,” Huang said. “Chip manufacturing is an ideal application for Nvidia’s accelerated and AI computing.”

AI-accelerated chip manufacturing

In his keynote, Huang outlined several ways in which Nvidia’s GPUs are being used to advance chip manufacturing processes. For instance, D2S Inc., IMS Nanofabrication GmbH and NuFlare Technology Inc. are using Nvidia’s hardware to aid in the building of “mask writers,” which are specialized machines that create photomasks, or stencils that transfer patterns onto silicon wafers using electron beams. With this, Nvidia’s GPUs are helping to accelerate the computationally demanding tasks of pattern rendering and mask process correction, Huang said.

Meanwhile, Taiwan Semiconductor Manufacturing Co., KLA Corp. and Laser Technology Inc. are leveraging Nvidia’s GPUs to aid in extreme ultraviolet light and deep ultraviolet light mask inspection. The GPUs help by processing classical physics models and deep learning to generate synthetic reference images and detect defects, Huang explained.

Moreover, Nvidia is working with various chipmakers to accelerate computational lithography with its GPU hardware. Huang said computational lithography involves simulating Maxwell’s equations of light behavior as they pass through optics and interact with photoresists. It’s the largest computational workload in chip design and manufacturing, typically consuming tens of billions of CPU hours each year. He explained that chipmakers will run enormous data centers 24/7 on such workloads as part of the design process of reticles for new chips.

To make these workloads more efficient, Nvidia recently announced a new software library called Nvidia cuLitho, which includes various optimized tools and algorithms for GPU-accelerated computational lithography. Huang said he believes this will be the single most important advance in pushing chip design beyond two nanometers.

“We have already accelerated the processing by 50 times,” Huang said. “Tens of thousands of CPU servers can be replaced by a few hundred Nvidia DGX systems, reducing power and cost by an order of magnitude.”

Embodied AI

Looking to the future, Huang believes we’re on the verge of creating a new generation of AI systems that he terms “embodied AI.” These are intelligent systems that will be able to understand, reason about, and interact with the physical world, he said. Some examples include robotics, autonomous vehicles and chatbots that can understand the physical world.

Nvidia has taken its first steps toward embodied AI. Huang introduced a new multimodal embodied AI called Nvidia VIMA, which can perform tasks from visual prompts, such as rearranging a number of objects to match a scene that it has been given. It can learn from concepts and act accordingly, such as “This is a widget,” “That’s a thing” and then “Put this widget in that thing,” he said.

In a second project, Nvidia is building a digital twin of the Earth, called Earth-2, which will be able to make predictions about the weather and climate change, Huang said. Earth-2 lives within Nvidia Omniverse, a 3D development and simulation platform, and incorporates a physics AI model called FourCastNet that’s able to emulate global weather patterns 50,000 to 100,000 times faster than existing models. It runs on Nvidia’s GPUs, and will be used to address the need for cheap, clean energy, solutions to climate change, and the advancement of semiconductor manufacturing, the CEO promised.

“I look forward to physics-AI, robotics and Omniverse-based digital twins helping to advance the future of chip manufacturing,” Huang said.

Photo: Nvidia