Dive Brief:
- Nvidia raised the stakes in the global race to deploy bigger, better and faster AI processors Sunday during the Computex 2024 keynote in Taiwan.
- President and CEO Jensen Huang announced the company would follow up its new Blackwell architecture with Blackwell Ultra next year and an accelerator called Rubin in 2026.
- The race to develop AI-optimized silicon intensified as rival chipmaker AMD unveiled a new generation of neural processing unit PC chips named Zen 5 Ryzen Sunday. Leading manufacturers Acer, Asus, HP, Lenovo and MSI plan to manufacture PCs embedded with the processors, AMD said.
Dive Insight:
As large language models crowd the IT landscape, Nvidia is reaping the rewards.
Hyperscaler hunger for the chipmaker’s GPUs helped balloon Nvidia’s revenue to $26 billion during the three-month period ending April 28. Cloud provider demand for Nvidia processors drove the company’s largest reported segment, data center revenue, up 427% to $22.6 billion year over year.
While the company races to fill orders for its newest line of GPU processors, Huang reiterated the GPU giant’s commitment to advancing chip technology on a yearly cadence.
“Our basic philosophy is very simple — build the entire data center to scale, disaggregate it and sell it to you in parts on a one-year rhythm,” he said, speaking during the keynote.
The company has also been quick to move into the AI PC space. ASRock Rack, Asus, Gigabyte and a number of other computer manufacturers are incorporating Nvidia GPUs in cloud-based, on-premises and edge devices, the company said.
Gartner expects AI-optimized PCs to dominate the enterprise computer market by 2026. The analyst firm forecast 2024 semiconductor revenue to rise 33% year over year.
Intel is looking to get its share. The Nvidia competitor rolled out a new client processor, dubbed Lunar Lake, in May. The chips, which will ship in Q3, will power more than 80 new laptops across more than 20 PC manufacturers, including the Microsoft Copilot+ line unveiled last month.
AI workloads continue to run on central processing units, the workhorse of general-purpose compute and the core of AMD’s and Intel’s business. Both chipmakers have agreed to provide CPU designs for Nvidia’s modular MGX inference and retrieval-augmented generation and data processing platform, according to the announcement Sunday.
Huang is betting on processing speed and efficiency improvements to offset hardware development costs.
“We are seeing computation inflation,” Huang said. “Companies spend hundreds of millions of dollars processing data in the cloud… there’s an enormous amount of captured loss that we can now regain.”
