AMD CEO says rapid AI growth brings CPUs back into focus
Taipei, May 23
The artificial intelligence infrastructure market is growing at a rapid pace, with rising inferencing demand bringing central processing units back to the core of computing, AMD Chair and CEO Lisa Su said Friday, according to reports by Focus Taiwan.
During a Q&A session at the CommonWealth Magazine 45th Anniversary Summit in Taipei, Su was asked how AMD can compete with Nvidia, which dominates the AI infrastructure market.
She said the rapid expansion of the AI infrastructure market was a key factor to consider, given that the data center segment alone could exceed US$1 trillion within the next three to four years.
That growing market will require a wide range of technologies, including CPUs, graphics processing units (GPUs) and application-specific integrated circuits (ASICs), Su said.
"I think the advantage that we have at AMD is that we have all of these components," she said.
Just six to 12 months ago, CPUs were not a focal point of discussion and were not seen as in short or tight supply, but with rising inferencing demand, CPUs are now back at the centre of attention, Su said.
"This is where AMD is very, very strong," she said, adding that she expects the company to perform well across the broader AI infrastructure as the market evolves to include a wide range of use cases.
During the event, she was also asked how AMD views the growth potential of CPU technology in AI racks and how it is positioning itself for that opportunity.
Su said AMD was highly optimistic about CPU growth in AI racks, noting that CPU demand had been relatively flat in recent years as attention focused on GPUs.
That has changed with rising inferencing and agentic AI demand, she said, and she projected that the CPU market could grow by more than 35 per cent annually over the next five years.
Su said AMD's strategy is not to build a single CPU, but a full family of processors. Its new flagship CPU, for example, is built on 2-nanometer technology and code-named Venice, is optimised for cloud, throughput, and head node workloads.
"I think that's how you win. It's not just one design point, but actually we have a whole family of design points," she said.
During her speech at the CommonWealth event, the AMD CEO said AI is not only entering the inference stage but also an era in which it can perform "truly intelligent tasks." It will be able to solve complex problems and reshape how products are built, companies are run, and science is explored, she said.
Earlier on Thursday, AMD announced in a press statement that it will invest more than US $10 billion in Taiwan's AI supply chain ecosystem as global demand for AI infrastructure continues to surge.
— ANI
Reader Comments
As someone working in Bengaluru's tech sector, this makes sense. The inference stage of AI needs balanced compute—not just brute-force GPUs. AMD's strategy of having CPUs, GPUs, and ASICs under one roof is smart. But I worry about Taiwan's geopolitical risks. India should definitely push for more local semiconductor manufacturing. 🇮🇳⚙️
Data center market exceeding $1 trillion in 3-4 years? That's huge. CPUs coming back into focus for inferencing makes technical sense—many AI workloads don't need the peak parallel processing of GPUs. AMD's 'family of design points' approach is a smart hedge. Let's see if they execute.
Lisa Su is a rockstar—Indian-origin CEO leading AMD in this AI race. The 35% CPU growth projection over 5 years is ambitious. But I wish AMD focused more on power efficiency for data centers in India, where electricity costs are high. Still, great to see competition in the AI hardware space. 💻🔥
Agentic AI demand driving CPU growth makes sense—these systems need efficient task management and orchestration. AMD's $10 billion Taiwan investment shows serious commitment. But with Nvidia's CUDA ecosystem lock-in, winning mindshare won't be easy. Open-source frameworks could level the playing field though.
From a Chennai-based IT perspective, this shift back to CPUs is welcome. Many legacy enterprise systems here run on x86, and upgrading to pure GPU-based AI is costly. AMD offering a full CPU family means we might see more balanced, cost-effective AI
We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.