Growing use of artificial intelligence (AI) across all spheres is set to take the overall market for compute infrastructure to $2 trillion by 2030 from $365 billion in 2025, global semiconductor major AMD said on Thursday.
Eyeing the 40 per cent compounded annual growth in the market that it expects, the company unveiled a slew of products, including advanced microprocessors at its annual Advancing AI Summit here.
Lisa Su, the Chairman and Chief Executive of the company that has been competing with bigger rival Nvidia, said AI had moved beyond chatbots to reasoning models and was now entering the era of agentic AI, where AI systems autonomously execute multi-step tasks, dramatically increasing demand for computing power.
“AI is the most important technology of the last 50 years, and frankly, the progress that we’re seeing in the industry is just incredible. Every few months, we see something new that was far beyond our imagination,” Su said, adding that the industry was still in the “very, very early innings” of its evolution.
She said the rapid shift from training AI models to running them at scale would make inference the dominant AI workload, with about 60 per cent of global AI computing capacity expected to be devoted to inference this year.
“I think we’re all going to use a slew of different models, depending on what task you’re trying to solve and which industry you’re in. But the bigger shift is actually in inference,” Su said.
According to Su, AMD now expects the global AI accelerator market to expand to about $1.4 trillion by 2030, while the server CPU market is projected to exceed $200 billion, driven by rising deployment of AI agents across enterprises.
She said AMD’s strategy was centred on leadership in computing, open hardware and software platforms, and extending AI from cloud data centres to personal computers and edge devices, arguing that “an open ecosystem is essential to the future of AI”.
Highlighting the growing role of processors beyond AI accelerators, Su said AMD expects the global server CPU market to expand more than eight-fold to over $200 billion by 2030 from about $25 billion today, driven by the rapid adoption of agentic AI.
In the summit, AMD also unveiled Helios, its rack-scale AI infrastructure integrating Instinct GPUs, EPYC processors, Pensando networking and ROCm software, with production shipments expected to begin later this year.
The architecture is also expected to help data centre operators address rising requirements for power, cooling, and networking.
Without naming rival Nvidia, Su said the AI market would not be served by a single architecture or vendor, arguing that “there’s no one-size-fits-all” approach to AI chips and reiterating AMD’s push for an open hardware and software ecosystem.
She said customers increasingly wanted flexibility to deploy different models and workloads, making open platforms critical for the future of AI.
Making an indirect comparison with competing AI systems, Su said AMD’s newly launched Helios rack delivers 15 per cent more compute, 50 per cent more HBM4 memory capacity and bandwidth, and 50 per cent more scale-out bandwidth than competing offerings.
She added that the system provides 10-15 per cent higher performance across inference workloads and up to 30 per cent more tokens per dollar, positioning AMD’s focus on performance and cost efficiency as key differentiators in the AI infrastructure market.
The Advancing AI 2026 summit drew more than 10,000 attendees, including executives from leading AI companies, cloud service providers, enterprises, developers, researchers, and technology partners from around the world.
The event served as AMD’s flagship platform to unveil its latest AI hardware, software and infrastructure roadmap, with executives from companies such as Anthropic, OpenAI, Meta, Oracle, Microsoft and xAI participating in keynote sessions and discussions.
Meta Head of Infrastructure Santosh Janardhan said AI demand across inference, training and recommendation systems was growing exponentially, requiring companies to design entire data centres as integrated systems rather than focusing on individual chips.
He added that “it’s not just a GPU game anymore. It’s CPUs and GPUs. If anything, CPUs are becoming at least as important, if not more,” reflecting the growing importance of server processors in the agentic AI era.
The announcement also comes amid growing global investments in AI infrastructure.
AMD said Anthropic plans to deploy up to 2 GW of its Instinct MI450 Series GPUs through Helios, with the first gigawatt expected to begin deployment in the first half of 2027, while Microsoft plans to use Helios for Azure AI services and frontier model inference.
The large-scale deployments underline AMD’s strategy to build infrastructure for emerging agentic AI applications and provide a template for markets such as India, where demand for AI computing is expected to rise.
India is already a part of AMD’s global Helios rollout as AMD and Tata Consultancy Services (TCS) had announced plans earlier this year to deploy the country’s first Helios-powered AI infrastructure, supported by a data centre blueprint with up to 200 MW capacity.
With commercial shipments of Helios scheduled to begin in the second half of 2026, the TCS deployment places India alongside global AI infrastructure projects announced by AMD with customers such as Anthropic and Microsoft, signalling the country’s increasing role in the next generation of AI computing.
Published on July 24, 2026
