AMD deepens enterprise AI push with new data centre chips

Advanced Micro Devices Inc. unveiled a new chip aimed at smaller corporate data centres while outlining a longer-term roadmap designed to narrow the gap with market leader Nvidia Corp.

The move reflects a growing focus on companies that want to deploy AI workloads on local infrastructure, rather than relying entirely on large cloud providers.

At the same time, AMD signalled confidence that demand for advanced computing will remain strong, underpinned by the rapid expansion of AI applications across industries.

The announcement comes at a time when AMD is under pressure from investors to demonstrate tangible progress in a market that has been dominated by Nvidia.

Over the past two years, AI chips have become a multibillion-dollar business for AMD, lifting revenue and earnings.

Yet the company is still chasing a share of the tens of billions of dollars in annual AI hardware orders flowing to its larger rival, Nvidia.

New chip for local data centres

At the heart of the update was the MI440X, a new addition to AMD’s existing AI accelerator lineup.

The chip is designed for compact systems used in smaller corporate data centres, allowing enterprises to run AI models on premises and keep sensitive data within their own facilities.

By focusing on flexibility and local deployment, AMD is positioning the chip as a practical option for companies taking early or incremental steps into AI adoption.

Flagship systems and performance claims

Alongside the new model, AMD highlighted its higher-end MI455X accelerator.

Systems built around this chip are described as a significant advance in overall capability, reflecting improvements in performance and efficiency compared with earlier generations.

These systems form the backbone of AMD’s push into more demanding AI workloads, including training and inference at scale.

AMD also confirmed that its Helios system, which combines the MI455X with the upcoming Venice central processing unit design, is scheduled to go on sale later this year.

The system is intended to offer a tightly integrated platform for enterprise and research customers looking for end-to-end AI infrastructure.

AI demand and competitive pressure

AMD chief executive Lisa Su reinforced the view that AI-driven demand for computing power is far from peaking.

She echoed comments from other US technology leaders who argue that the rapid pace of AI innovation continues to push hardware requirements higher, rather than easing them.

AMD is widely seen as Nvidia’s closest competitor in the market for chips that power artificial intelligence software.

While the company has made progress in establishing itself as a credible alternative, the scale of Nvidia’s current lead means that each new product launch is closely scrutinised for signs of competitive momentum.

AMD also previewed its MI500 series of processors, scheduled to debut in 2027.

According to the company, the new range is expected to deliver performance gains of up to 1,000 times compared with the MI300 series introduced in 2023.

If achieved, such improvements would represent a major step change in AMD’s AI hardware capabilities.

AMD’s CES appearance also featured collaboration signals. OpenAI co-founder Greg Brockman appeared alongside Su to discuss ongoing cooperation with OpenAI and future system deployments.

The post AMD deepens enterprise AI push with new data centre chips appeared first on Invezz