TECHnalysis Research Blogs
TECHnalysis Research president Bob O'Donnell publishes commentary on current tech industry trends every week at LinkedIn.com in the TECHnalysis Research Insights Newsletter and those blog entries are reposted here as well. In addition, those columns are also reprinted on Techspot and SeekingAlpha.

He also writes a regular column in the Tech section of USAToday.com and those columns are posted here. Some of the USAToday columns are also published on partner sites, such as MSN.

He also writes occasional columns for Forbes that can be found here and that are archived here.

In addition, he has written guest columns in various other publications, including RCR Wireless, Fast Company and engadget. Those columns are reprinted here.

July 24, 2026
AMD Lays Out AI Vision Across Cloud, Client and Robotics

By Bob O'Donnell

When a market is utterly dominated by a single player, as the AI infrastructure world has been by Nvidia, it can be tough to make an impact. No company knows this better than AMD. They’ve been battling in many segments against Nvidia, as well as other semiconductor giants like Intel, for decades now.

Clearly the challenge hasn’t deterred AMD in the least. In fact, based on the impressive array of announcements the company made at its recent Advancing AI event in San Francisco, it’s actually inspired them. The company used its now annual event to make big announcements in AI infrastructure for the cloud and enterprise data centers, AI-powered client devices, and the physical AI world of robotics.

The big news was the official launch of the Helios rack, AMD’s first full AI infrastructure rack solution powered by the company’s latest 6th generation Epyc CPUs, their Instinct MI455 datacenter GPUs and several Pensando networking chips. The company first announced the system at last year’s event, but bringing it to market is an extremely important step both philosophically and practically. Thanks to Nvidia’s initial efforts, the world has quickly come to expect rack and even datacenter-level solutions as requirements for AI infrastructure and without a shipping rack product, AMD was at a decided disadvantage. With shipments of Helios now on tap for later this quarter, it’s clear that AMD is officially now a serious contender for AI infrastructure and organizations can feel confident that they have a viable alternative from which to choose.

Not surprisingly, that confidence was reflected by the impressive array of customers and partners that AMD brought out on stage at the event. From Anthropic and OpenAI on the frontier model side, to Meta and even enterprise customers like AT&T, there was a strong sense of energy and enthusiasm behind AMD’s efforts and a genuine sense of partnership and collaboration with many of them. In fact, one of the things that AMD CEO Lisa Su went out of her way to point out was the level of co-design and cooperative engineering efforts that occurred between AMD and some of its earliest customers. It was a textbook example of how important it is to truly listen to what potential buyers of a product want, and then incorporate those requests into the final product.

Beyond just building a credible alternative, AMD made some pretty aggressive claims that it was building some even higher performing options than its competition. While it’s always to tough to interpret exactly how reflective of real-word performance certain benchmark comparisons really are, AMD touted 20% greater performance for its latest Epyc CPUs vs Nvidia’s new Vera CPU among other points. In addition, the company discussed the larger amount of HBM4 memory and faster connections to that memory in Helios versus similar Nvidia rack solutions. Regardless of how the performance claims play out—and more independent 3rd party benchmarks will undoubtedly be coming soon—it seems fair to say that AMD is providing a credible hardware alternative to Nvidia. Plus, for those looking for low-latency inferencing options, AMD also made a surprising new announcement with Cerebras that will enable a Cerebras rack to sit side-by-side with a Helios rack for certain types of demanding applications.

Of course, there’s more to the AI infrastructure game than just hardware and one of the more intriguing announcements from Advancing AI was a new version of the company’s software stack that they called ROCm.ai. Cleverly, it’s basically designed to use advancements in AI-powered programming to help make the process of creating software that’s optimized for AMD hardware much easier. In practical terms, that means for example, that it could be used to make the process of porting code originally written in CUDA over to ROCm significantly easier. Given Nvidia’s roughly 18-year history and advances in CUDA development a wholesale change won’t happen overnight, but it certainly seems to make the “CUDA moat” look a less daunting than it first appeared. In addition, a feature of ROCm.ai called Hyperloom, which is the part specifically designed to optimize AI applications for ROCm, could prove to be particularly compelling. A representative from Anthropic, for example, talked about how they were able to port some of their applications to run natively on a Helios rack over a weekend.

On top of all the datacenter and rack system announcements and forward-looking timelines, AMD also made several interesting PC and robotics-based announcements at Advancing AI. For PCs, the company disclosed a new version of an AI-focused deskside computing device based on an enhanced version of their Ryzen AI Max APU, which incorporates both a Zen5-based CPU and a RDNA 3.5 GPU. Codenamed Gorgon Halo, the device will offer up to 192 GB of unified memory—at a price still TBD but likely to be quite high given the cost of DRAM these days. Still, it’s an interesting example of the new class of developer-focused systems we’ve started to see on the market.

What was even more surprising, however, was a partnership with Cisco to bring AI agent management, observability and tokenomics tools to this class of new devices. Leveraging Cisco’s latest Cloud Control, AI Defense and Splunk technologies, the new software solution is being specifically designed for organizations who are looking to deploy larger numbers of AI-focused PCs. More details still to come, but it’s an interesting partnership that highlights how AMD (and Cisco) is/are looking to provide more comprehensive AI tools to the enterprise.

Finally, AMD also made a number of big announcements in physical AI and robotics. While most people don’t realize it, AMD has had a large presence in industrial robotics for quite some time, particularly because of the technologies and products it acquired when it purchased Xilinx back in 2022. The kinds of advanced FPGAs that Xilinx has been creating for years are an essential part of many robotics systems. The ability to reprogram them as new AI-powered algorithms are developed for robotic operations as well as the role they can play in sensor fusion (which is used to tie together all the camera and other sensor feeds found on most robotics systems) make them very important for robotics applications. Combining this with AMD’s traditional CPUs and GPUs gives the company a strong portfolio of choices. They leveraged all these capabilities into what they’re calling the Kria AI Module, as well as the Kria AI SOM (System on Module) board for use in other designs. Both devices are based on the company’s Ryzen AI Embedded X100 series SOCs as well as the new open-source Kria AI Robotics Platform software. All told, it’s a comprehensive line of new tools for robotics developers and helps solidify AMD’s position as a significant player in the burgeoning field.

While there’s no question that Nvidia will continue to be the leader in AI infrastructure for some time to come, it is now clear that AMD has definitely come to play and the result is good news for everyone. As in any market, competition always drives better, faster advancements and it will be exciting to watch how the world of AI semiconductors and solutions continue to evolve.

Here’s a link to the original column: https://www.linkedin.com/pulse/amd-lays-out-ai-vision-across-cloud-client-robotics-bob-o-donnell-uwrlc

Bob O’Donnell is the president and chief analyst of TECHnalysis Research, LLC a market research firm that provides strategic consulting and market research services to the technology industry and professional financial community. You can follow him on LinkedIn at Bob O’Donnell or on Twitter @bobodtech.