One of the major announcements from Intel's Tech Tour was the company's decision to adopt an annual AI product cadence, in an effort to compete in the market.
Intel's Transition To Annual Product Cadence Is a Step In The Right Direction, But Execution Will Take Time
When examining Intel's position in the AI compute market, it has been an uncertain one, as neither the administration nor the team itself has yet concluded how to capitalize on the 'AI hype'. At the opening keynote on ITT, Intel announced that it will switch to an annual product cadence, meaning new AI architectures will be released every 12 months. More importantly, the first showcase is expected to feature an inference-optimized GPU, for which the firm plans to unveil more information at the OCP.
Of course, to play in this space, we need to get onto an annual predictable cadence of GPUs. That's what we are working on next as we think about the next stage of scaling our agentic AI solutions. We are working hard on an inference-optimized GPU that we will be talking about more at OCP in a couple of weeks.
Really excited about this product. Has enhanced memory bandwidth, lots of memory capacity, and a fantastic product for token clouds and enterprise-level inference. But while we do that, the software abstraction stays exactly the same.
- Intel's CTO Sachin Katti
Apart from the announcement of an inference-optimized GPU, Intel also confirmed that Jaguar Shores (JGS) is the company's primary choice for the most capable AI infrastructure product. Let's talk about JGS first. So, we are aware that Falcon Shores, the lineup that was intended before JGS, was canceled since Intel made an effort to 'give in' everything with Jaguar Shores, and also included a rack-scale solution around the lineup, which is why it is seen as a massive release from the firm. It is claimed to feature HBM4 and will be a multi-domain and multi-IP product.
Now, for the inference-optimized solution Intel is discussing, we are looking at an overlap between the company's consumer and AI lineups, or the GPU could be introduced in an entirely new series. It was recently rumored that Intel plans to release a low-power AI chip for inference workloads. According to Intel's CTO, the chip will feature high memory capacities. At this stage, the best approach would likely be to employ a combination of Battlemage architecture with high-capacity GDDR7 modules, but this is just speculation.
The switch to an annual product cadence comes at a time when competitors like NVIDIA and AMD are at least three to five years ahead of Team Blue, so opening up a new frontier in the AI industry becomes a lot difficult. More importantly, existing platforms, such as the Gaudi AI lineup, haven't seen widespread external adoption, so Intel ultimately needs to do a lot more to gain ground in the AI segment.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
