Intel's AI plans have been in limbo for several quarters, but the company looks set to target two main areas: ASICs and edge AI.
There is no doubt that in the field of computing solution providers, Intel has been dominant in the past, especially in general-purpose computing and server-level computing. However, in the field of artificial intelligence, the gap between Intel and Nvidia and AMD is huge. Former Intel CEO Pat Gelsinger also admitted that Intel's artificial intelligence strategy is not satisfactory. Even during Chen Libu's tenure as CEO, we may not see a clear structure until now. At Barclays' annual global technology conference, Intel Vice President John Pitzer gave his insights on the company's future direction in artificial intelligence:
"As AI expands from the center to the edge, we are working on a power-optimized GPU for inference. We still have a lot of work to do in this area, so please look forward to it. This will take some time. Another dynamic that I think is worth exploring Lip-Bu introducing is the ASIC business segment.
I think the other interesting thing about the ASIC business is that we can adopt a model similar to Broadcom Marvell. But given the Intel foundry model, many ultra-large-scale data center operators are looking to work directly with foundries in an attempt to bypass the traditional model."
Today, one of Intel's main strategies is to take full advantage of edge AI, and the company is achieving this goal through its "AI PC" product portfolio, which includes Meteor Lake for laptops,Lunar LakeAs well as the upcoming Panther Lake series. We have seen Intel improve the performance of edge AI by integrating a new generation of NPU, which significantly improves the computing power of mobile SoCs. What's more, Intel also plans to expand its edge product line with products such as Crescent Island, Crescent is an inference-focused processor with on-board LPDDR5X memory.
A more attractive strategy is to set up a dedicated ASIC unit, which is currently headed by Iyengar Srini and is part of the Central Engineering Group (CEG). We know that companies are looking to customize chips to meet their specific workload needs. To date, Google's TPU and Amazon'sTrainiumsolutions have attracted wide attention in the industry. ASIC is an emerging market segment in the field of artificial intelligence, so it is expected that Intel will develop its artificial intelligence strategy around ASIC, providing customers with a "one-stop" solution for manufacturing and advanced packaging.
Intel Vice President John Pitzer said the company is customizing the network.ASIC chipThe field has a "booming" business, and they have acquired many customers for smart network card ASIC chips. These chips are designed for network-intensive workloads such as network packet processing, telemetry, traffic management, and more. Intel ultimately wants to keep pace with the likes of Broadcom and Marvell, but the company plans to differentiate itself by offering in-house foundry services to customers, reducing time to market and working closely with ASIC chip customers.
It is worth mentioning that Chen Liwu has deep experience in promoting the business model of customized chips. This is evidenced by his position at Cadence, where he has been working on intellectual property, design tools, design ecosystem collaborations, and custom chip verticals. His experience and market contacts will help Intel better grasp the "ASIC boom".
Intel's new ASIC business will "serve a wide range of external customers"
Intel has decided to move into the ASIC and design business as its next big business. According to the latest comments from CEO Chen Liwu, the business will play a vital role in the company's operations.
For those who don't know, Intel recently announced the formation of the Central Engineering Group (CEG), which aims to consolidate all the engineering talent within the company into one department. The move would give the company's engineers "more power". The division is led by Iyengar Srini, who joined Intel in July from Cadence Systems. Although it seems that the department is not an indispensable part of the company, according to CEO Chen Libu's past statements, Intel's CEG Group is likely to open up a new source of revenue for the company and make up for the mistakes in the artificial intelligence boom.
And, just as importantly, the [CEG] team will lead the way in building our new ASIC and design services business, providing custom chips to a wide range of external customers. This will not only expand the reach of our core x86 IP, but will also leverage our design strengths to provide a range of solutions from general-purpose computing to fixed-function computing. -Intel CEO Chen Liwu said in the third quarter earnings conference call.
Okay, this all sounds a little complicated, so let's read it through. If you look at the current state of Intel, the company is not offering a competitive product line for AI customers, and the next "blockbuster" is likely to beJaguar Shores AIThe rack-mounted accelerator product line is expected to be released in 2027. What's more, both Nvidia and AMD have built well-established AI hardware portfolios, leaving Intel with little room to compete; however, this is where CEG and Intel's ambitions in the ASIC space come into play.
Intel has chip technology expertise, x86 IP, and an in-house foundry that provides manufacturing services, so customers looking for custom AI chips can actually get a "one-stop shop" for all needs. This is an advantage that no other ASIC designer in the market can provide, even Broadcom, Marvell or Alchip, because Intel's foundry service is the key to its differentiation from other companies. More importantly, with the CEG Group, Intel has centralized horizontal engineering, which means that the cost of combining design services with manufacturing and packaging has been significantly reduced.
Of course, there are also many opportunities in the "middle" of the AI supply chain, such as profiting from the profits of mass production and even ASIC design fees. It is worth mentioning that Chen Libu has deep experience in promoting the business model of customized chips.CadenceThe position is the best proof. In Cadence, he focuses on the intellectual property business, design tools, design ecosystem collaboration, and vertical markets for custom chips. His experience and market contacts will accelerate Intel's profit from the "ASIC boom.