Home Web graphics Intel plans to make product cuts, where will they be? • The register

Intel plans to make product cuts, where will they be? • The register


Analysis With a major drop in revenue and profitability over the past six months, Intel has some tough decisions to make as it seeks to make billions of dollars in cuts while the beleaguered semiconductor giant attempts to implement is carrying out its grand comeback plan.

As announced last week, the US chipmaker plans to cut spending by $3 billion a year starting next year and then increase it to $10 billion by 2025.

While a good chunk of those expense cuts will come from laying off a “significant number” of people, the company also said it would also consider “portfolio reductions.” This means Intel will be looking to get rid of some of the products it makes and sells.

In some cases, Intel may seek to offload products to a willing buyer, as it did with the sale of its flash memory and NAND SSD business. But in other cases, the company may decide to discontinue products and write them off, as it did with its Optane memory business in July.

“We remain committed to maximizing our value creation efforts through portfolio refinement; reallocation of resources to higher yielding, higher growth businesses; mergers and acquisitions; and, where appropriate, divestitures. “Intel CEO Pat Gelsinger said during the company’s third-quarter earnings call. A week ago.

While we don’t yet know where Intel will be making product cuts, we’ve decided to make some educated guesses about where the company can cut its portfolio.

Habana AI Chips

In December 2019, Intel made the seemingly hasty decision to pay $2 billion to acquire AI chip startup Habana Labs, just weeks after the company showed off sophisticated deep learning processors from a previous acquisition. Nervana Systems.

Knowing it needed a solid response to Nvidia’s rapidly growing AI chip business, Intel apparently realized it had a better chance with Habana, so it decided to end development of its Nervana deep learning chips to focus on the new acquisition.

Three years later, Intel has made some progress with Habana. Its Gaudi training chip is now available in Amazon Web Services cloud instances, there are Habana-powered systems available from Supermicro, and the division is working on the second generation of its training and inference chips, the latter to be installed in new servers from Supermicro, Inspur and WiWynn.

The problem is that Habana probably isn’t moving fast enough to deal a significant blow to Nvidia’s multi-billion dollar AI chip business anytime soon. So far, Intel has released few details about end users who have used Habana chips, and one of them is its recently created Mobileye division.

Habana chips are still not supported by oneAPI, Intel’s answer to Nvidia’s CUDA platform. Then there’s the fact that Intel didn’t make enough money from Habana to disclose the sales in earnings.

Intel has positioned Habana chips as more efficient alternatives to GPUs for large-scale AI training and inference, so if the CPUs can do their job, it may be worth Intel sticking with them for a long time. term. But with the company badly in need of the extra money and Habana not making much money, Intel might be better off parting ways with the division to focus on its more flexible GPU architecture.

And yes, we know Intel’s GPUs aren’t making a lot of money right now either. But if Intel indeed wants to become a more viable competitor to Nvidia’s AI computing prowess, it needs some sort of dedicated acceleration chip. And with the way Intel can recoup GPU development costs through multiple product lines, GPUs may make more sense financially.

Block-scale ASIC

Here’s a question worth asking: Is the cryptocurrency mining industry stable enough for Intel to keep going? The x86 goliath just started shipping its first Bitcoin mining acceleration chip, the Blockscale ASIC, in June, promising to provide a power-efficient alternative to power-hungry GPUs.

The problem is that things weren’t going well for crypto mining companies when Blockscale hit the market, and those conditions haven’t improved.

The FinancialTimes reported on Wednesday that crypto mining companies are on shaky ground due to high energy costs and the fall in cryptocurrency values ​​that began earlier this year. The so-called crypto winter has already caused a data center provider for crypto firms, Computer North, to file for bankruptcy, and other companies are facing significant financial issues.

This includes one of Blockscale’s first customers, Argo Blockchain, which warned last week that it was at risk of going out of business due to a lack of funding. Another early Blockscale customer, Griid, is facing allegations from its lender that the company defaulted on its main line of credit.

Blockscale’s revenue may not amount to much at the moment since the product’s business unit, Accelerated Computing Systems and Graphics, only brought Intel $185 million in Q3, and most of it likely comes from GPUs from the company.

Perhaps conditions in the crypto mining industry will suddenly improve, and Intel’s bet on Blockscale will be worth it in the long run. But amid a struggling economy that could worsen, the company’s patience for Blockscale’s viability could run out sooner rather than later.

RealSense stereo cameras

Remember when Intel said last year that it was “terminating” its RealSense computer vision business? Turns out that wasn’t quite the case, as the company decided to keep its stereo cameras while phasing out its LiDAR, facial authentication, and tracking product lines.

Intel clearly has enough customers to justify further development and sales of its stereo cameras, but we’re sad to say that these sophisticated devices for robots and IoT devices would be a good choice for the chopping block, whether that means giving them up. or sell them to another company.

Simply put, RealSense cameras are outside of Intel’s chip core. Yes, Intel benefits from being able to design the chips that go inside these cameras, but the sensors themselves? Developing such devices may not be a priority when the company needs to focus on making its chips – primarily processors – more powerful and efficient than competing products.

Intel CEO Pat Gelsinger said so himself when this reporter asked him about his decision to end the RealSense business.

“Hey, there are good assets that we can harvest, but it doesn’t fit into any of these six business units that I’ve defined,” he said last year. The six divisions he referred to are Intel’s PCs, Data Centers, Networks and Edge, Graphics, Foundry, and the newly created Mobileye business.


Do not mistake yourself. Intel’s NUC mini PCs are small feats of engineering, especially ones that can pack a Core i7 processor and mid-range graphics card in a box that fits easily in your backpack.

Intel deserves plenty of credit for pushing the boundaries of miniature PC design and finding new ways to balance power, performance, and space in small form factors. But now that the company has to make tough spending decisions, it might be time to reevaluate the future of NUC under Intel.

In May, the chipmaker revealed that it had sold more than 10 million NUCs since launching the Next Unit of Computing brand ten years ago. Ten million of anything is impressive, but we have to remember that hundreds of millions of PCs are shipped each year. Dell alone shipped nearly 60 million last year of the 340 million total computers that were shipped in 2021.

It’s not that NUCs don’t provide value. While NUCs are loved by enthusiasts, they are also used by many companies around the world. It’s just that NUC sales are probably a small percentage of the company’s total revenue, and making tiny, well-designed PCs isn’t part of the company’s core chip strategy. Fortunately, Intel is far from the only company making mini-PCs.

Intel had no comment on this at the time of publication. ®

Do you have any ideas on where Intel could make cuts to its product portfolio? Leave us a comment. Or, better yet, if you know what Intel’s plans are, you can always email us confidentially.