Intel CEO: Meteor Lake Will Be Our Next 'Centrino Moment'

Intel
(Image credit: Intel)

Having experienced two consecutive quarterly losses, Intel bounced back to profitability in the second quarter and is looking positively toward the second half of the year. Intel CEO Pat Gelsinger expressed optimism about the company's future in a conversation with Yahoo Finance Live, stating that the worst is likely over for the PC market, and the upward revision of the Q3 guide indicates a brighter future. Meanwhile, Gelsinger expects the upcoming Meteor Lake CPU family to be Intel's new 'Centrino Moment.'

"I do think we have […] seen the worst," Gelsinger said. "Obviously, the PC inventory levels are now healthy again. And we are seeing the demand signals. You know, we think we still have a little bit of inventory on the networking and the datacenter to fight through in the third quarter of the year. But I really think the worst is behind us at this point." 

The head of Intel attributed his positive outlook to expected improvements in the PC business, noting that Intel's inventory levels are now healthy, which is why the chip giant now expects quarter-on-quarter improvements due to inventory level adjustments.

Gelsinger also highlighted the upcoming introduction of Intel's codenamed Meteor Lake processor, the company's first client CPU to use a multi-chiplet design and a built-in AI accelerator to boost applications like real-time language translation, transcription, and video effects, among other things. 

Interestingly, Gelsinger sees AI accelerator integration as the company's new Centrino moment. Way back in 2003, Intel was the only company to offer a fully integrated PC platform with Wi-Fi and an appropriate software stack. By contrast, AMD and Apple today have processors with integrated AI accelerators and while software support for AMD's unit may be scarce, Apple's Neural Engine AI capabilities are used by its own software. 

"We see that being like the Centrino moment for bringing Wi-Fi into the PC two decades ago," said Gelsinger. "We see this AI PC as ushering in a new class of applications, major application upgrades, and ISV excitements. We are leading the way with our Meteor Lake product in the second half of this year. So healthy and great opportunity in the future."

Gelsinger also addressed Intel's position in the datacenter AI market, acknowledging that the company has not been as proactive as it should have been. He recognized Nvidia as a strong competitor that has led the way in this space and currently dominates the arena. Yet, he mentioned that Intel has a robust roadmap for its Gaudi product line, which is gaining market interest. As it turns out, the company already has Gaudi 3 in the lab, which is twice as fast as Gaudi 2.

"Our Gaudi product line really starting to get interest in the marketplace, customer pipeline is a building very rapidly, and we have a very robust roadmap," said Gelsinger. "Gaudi 2 is shipping in volume today. Gaudi 3, I just received the first silicone it. So it is sort of like a 2x level of capability. And, you know, we will be launching [Gaudi 3] for volume shipments next year. And then our 2025 product is already well underway."

He expressed confidence that the company will receive more credit for its AI efforts as they continue to execute its plans. Yet, he refrained from commenting on the prospects of the company's CPUs and compute GPUs in the AI space.

 

Anton Shilov
Contributing Writer

Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.

  • bit_user
    For AI acceleration to really gain strong market penetration and become a "must-have" feature, what you need is for it to be as ubiquitous and seamless as Direct3D acceleration. I'm not sure we're there, yet.
    Reply
  • Jimbojan
    Intel did not make a profit last quarter, it recovered the tax it paid from previous years to make a profit, although Intel did invest many $billions in its factories to cause this loss. By the way, AMD also lose money last quarter, also recovered from previous taxes, but AMD has hardly paid any taxes in the previous years, so it cannot recover much money ($27M). What is interesting is that Intel has caught up with TSMC in manufacturing process; in fact, Inel 4 is ahead of AMD's 5nm processor with TSMC, thus, Intel's Meteor Lake will have a better power efficiency than AMD's chips in additional to its higher performance, which is a first in the last 3 years. This trend will be continuing for the foreseeable future as Intel 3 and 18A will be the leader in the market place; AMD will fall behind again from here on. That is because Intel invested heavily in hardware fab and design, plus the software ( Intel’s OneAPI), no one will be able to compete shortly.
    Reply
  • Eximo
    Jimbojan said:
    Intel did not make a profit last quarter, it recovered the tax it paid from previous years to make a profit, although Intel did invest many $billions in its factories to cause this loss. By the way, AMD also lose money last quarter, also recovered from previous taxes, but AMD has hardly paid any taxes in the previous years, so it cannot recover much money ($27M). What is interesting is that Intel has caught up with TSMC in manufacturing process; in fact, Inel 4 is ahead of AMD's 5nm processor with TSMC, thus, Intel's Meteor Lake will have a better power efficiency than AMD's chips in additional to its higher performance, which is a first in the last 3 years. This trend will be continuing for the foreseeable future as Intel 3 and 18A will be the leader in the market place; AMD will fall behind again from here on. That is because Intel invested heavily in hardware fab and design, plus the software ( Intel’s OneAPI), no one will be able to compete shortly.

    This assumes that everything remains as it is and all projected targets are met for mass production. Would not be the first time that we've been promised a node or release that ends up delayed.

    TSMC is not going to sit around doing nothing either. '3nm' already available to Apple, won't be too long before Apple switches to the next improvement. AMD could even switch to Intel as a fab when their capacity becomes sufficient. Not that Intel would be a huge fan of that, but if they have the capacity and a willing customer...

    All speculation until the products exist for review.
    Reply
  • DaveLTX
    Eximo said:
    This assumes that everything remains as it is and all projected targets are met for mass production. Would not be the first time that we've been promised a node or release that ends up delayed.

    TSMC is not going to sit around doing nothing either. '3nm' already available to Apple, won't be too long before Apple switches to the next improvement. AMD could even switch to Intel as a fab when their capacity becomes sufficient. Not that Intel would be a huge fan of that, but if they have the capacity and a willing customer...

    All speculation until the products exist for review.
    3 things that is counter to that :
    First, IFS is now wholly independent to the design team. Which is why the design team is struggling to get to grips with the new way of doing things. In the past they simply asked production to tweak the process until they wrangle some life into the chips. Therefore the umpteen amounts of node revisions and steppings.

    AMD has a very strong relationship with TSMC for packaging as well, simply switching to Intel is not entirely feasible due to packaging reasons.
    However, SRAM density on Intel 4 is still behind TSMC N5 and will result in larger CPU dies on everything. Especially 3D v cache.
    And a bigger die and lower density also means more latency and less efficiency

    Therefore, claims of higher efficiency is not entirely sound. This was also Intel's old metric boast (SRAM density) when comparing TSMC N7 and what was Intel 10 which they later loosened extremely to make Intel 7 for alder lake
    So in other words, I don't buy it. Samsung also claimed higher density from their nodes but where are the results?
    Given modern CPUs and GPUs are mostly SRAM, Intel 4 doesn't look more efficient, or even area efficient. I suspect that was the main reason behind using "atom class cores" because they're trying to save area but Intel being Intel doesn't think out of the box
    Reply
  • Co BIY
    Eximo said:
    All speculation until the products exist for review.

    True statement.

    And then compared with the contemporaneously available competing product.
    Reply
  • JayNor
    PVC includes 408MB of L2 SRAM . 288MB of it is on the two base tiles.
    Reply