Ex-Apple executives take aim at datacenter processor market

Posted:
in AAPL Investors
A trio of former Apple executives who worked on the iPhone maker's mobile chips have launched their own startup to design processors destined for use in data centers, with the company Nuvia recently raising $53 million in funding.

From left to right: John Bruno, Gerard Williams III, and Manu Gulati
From left to right: John Bruno, Gerard Williams III, and Manu Gulati


The three founders of Nuvia are Gerard Williams III, Manu Gulati, and John Bruno, who all worked for Apple for multiple years. Williams left Apple earlier this year after spending nine years at the company, leaving his position as senior director of platform architecture, and having helped architect Apple's CPU and Systems-on-Chip development for Apple's self-designed A-series processors.

According to Williams' LinkedIn, he was the "Chief Architect for all Apple CPU and SOC development," including leading work on the Cyclone, Typhoon, Twister, Hurricane, Monsoon, and Vortex architectures.

Gulati worked on mobile SoC development for eight years at Apple, before being hired away by Google in 2017. Bruno worked on Apple's platform architecture group for five years after spending time at AMD, before making a similar exit to Google.

Reuters reports the trio are using their backgrounds in mobile chip development and the creation of power-efficient but powerful processors for the iPhone and other Apple products in Nuvia, but for data center usage. By targeting a processor market that typically uses power-hungry chips, the team are hoping their self-designed chip codenamed "Phoenix" will offer performance gains and lower energy usage, as well as more security than current server processors.

"We want to bring all these aspects that we have developed over time through our careers to this new market and really exploit them in this market, because it's an area ripe for innovation and advancement," Williams advised.

The effort puts them against industry giants like AMD and Intel who already make up the majority of server processors used today. A similar concept is also being made like other chip producers, such as Qualcomm and Marvell, who are keen to pivot their knowledge of mobile chip design towards server usage.

So far, the project has caught the attention of major server vendor Dell, who among with a number of other investors has put $53 million in funding into the startup. Dell is a major customer of Intel, so investing in potential alternatives offering power savings could be worth investigating, but the company advised it could not comment on whether Dell would use Nuvia's chips in its servers.

Comments

  • Reply 1 of 15
    thttht Posts: 5,437member
    Surprised that Apple didn’t at least do this in-house (or is taking awhile to do it), with their own custom ARM servers. If they have a 2-to-1 perf/W advantage, with competitive peak performance and Apple has competitive peak performance, in these servers, that’s money saved and probably enough to fund the server development itself for a company with huge server infrastructure needs.
    applesnorangeswatto_cobra
  • Reply 2 of 15
    wizard69wizard69 Posts: 13,377member
    tht said:
    Surprised that Apple didn’t at least do this in-house (or is taking awhile to do it), with their own custom ARM servers. If they have a 2-to-1 perf/W advantage, with competitive peak performance and Apple has competitive peak performance, in these servers, that’s money saved and probably enough to fund the server development itself for a company with huge server infrastructure needs.
     Not to mention that such processors could be used in Macs.   Further Apple could easily sell server chips to others without it being used against them.   Sell a cell phone suitable chip and you undermine your advAntage.  Sell a server based chip and also use it in high end Macs and you actually aid in bringing services to your platform.    The biggest problem with ARM is the lack of hardware I/O standardization, something that Apple could define giving ARM a boost.  

    As for this start up, frankly they are one of many.  I don’t see any of them having the volume to stay in business.  Servers are a small pie and ARM is an even smaller part of the pie.  I’d suggest a two pronged approach that covers servers and volume.  The volume to be achieved with either a mobile solution or an embedded solution.    Embedded actually provides incredible opportunity due to the rapid increase in automation, robotics and self driving vehicles.   

    davebarneswatto_cobra
  • Reply 3 of 15
    rob53rob53 Posts: 3,249member
    Other than the Dell comment, I hope these have a chance being extremely compatible with a Mac environment. It would be nice to have a Mac server again. 
    applesnorangescornchipwatto_cobra
  • Reply 4 of 15
    nchianchia Posts: 124member
    Phoenix - runs so hot eventually it’ll die and turn into ash. Then you buy a new one.  :D
    cy_starkmanwatto_cobra
  • Reply 5 of 15
    jd_in_sbjd_in_sb Posts: 1,600member
    Good luck navigating the gauntlet of processor patents 
    mdriftmeyerjbdragonwatto_cobraeyeakel
  • Reply 6 of 15
    jd_in_sb said:
    Good luck navigating the gauntlet of processor patents 
    Whether PowerPC, x86/x86_64 ISA, or ARM patents already this is DOA.
    williamlondonwatto_cobra
  • Reply 7 of 15
    Rayz2016Rayz2016 Posts: 6,957member
    It seems rare that executives who leave Apple enjoy the same level of success elsewhere, but the processor market needs a shakeup, so good luck to them. 
    williamlondonwatto_cobra
  • Reply 8 of 15
    avon b7avon b7 Posts: 7,664member
    Things seem to be picking up again in the non-intel server space.

    Samsung and Nvidia wound things down a few years ago in the ARM server space but new players have stepped in and it looks like the idea could take things up a notch again.

    My guess is that Nuvia will use ARM based silicon but try to sell it to existing players instead of creating an in-house server solution. That would be biting off more than they could chew. Having Dell financing part of the initial round also points in that direction.

    Taishan/Kunpeng setups and other systems are already using ARM based SoCs and hoping they will get a foothold in this space.

    Whether Nuvia is ARM based or not, it should still lead to a product that serves the same goals if financing is solid.
  • Reply 9 of 15
    Nuvia sounds like a non starter.
  • Reply 10 of 15
    Rayz2016 said:
    It seems rare that executives who leave Apple enjoy the same level of success elsewhere, but the processor market needs a shakeup, so good luck to them. 
    Such is the human spirit, particularly in entrepreneurs. Talent are the crew, optimism is the captain. They’ve been funded enough to pay their salaries so it will be a nice adventure at the least. 
    williamlondonwatto_cobra
  • Reply 11 of 15
    thttht Posts: 5,437member
    Rayz2016 said:
    It seems rare that executives who leave Apple enjoy the same level of success elsewhere, but the processor market needs a shakeup, so good luck to them. 
    Such is the human spirit, particularly in entrepreneurs. Talent are the crew, optimism is the captain. They’ve been funded enough to pay their salaries so it will be a nice adventure at the least. 
    Rayz2016’s observation can be applied to basically every executive who leaves a successful company and ends up not finding success elsewhere. Big success is really rare, hard, and repeat successes are truly rarified air.

    Jim Keller obviously has been successful at moving to different companies and being successful, but even he would find it really hard to start from ground zero. If there was a person who could provide a good competitive chip product from ground zero, it’s him, but that’s like 25% of the business, whatever low number. There’s a huge skill to navigating the business side of things and there isn’t a magic bullet. It seems only the rare individual with the right personality trait like a Jobs or a Musk can do it, and they are [not] successful all the time.

    These people can’t simply build an awesome server solution and buyers will come. They need to produce it on time, actively sell it, lock in big customers, continuous increment their product, and navigate a business field full of bombs.

    Edit: I always forget my nots.
    edited November 2019 watto_cobra
  • Reply 12 of 15
    thttht Posts: 5,437member
    wizard69 said:
    tht said:
    Surprised that Apple didn’t at least do this in-house (or is taking awhile to do it), with their own custom ARM servers. If they have a 2-to-1 perf/W advantage, with competitive peak performance and Apple has competitive peak performance, in these servers, that’s money saved and probably enough to fund the server development itself for a company with huge server infrastructure needs.
     Not to mention that such processors could be used in Macs.   Further Apple could easily sell server chips to others without it being used against them.   Sell a cell phone suitable chip and you undermine your advAntage.  Sell a server based chip and also use it in high end Macs and you actually aid in bringing services to your platform.    The biggest problem with ARM is the lack of hardware I/O standardization, something that Apple could define giving ARM a boost.  
    Apple can just use PCIe 4 for I/O. I think the issue for a new server provider is that there is very little support infrastructure. This makes it very hard for a buyer to make the switch, and Apple isn’t really into this business and I have a hard time believing they can get up for this sort of thing. They would have to support Linux/ARM in a very serious way, and make sure all the tools become available. I suppose they could do DarwinOS, but that’ll be a bigger hill to climb.

    I suppose they can build a merchant support infrastructure for their own data centers, but you begin to see that it is a 10 year time line this way, and they say no to being a merchant ARM provider every time when they see this.

    For their Macs, if it transitions to ARM, I’m still conflicted on whether they’ll go the AMD MCM route or the Intel monolithic route, though everyone is going MCM these days so perhaps moot. It seems they have no problems adding cores, be CPU or GPU or even memory channels, to their iPhone cores for iPad A-X style packages, and they only sell about 10m of those per year. If they are comfortable with that, I don’t see why they wouldn’t just add more cores, more memory channels and PCIe 4 for Macs, in a single monolithic chip and save themselves the integration work.

    wizard69 said:
    As for this start up, frankly they are one of many.  I don’t see any of them having the volume to stay in business.  Servers are a small pie and ARM is an even smaller part of the pie.  I’d suggest a two pronged approach that covers servers and volume.  The volume to be achieved with either a mobile solution or an embedded solution.    Embedded actually provides incredible opportunity due to the rapid increase in automation, robotics and self driving vehicles.   
    They need a big backer. Or many big backers, like Apple, Cloudflare, Facebook, Tencent, etc. If they don’t have big customers that buy their hardware across several cycles of hardware, they can’t fund themselves to iterate and to develop that support infrastructure so that it becomes easier for more customers to buy.

    The only should have one goal. Sounds like it is servers, and it better be at least a 2x advantage in $/hr runtime. They have to race to get the product out, and can’t be dawdling for 5 years. This product needs to come out within 3 years, with iterative improvements every year. Oh, and they need multiple big buyers when the product comes out, so they have to get contracts in by the 2nd year. If they haven’t demonstrated their hardware announced a big buyer within a couple of years, I think they are dead.

    cornchipwatto_cobra
  • Reply 13 of 15
    1st1st Posts: 443member
    NPU? grid layout with limited logic to direct traffic?  Cool (low temp).  Qualbit? that would be even better ;-) and faster not mention lower power... imagination runs wild... Can't wait to see the details.  Best of luck.  watch out $.  50 mil is not big sum, limit FPGA interation and rev.... hopefully, hand route some of it rather than rely on simulated annealing... otherwise, need more dough and time  ;-).  
    watto_cobra
  • Reply 14 of 15
    Quantum computing, not binary computing, is where we need to be investing more. The days of 1's and 0's are done. It's time for a whole new scale! A new approach!

    watto_cobra
  • Reply 15 of 15
    Quantum computing, not binary computing, is where we need to be investing more. The days of 1's and 0's are done. It's time for a whole new scale! A new approach!

    Quantum computing solves a completely different set of problems. Invest needs to be in both, but with Quantum accelerating more.
    watto_cobra
Sign In or Register to comment.