Yesterday during Marvell’s quarterly earnings call, the company had made a surprise announcement that they are planning to restructure their server processor development team towards fully custom solutions, abandoning plans for “off-the-shelf” product designs.

The relevant earning call statements are as follows:

"Very much aligned with our growing emphasis on custom solutions, we are evolving our ARM-based server processor efforts toward a custom engagement model.

[…]

Having worked with them for multiple generations, it has become apparent that the long-term opportunity is for ARM server processors customized to their specific use cases rather than the standard off-the-shelf products. The power of the ARM architecture has always been in its ability to be integrated into highly customized designs optimized for specific use cases, and we see hyperscale data center applications is no different. With our breadth of processor know-how and now our custom ASIC capability, Marvell is uniquely positioned to address this opportunity. The significant amount of unique ARM server processor IP and technology we have developed over the last few years is ideal to create the custom processors hyperscalers are requesting.

Therefore, we have decided to target future investments in the ARM server market exclusively on custom solutions. The business model will be similar to our ASIC and custom programs where customers contribute engineering and mask expenses through NRE for us to develop and produce products specifically for them. We believe that this is the best way for us to continue to drive the growing adoption of ARM-based compute within the server market."

We’ve had the opportunity to make a follow-up call with the teams at Marvell to get a little more background on the reasoning for such a move, given that only 6 months ago during the launch of the ThunderX3, the company had stated they were planning to ship products by the end of this year.

Effectively, as we’ve come to understand it, is that Marvell views the Arm server market at this moment in time to be purely concentrated around the big hyperscaler customers which have specific requirements in terms of their workloads, which require specific architecture optimisations.

Marvell sees the market beyond these hyperscaler customers to not be significant enough to be of sufficient value to engage in, and thus the company prefers to refocus their efforts in towards closer collaborations with hyperscaler costumers and fulfilling their needs.

The statement paints a relatively bleak view of the open Arm server market right now; in Marvell’s words, they do not rule out off-the-shelf products and designs in several years’ time when and if Arm servers become ubiquitous, but that currently is not the best financial strategy for the current ecoysystem. That’s quite a harsh view of the market and puts into question the ambitions of other Arm server vendors such as Ampere.

The company seemed very upbeat about the custom semicon design business, and they’re seemingly seeing large amounts of interest in the latest generation 5nm custom solutions they’re able to offer.

The company stated during the earning call that it still plans to ship the ThunderX3 by the end of this year, however it will only be available through customer specific engangements. Beyond that, it’s looking for custom opportunities for their hyperscaler customers.

That also means we won't be seeing public availability for the dual-die TX3 product, and neither the in-design TX4 chip, which is unfortunate given that the company had presented their chip roadmap through 2022 only a few weeks ago at HotChips, which is now out of date / defunkt.

Although the company states that it’ll continue to leverage its custom IP for the future, I do wonder if the move has anything to do with Arm’s recent rise in the datacentre, and their very competitive Neoverse CPU microarchitectures and custom interconnects, essentially allowing anybody to design highly customizable products in-house, creating significant competition in the market.

From Marvell’s perspective, this all seems to make perfect sense as the company is simply readjusting towards where the money and maximum revenue growth opportunities lie. Having a hyperscaler win and keeping it is already a significant pie of the total market, and I think that’s what Marvell’s goal is here in the next several years.

Related Reading:

POST A COMMENT

41 Comments

View All Comments

  • webdoctors - Friday, August 28, 2020 - link

    Makes sense, when there's like 5 customers that dominate the entire dataserver market (Alibaba, Google, Amazon, Facebook, Microsoft), you might as well cater to them.

    It would be good to add some numbers in this article. Like I think Amazon's internal networking company produces enough network HW for Amazon that if they were a regular networks company they'd be like 4th biggest in the world. Maybe a similar stat if they sold HDDs. That's how big these customers are.
    Reply
  • Industry_veteran - Saturday, August 29, 2020 - link

    Actually it makes no sense. All the big or hyper scale customers today have the money and technological reach to hire their own team to design and manufacture CPU as per their own requirement. Why would they need Marvel? Besides that this team that Marvel has is essentially not a team experienced in server processors. They are basically an embedded CPU team that was inflated to make servers. They never understood the demands of server market and what it takes to knock off intel. Reply
  • pifourth - Monday, August 31, 2020 - link

    We'll see how Marvell(Cavium) does with hyperscaler customers.

    Marvell (Cavium) does come from a background in multi-core ARM processors and high speed crypto operations that goes back many years. Marvell (Cavium) might see some business from
    the combination of ARM server and high speed crypto of various sorts ( iPsec, Sha, Public Key
    Crypto, PKI, IKE, AES, Crypto that is resistant to Quantum Computer attack, etc. ). Time will tell how much business it gets.

    Tom

    If Intel charges quite a bit extra for server chips that also do high speed crypto operations, Marvell (Cavium) might get some sales on a much cheaper ARM server that does high speed crypto operations. On the other hand, if Intel doesn't charge much extra for
    server chips that do high speed crypto operations, then Marvell (Cavium) may not see much
    business for customized server chips. We'll see what develops.
    Reply
  • ksec - Saturday, August 29, 2020 - link

    Well HyperScalers ( The Big 5 you mentioned ) only represent half of the market. There are still many smaller players that make up the other 50%.

    But again you need HyperScaler to kick start those volume before moving on to more general market. So if they do succeed I am sure they will come back to with off the shelf solutions.
    Reply
  • rahvin - Saturday, August 29, 2020 - link

    No it would make sense because like Thunder I and Thunder II, they got Thunder III to actual production silicon and realized it wasn't competitive just like every other ARM server design.

    It would seem the ARM server future resets on Amazon's shoulders alone.
    Reply
  • Wilco1 - Sunday, August 30, 2020 - link

    Arm servers have been competitive at the time they were released. TX2 looks great compared with Skylake: https://uob-hpc.github.io/assets/cug-2018.pdf , and Cloudflare wanted to move to Centriq because it beat their x86 servers on every aspect (it looks they will use Ampere Altra now).

    No, Graviton is definitely not alone. Ampere Altra already beats the fastest EPYC server, there is Nuvia and several other companies designing Arm servers. Arm's next generation Neoverse will obviously be even faster and more efficient, so the future looks very promising.
    Reply
  • Industry_veteran - Friday, August 28, 2020 - link

    So after spending hundreds of millions of dollars on false promises they are finally admitting that ARM server market was never there. This is no surprise. The Broadcom leadership had already reached this conclusion in 2016.
    This Marvell team is actually a team that spend close to 500 million dollars in Broadcom and never came up with working CPU despite spending so much cash for 4 years. Finally Broadcom leadership got fed up of this team and offloaded them.
    To my surprise, they found employment in Cavium and spent hundreds of millions more in both Cavium and then Marvell before reaching the conclusion that there is no demand for ARM server in general purpose server market. The ARM technology is certainly not good enough to abandon Intel.
    Makes me wonder how decisions are made in hi-tech companies!. People without commonsense are reaching the top of corporate ladder due to inherent buddy system in Silicon Valley.
    Question is why SEC doesn't investigate these people who supposedly spend hundreds of millions of dollars on technology that never had much promise to begin with.
    Reply
  • FunBunny2 - Saturday, August 29, 2020 - link

    "People without commonsense are reaching the top of corporate ladder due to inherent buddy system in Silicon Valley."

    not specific to SV. it's the decade long invasion of the MBA. the MBA 'knows' that management is a separable skill, which skill is only truly possessed by the MBA, and such a skill that can be applied by any MBA to any industry. that's what has killed American industry, not China.
    Reply
  • Industry_veteran - Saturday, August 29, 2020 - link

    While I agree with you about MBAs, the buddy system is not just limited to MBAs and people without common sense are not just MBAs but even engineers.
    The chief architect of the team in Broadcom knew fully well by 2016 that ARM server market today is all about getting hyper scale customers. That is the time Broadcom shutdown the ARM server project and let the team go. The same team then went to Cavium and then Cavium was bought by Marvell touting the promise of ARM server market. So how can anyone say that not knowing the realities of ARM server market is only the fault of MBAs?
    Reply
  • TomWomack - Saturday, August 29, 2020 - link

    I don't believe it was obvious in 2016 or 2017 that the ARM server market was only hyperscalers; certainly at ARM we were really hoping for somebody to produce motherboards that could go in a standard Supermicro 1U case and be bought by anyone who might have bought a Supermicro Xeon E5 machine.

    But ARM wasn't willing to compete with its customers by providing a reference model, in the way that every Intel server outside the hyperscalers uses an Intel chipset on an Intel motherboard, and ARM didn't have enough pull to get the customers to be broadly compatible with one another - being the first to do the R&D to produce a completely commoditised system is difficult to sell to a company's board.
    Reply

Log in

Don't have an account? Sign up now