Dell AI server revenues leap however storage waits on Mission Lightning
Dell’s quarterly outcomes present an enormous development in server gross sales, pushed by synthetic intelligence (AI) initiatives, however a relative lag in storage. Key causes behind that could be that Dell’s present storage lags somewhat behind the curve in AI efficiency, whereas its huge parallel network-attached storage (NAS) that goals to plug that hole, Mission Lightning, is in gestation.
Dell’s PC division normally massively outsells its datacentre merchandise, however that’s not the case in its newest (second) quarterly outcomes, which present 69% development in gross sales of servers and networking tools year-on-year. That equated to a income of $16.8bn for the infrastructure division that put the consumer providers – i.e. private tools – into the shade with $12.5bn of gross sales.
Dell has benefited right here from being the primary to the AI market, with servers, the most recent Nvidia graphics processing unit (GPUs), and switches appropriate with excessive throughput Nvidia Spectrum-X networking and Extremely-Ethernet playing cards.
“Within the final six months we have now delivered $10bn value of servers for AI,” mentioned Jeff Clarke, vice-chairman and chief working officer for Dell Applied sciences. “That’s greater than was attained in the entire earlier yr. Demand is robust and gross sales of the brand new AI {hardware} has totalled $20bn for the yr.”
General, Dell’s Q2 outcomes confirmed file income of $29.8bn, which was up 19% on the earlier yr. Of the $16.8bn of income contributed by the infrastructure group – development of 44% in a yr – servers and community tools contributed $12.9bn.
In the meantime, nevertheless, storage arrays – flash and disk – noticed revenues decrease by 3% over the yr at $3.9bn within the quarter.
In the meantime, the consumer providers group’s income development was a mere 1% year-on-year, with enterprise PCs reporting $10.8bn income (+2%) and client merchandise $1.7bn (-7%).
Storage the poor relation in infrastructure gross sales
A salient characteristic of those outcomes is that the calls for of AI appear to favour compute {hardware} greater than storage.
That could be confirmed by the most recent outcomes from NetApp, which is primary in flash storage arrays, in accordance with IDC. Right here, the array maker posted quarterly outcomes of $1.56bn in August, which equated to annual development of 1%.
In the meantime, Pure Storage introduced income of $861m, and that was a rise of 13% on gross sales in a yr – however there’s a catch. That set of figures included its supply – unprecedented – of SSD DirectFlash Modules (DFM) to hyperscaler Meta.
Pure’s DFMs are a proprietary format through which the seller has packed a a lot larger density of storage onto SSD playing cards. That’s as a result of it offloads loads of on-board cache to the array and handles knowledge there as an alternative.
HPE’s third quarter outcomes confirmed income development for servers at $4.9bn, up 16% year-on-year, however doesn’t seem to interrupt out storage income.
In the direction of evolution in storage
Why have we seen a lift in income for servers for AI, however probably not with storage?
There’s little doubt from a technical standpoint that storage is a vital assist for compute for AI; it’s attainable that enterprises have staged their funds spend and centered first on processing energy.
On the similar time, it’s true that storage merchandise have lagged behind by way of efficiency in contrast with compute. For instance, servers that feed GPUs are in a position to transfer knowledge at a fee of 400Gbps and even 800Gbps. Present storage merchandise supply round 100Gbps.
Storage suppliers have, nevertheless, centred efforts to develop AI storage round parallel file system storage for AI.
Huge Knowledge led the way in which right here, with huge parallel entry to storage, whereas Hammerspace and Weka additionally adopted.
Dell responded with Mission Lightning – which contains Powerscale, the rebranded Isilon scale-out NAS – however that doesn’t appear to have a launch date but. In the meantime, NetApp has Ontap Knowledge Platform for AI, whereas Pure has FlashBlade//Exa.
