Las Vegas is a city that loves to put on a show. Its streets are filled with billboards and fliers – they even cover entire sides of buildings – advertising one spectacle or another.
The latest addition to the already comprehensive list of event venues is the Sphere. Since it opened in September 2023, it has hosted performances from U2, Phish, Dead & Company and, most recently, HPE CEO Antonio Neri
From 7.00am on the first day of HPE Discover 2024, delegates dutifully made their way through assigned lines, submitted themselves to airport-style security checks, and wove past various gimmicks to take their seats in the capacious auditorium.
There, they waited patiently to see the “first-ever keynote at Sphere” (a fact that was signposted almost everywhere ahead of Neri’s talk).
Their perseverance was rewarded. Irrespective of any announcements, the Sphere is an impressive venue that wowed even the most cynical attendees. This, I imagine, was the point. A striking show like this generates anticipation and warm feelings for the main announcement: Nvidia AI computing by HPE.
This, in short, brings integrated HPE and Nvidia hardware optimized for AI to mid-market and enterprise. As ever, it’s available through HPE GreenLake, although it strikes me that the four available builds – small, medium, large and extra large – don’t quite fit into the ‘consumption-based’ model that GreenLake is predicated on.
Speaking of GreenLake…
Despite all the talk of generative AI, a scale model of the Venado supercomputer, and boasts of the company’s high performance computing (HPC) prowess, last year’s big announcement – HPE GreenLake for LLMs – was conspicuous by its absence.
As we left things last year, there was a waiting list to put in orders, but extra capacity was expected come the end of 2023 in North America. Availability in Europe was supposed to follow in 2024, but as far as I can tell it’s yet to leave the pilot stage.
Arguably, there’s more money to be made from the familiar hardware emanating from this latest partnership with Nvidia. GreenLake for LLMs is not quite esoteric, but feels less grounded in businesses’ needs as they stand today than a box they can put in their data center.
On premises is back in fashion
Another notable element of Neri’s key messaging this year was the emphasis on delivering services on premises.
Now, let’s be clear about one thing. GreenLake has been pitched as a cloud product, but this is only strictly true in the ‘consumption-based pricing’ element. The hardware is still delivered on premises or potentially via a managed services provider (MSP) in a colocation facility – HPE is in the hardware business, not the hyperscale cloud business.
The reality of this was in evidence when Neri said AI has given a “triple shot in the arm to the server business”.
He claimed that the nature of the data used by businesses’ own generative AI tools will make them reconsider where it’s stored, particularly from a compliance and data sovereignty perspective. This, the company believes, will lead to a reticence to put data into the public cloud and thus generate more on-premises infrastructure sales.
There’s clearly a “well they would say that, wouldn’t they” element at play here. It benefits HPE – and its competitors – if an explosion in custom generative AI tools leads to a surge in the enterprise server market. Whether it plays out like that in reality is another matter.
Sustainability, maybe
Sustainability is a watchword for most vendors nowadays and HPE is always more than happy to talk up its credentials here. Delegates were reminded that seven of the 10 most energy-efficient supercomputers on the global Green500 list were created by HPE.
HPE executives were regularly telling everyone how efficient direct water cooling is, and how waste heat from the cooling process can be used to heat buildings. In the case of supercomputers, this can even include defrosting pavements in the town where it’s housed.
This is largely the same story as was told last year, when I wrote about the company’s plan to combat generative AI’s dirty secret. It sounds great – but it only goes so far.
What happens to the waste heat in summer, for example? Or if the winters are relatively mild, as with its supercomputer in Bristol, South West England, or Oak Ridge National Laboratory, Tennessee. Additionally, as more generative AI hardware comes online it ultimately doesn’t matter how efficient it is as the power consumption will increase or, at best, remain static.
This problem isn’t unique to HPE – it’s an acknowledged issue with hyperscale cloud or social media, for example – but it does need solving if we’re to match technology ambitions with climate imperatives.
Come for the big screen, stay for the gimmicks
On a lighter note, it wouldn’t be a trade show if there weren’t some gimmicks on offer and Discover was no different. With AI inevitably comes robots and avatars, both of which were on display at the event.
Once delegates had made it past the metal detectors and bag search into the Sphere’s atrium we were greeted by a wise-cracking female robot. People were invited to come forward to ask questions into a microphone, which the android was able to answer with some top-tier dad jokes courtesy of natural language processing (NLP).
If you missed your opportunity with ‘her’, there was an even more meta opportunity to experience basically the same thing – Antonio Nearly. This replica of HPE’s CEO, which moved rather like an NPC from a video game, was similarly able to provide answers to spoken questions with a convincing simulation of the real Neri’s voice.
Attendees were encouraged to ask ‘him’ about anything, whether HPE-related or not. From my observations, it seemed most people were opting for “not”, often treating the avatar like a high-tech soothsayer.
“Who will win the World Series?” One person asked. ‘Nearly’, presumably sticking firmly within his guardrails, demurred and instead offered a response that his human counterpart prefers football (soccer, rather than American) and is a supporter of the Argentinian national team.
“It’s true!” the real Neri told me when I relayed this story, having confessed I hadn’t interacted with it myself. Not an oracle, then, but given it also referenced the outcome of the 2022 FIFA Football World Cup final, certainly an avatar up to date with its sports references.
As with Discover 2023, generative AI has once again been center stage at Discover 2024 and – playing the clairvoyant myself – I don’t see that changing in 2025. Generative AI is a juggernaut that shows no sign of slowing down or waning in popularity.
Whether we see anything more of GreenLake for LLMs is a big question mark. It was completely absent from the discussion and representatives from the company were very reticent to answer any questions about it; the most I got in terms of a response was “we’re working to refine the offering and we’ll share details later this year". I’m sure I can’t be the only attendee whose mind turns to the ill-fated The Machine in-memory computing initiative that was once front and center in former CEO Meg Whitman’s strategy, but quietly set aside in 2018.
How Nvidia AI Computing by HPE unfurls and more importantly, how enthusiastically customers invest in this new hardware will be the real measure of the company’s success in attaching itself to the enthusiasm for this new technology.
