Antonio Neri (CEO - Hewlett Packard Enterprise)
As the dust settles on the industry’s most recent data-driven acquisition, is the coming together of Hewlett Packard Enterprise (HPE) and Cray a match made in supercomputing heaven?
The official line naturally paints a rosy picture; the creation of the “most comprehensive end-to-end portfolio” in the market, spanning compute, storage, interconnect, software and services.
Billed as having “significant economic upside”, the $1.3 billion deal is in response to an “explosion of data”, an explosion which continues to drive strong high performance computing (HPC) growth.
With the ink barely dry on yet another hardware-centric acquisition, will HPE’s gamble pay off?
“A definition of insanity is to engage in the same activity over and over again while expecting a different outcome,” observed Geoff Woollacott, principal analyst at Technology Business Research. “HPE’s history has been to acquire struggling firms in niche hardware areas in hopes of growing share.
“With fewer and fewer silicon-centric vendors left standing, the odds of success can certainly increase in time.”
As reported by Channel Asia, terms of the agreement will see future HPC-as-a-service offerings, combined with artificial intelligence and machine learning analytics delivered through HPE GreenLake.
Founded in 1972, Seattle-based Cray houses approximately 1,300 employees worldwide, and delivered revenue of $456 million in its most recent fiscal year, up 16 per cent year-over-year.
“We know Moore’s law drives consolidation in the industry,” Woollacott added. “What we do not know, however, is if any two hardware-centric vendors can come together and build a business accretive to the top line.
“Michael Blumenthal tried this strategy by combining Burroughs and Sperry to create Unisys, and that certainly did not work. More recently Dell acquired EMC, and while jury remains out on that consolidation play, early indications have been positive.”
As explained by Woollacott, HPE has deployed such a strategy “multiple times” over the years.
Specifically, similar hardware-centric deals the vendor has conducted over the years include acquiring Apollo after its first-mover advantage in engineering workstations was eclipsed by Sun Microsystems.
“Acquiring Compaq after it had acquired Tandem and Digital Equipment Corporation (DEC), which had likewise struggled as much in business model integration as with technology integration,” Woollacott summarised.
“Acquiring SGI, which was hemorrhaging cash but was a strategic HPE OEM partner that HPE could not afford to let fail or be acquired by a rival. And now Cray, the last of the venerable high-end niche vendors to double down on higher-margin HPC.”
Mainstream business
In assessing the wider market potential, Woollacott acknowledged that HPC “certainly has growing appeal”, which stems from several economic drivers.
“As always, Moore’s law theory gets borne out in reality as cost and form factors decrease to the point where distributed computing (a fundamental tenet of Ken Olsen’s original business plan for DEC in the early 1960s) can be done at the board level if not the chip level,” Woollacott added.
“Graphics processing units (GPUs), tensor processing units (TPUs) and field-programmable gate arrays (FPGAs) can keep pace with increasing demands coming from big data analytics.”
In addition, Woollacott said supply chain excellence and software tuning of these commodity components can allow for custom-designed systems, purpose-built to the compute demands of the HPC customers.
“IBM certainly keeps innovating in HPC, especially with its RISC-based power chips suited for analytics,” Woollacott observed. “Lenovo has taken a huge bite out of HPE’s share of the HPC space through its design engineering and supply chain flexibility, manufacturing commodity Intel boards at scale through Lenovo’s global manufacturing space.
“Per Lenovo it went from having none of the top 500 HPC installations in the world in 2014 to having 140 of them in 17 countries in 2018. Much of this success came at HPE’s expense.”
Success or failure?
For Woollacott, the Cray acquisition “may well aid” HPE in stalling Lenovo’s recent successes in the HPC space, but Lenovo’s operating best practices are "well suited to commoditising markets".
“Supply chain excellence honed to attack the hyper-scale market brings decided cost advantage to the HPC space,” he said. “Talent recruited from Intel and other firms likewise gives Lenovo the software tuning competencies necessary to extract fit-for-purpose performance from commodity chipsets.”
According to Woollacott, quantum also “looms large” on the horizon as the next chapter for the high-end compute requirements to help solve the world’s intractable problems.
“Seven nanometer wafers may not be the end of the line for silicon innovation, but it is certainly getting close,” Woollacott explained.
“This acquisition seems poised to satisfy the immediate here and now, while once again being eclipsed by niche innovation elsewhere, with that elsewhere coming in the quantum domain in three to five years.”
Woollacott alluded to recent articles in the media, which suggested that HPE is cutting back on quantum research, intending instead to extract more life out of the traditional computing space with processors for deep learning and analytics.
“HPE has certainly acquired a company that has been admired for decades as being the ‘tip of the spear’ in silicon innovation,” Woollacott added. “HPC innovations certainly can work for today, but that tip of the spear will be blunted by the inexorable laws of physics, making further silicon innovation increasingly more challenging.”
In looking ahead, Woollacott said future offerings in what has been Cray’s core market will come from quantum innovators.
“Once quantum reaches economic advantage over high-end classical computing, the industry will see yet another round of business exits for those vendors lacking transformation fearlessness,” Woollacott concluded.
“Like many of HPE’s other hardware-centric acquisitions, this move appears to have reasonable short-term impact and limited long-term upside.”
