Two HPs, One Dream

As the company gets ready to split, its future rests on “the Machine”
Photo Illustration: 731; Photographer: Plush Studios/Getty Images

Last year, Hewlett-Packard announced it was working on a new type of computer designed to be superfast, low-power, and capable of replacing a data center’s worth of equipment with one fridge-size box. Dubbed the Machine, the project would combine advanced research from throughout HP’s divisions, the company said, and if each hit its ambitious goals, the Machine could essentially replace today’s standard computer designs. The job sounded tough even before HP said a few months later that it planned to split down the middle, creating one company that sells business hardware and software while the other focuses on consumer PCs and printers.

The Machine is supposed to be ready in 2017 at the earliest, and HP has set a Nov. 1 target for the split. So Chief Technology Officer Martin Fink has drawn up a deal between the two companies that will allow the engineers working on the project to continue collaborating, even though they’ll be working for separate entities. At the same time, the engineers have had to rethink some pieces of the project. Instead of building a brand-new operating system for the Machine right away, they’re first working to customize the free Linux OS and see how far it can take them, Fink says. When the first prototype is completed next year, “there’ll be some ugly spots,” says Kirk Bresniker, chief architect at HP Labs, the company’s research and development arm, which is leading work on the Machine. “I just want something that you can demonstrate and prove is on the right track.”

In November, HP will become Hewlett Packard Enterprise (selling business IT) and HP Inc. (PCs and printers). Fink, who will be CTO at HP Enterprise, says the Machine has become “a daily point of conversation” with customers and among fellow executives, and HP Inc. engineers will keep working with his team. With the Machine, “I’m desperately hopeful we’ll find that most of everything we do today will get done better and simpler, and we’ll be able to do things we can’t conceive of,” says Fink. HP wouldn’t comment on the financial terms of the collaboration agreement or the Machine’s funding but says it is and will remain HP Labs’ biggest research project.

Photonic cables could transfer data far faster than copper wires.
Photonic cables could transfer data far faster than copper wires.
Source: HP

Plans for the Machine require HP to develop several technologies in tandem: a nanoscale, low-power memory chip called a memristor that’s supposed to vastly increase data storage; a data-transfer system that’s much faster than current models because it relies on lasers instead of copper wiring; and software that can put the new components to work. In an early chat about the Machine, Fink says, one client reacted as if he were talking about curing cancer. “That was scary to me,” he says. “I don’t want that on my shoulders.”

If all of those parts come together, the Machine would be able to reinvent the building blocks of every computer on earth, from PCs to smartphones, says Crawford Del Prete, chief research officer for market researcher IDC. Competitors have yet to announce challenges to HP’s moonshot. “We’re involved in some very advanced, very low-level silicon-level engineering, and you’re going to hear about it when it’s ready to ship,” says Oracle’s senior vice president for systems technology, Juan Loaiza. “We don’t need to talk about stuff five years from now.”

Tiny, low-power memory chips could store way more data than current models.
Tiny, low-power memory chips could store way more data than current models.
Source: HP

Over the past few months, to get developers used to programming on something like the Machine, HP has begun releasing hardware and software designed to emulate it. That includes a high-capacity version of HP’s Superdome servers and a programming language called Distributed R. South Korean chipmaker SK Hynix has signed a deal with HP to manufacture its memristors when the design is complete. (Fink says he’s discussing timing for a “test vehicle.”) But HP has been hyping memristors for years without managing to produce them en masse, says IDC’s Del Prete. “It’s not just demonstrating that a memristor works but that a memristor will be manufactured,” he says. “I’m not aware of anybody with this aggressive of a time frame.”

The researchers developing the Machine recognize they have one shot at it, says Rich Friedrich, HP Labs’ director for the Machine’s systems software. “We can’t afford either in time or in dollars or in engineering costs, once we evolve something, to throw everything out and start again,” he says. Whether one company or two, HP also needs to think about the reputational costs of failure, says Jim Handy, an analyst with researcher Objective Analysis. There’s a reason no other company has been so public about trying to do all the things a project like the Machine requires at the same time. Handy says: “If any one component is missing, then it’s as good as not having any of them.”

The bottom line: With HP’s split fast approaching, its engineers are figuring out how to keep working together on its biggest project.

(Updated third paragraph to clarify the distinctions between the forthcoming product lines of Hewlett Packard Enterprise and HP Inc.)

    Before it's here, it's on the Bloomberg Terminal.