Sunday, August 10, 2014

HP's Machine and IBM's $3bn R&D splash – aka how to survive Google


Tough roads ahead for IBM and HP. Aivars Lode avantce

HP's Machine and IBM's $3bn R&D splash – aka how to survive Google


Failure to crack next-gen semiconductors threatens to set back humanity
By | Jack Clark 23rd July 2014 11:02

IBM and HP are trying to invent their way out of severe problems that lurk in their future.
Both companies face slowing sales in some of their traditional high-margin areas, like enterprise servers and other on-premises hardware. And both are having a tough time converting existing business divisions over to selling the high-margin "as-a-service" style IT products that many companies are demanding.

Meanwhile, the rise of remote computing and storage from companies like Amazon and Google, and remotely maintained software from companies like Salesforce and Box and even Oracle, stand to reduce the need for the types of on-premises gear and lucrative service contracts they specialize in.
Both companies suffer from problems partly of their own making. IBM's previous chief Sam Palmisano pledged in 2010 that per-share earnings for the company would reach $20 by 2015. This has forced today's chief Virginia Rometty to fire people, sell businesses – like the server division to Lenovo – and contort the company to trim costs where possible.
HP, similarly, is grappling with problems from previous chiefs, whether that be Mark Hurd's cost-cutting to its R&D division and buy of EDS for $13.9bn (in 2012, HP wrote down the value of that purchase by $8bn), or Hurd's successor Leo Apotheker whose public dithering over HP's PC plans sapped the company's share price, and whose acquisition of Autonomy for $10.7bn in 2011 (which was subsequently written down by $8.8bn) drove shares even lower. This period of poor management has led to a spate of severe layoffs at the company, amid slowing growth of the overall PC market and greater competition from servers.
And unlike smaller rival Dell, neither company can easily take itself private to perform the necessary surgery to convert itself to a firm fit for a modern IT era.
To try to offset their declines, both IBM and HP have tried to build cloud businesses using open-source technologies like OpenStack and a multitude of data centers around the world. IBM also bought SoftLayer for about $2bn to bolster its cloud.
But cloud is unlikely to create significant growth to offset the declines in traditional businesses – Amazon Web Services is acknowledged to be the leader in cloud services and is projected to bring in revenue of between $4bn and $5bn this year for the company. That's good money, but nothing compared to the annual revenues of HP (2013: $112bn) and IBM (2013: $99bn).
What both IBM and HP need is a new set of products that are desirable, pricey, engender customer loyalty, and difficult to replicate by hungry Asian-based hardware competitors, and – ideally – things that have better profit margins than traditional commodity hardware.
So both companies have pressed their significant research organizations into developing some tech to give them some hardware with these characteristics – an ambitious, eye-brow raising strategy at a time when the rest of the industry is focusing on software.
The two schemes are HP's memristor-dependent R&D scheme, which was codenamed "The Machine" and announced at HP Discover in June, and IBM's unnamed $3bn-over-five-years scheme which was revealed in July.
Though subtly different in emphasis and implementation, both projects serve the same purpose and speak to the same problems bedeviling the companies.
The Machine will give HP a set of products and services built around a proprietary hardware platform made up of low-power storage based on memristors; custom processors packaged up in novel ways; speedy data transfer via the clever use of photonic comms tech and custom interconnects; and a new operating system designed to handle the properties of memristors.
If successful, The Machine will give HP a high-margin way to make money and take a lead on its competitors. However, it's based on some difficult technologies, like the memristor that was meant to come out in 2013but is now not forecast to be ready for sale till 2016. It also uses other long-running HP projects, like a plan to mush processors and storage together into a "nanostore". Like the memristor, this tech has been around for a while – an early paper on it was published in 2011 – and its eventual due date is unknown.
IBM's cash splash, by comparison, will fund the shrinking of chips beyond the 7nm process scale, and will also bankroll the design of chips that feature unusual architectures. These could include "neuromorphic" processors that try to implement brain-like arrangements of semiconductor logic to create chips that are good at pattern and object recognition. Like HP, IBM plans to use photonics to let it "connect processors and memory in a more integrated way."
If the schemes work, then HP and IBM will be able to survive and prosper. If they fail, then both companies will have to divest themselves of multiple business units, layoff many of their employees and, perhaps, cede much of their business to new IT firms like Amazon or Google.
Worse, neither Amazon or Google or even Microsoft appear to do much fundamental hardware and materials science research, so for IBM or HP to stumble would likely set the wider technology industry back by decades. Their plans may be mad, but they're the best hope they – and by extension we, the wider tech community – have for ongoing gains in performance of the basic components of our digital lives. 

No comments:

Post a Comment