IBM kicked off the new year in dramatic fashion, setting its direction for 1014 and beyond last week with three dramatic announcements, including the commitment of more than $2 billion in funding, and that direction is directly into the heart of the Cloud. It started on January 9th when CEO Virginia M. (Ginni) Rometty personally announced a $1 billion investment in a new company division focused on creating and marketing business products on top of its revolutionary Watson cognitive computing technology. While companies can purchase Watson and run it on IBM servers in house the focus of this new division, headquartered in New York City, is to deliver Watson-based services over IBM’s private cloud infrastructure.

System x

It followed that on Thursday with the announcement of a redesign of System x, its x86 Linux and Windows server line, with a modular design and new models designed for cloud and big data capture and analysis. This is IBM’s first modular-design server family, and the second from a major vendor. While System x is a very different kind of computer from HP’s revolutionary MoonShot, which is designed to bring hyperscale computing to the enterprise, the modular design works in a very similar way. Compute and storage is contained in “Compute Books” that plug into a backplane that provides IO and power. The modular design has distinct advantages for users who can upgrade the unit or replace a failing module simply by unplugging the old Compute Books and plugging in the new ones. IBM estimates that the backplane will be adequate to support three generations of server upgrades, meaning that users can upgrade their System x boxes by plugging in new Compute Books and rebooting, a process that might take an hour or less, without need for the forklift.

Another major advantage of the new design is that it puts large amounts of flash directly in the memory bus, getting the lowest possible latency and highest IO from the NAND flash technology. This is an inherent part of the modular design, since memory and flash are built into the Compute Books with the processor.

One modular computer line is an exciting innovation. Two is a trend. While I have no confirmation from IBM, I would be disappointed if it doesn’t bring modular design to its PureSystems and PowerSystems lines in their next generation. Even System z mainframes may be modularized eventually. And HP, which pioneered modular with Moonshot last April, should bring out modular redesigns of its standard server lines as well. If modular becomes a real trend it will mean a virtual end to forklift upgrades, and to all the expense and disruption they cause in the data center. Today this is an important market differentiator. By next year that will become table stakes that buyers will list as basic requirements for bidding on new server contracts.

IBM Cloud

IBM finished off Friday morning by announcing a $1.2 billion commitment to expanding its private worldwide cloud network. The plan includes building 15 new data centers and doubling the capacity of its SoftLayer data network, which is quickly becoming the platform on which IBM delivers all its cloud services to customers, including its growing portfolio of SaaS services, the new services it is building on top of Watson, and its PaaS and IaaS services. This really is the largest of the three announcements in part because that $1.2 billion is all budgeted to be spent this year, with further expansion, including data centers in Africa, tentatively planned for 2015. The Watson investment is a multi-year investment. The combination will put IBM into a leadership position in the cloud service market.

Clearly Rometty and IBM’s board sees the company’s future in the cloud. The big question, of course, is whether it can successfully challenge Amazon AWS, which continues to grow at an astounding rate. However AWS is definitely vulnerable particularly in the large-to enterprise size market due in part to its highly centralized architecture and dependence on the public Internet. For instance, AWS cannot guarantee the physical location of sensitive data, making it impossible to meet requirements in many countries that some types of data remain inside that country’s borders. It also makes it difficult to guarantee compliance with regulations governing sensitive data, such as personally identifiable medical information, that vary from country to country. IBM’s decentralized network of 25 data centers — scheduled to grow to 40 this year under last week’s announcement — can and does guarantee full auditability of its customers’ regulated data.

AWS so far has been unable to guarantee SLAs, which obviously disqualifies it from even bidding of providing most core services to IBM’s target markets. IBM can deliver on SLAs reliably, in part because of its high-speed private global network and its distributed data center network that provides enterprise-level backup for fast recovery of a disruption in any one physical location. The SoftLayer platform also provides higher data security.

This does not mean that AWS will see any diminution of its present huge growth rate, or that other service providers, including HP and the large colocators and outsourcers, all of whom are also players in this market, are going to be locked out. Indeed it is unclear whether AWS could handle an additional major increase in its business growth. However, as companies and government agencies move from trial and development to production use of cloud services, IBM is positioned for increasingly rapid growth in its target markets, which are largely different from the AWS sweet spot of startup entrepreneurial companies. It is hard to imagine that a mid-sized company would move its ERP system to AWS, whereas even a global manufacturer could consider putting its ERP in the IBM cloud, if not this year then by 2016.

The end of hardware

For IBM this really is a make-or-break strategy. For nearly two decades the corporate strategy has been to reinvent the company as a services and software vendor. While it clearly is highly profitable today, witness its multiple billion dollar commitments including the $1 billion it committed to Linux last fall, it has seen six consecutive disappointing quarters. Sales of its hardware, and its high-end Power servers in particular, have plummeted. And industry rumor is predicting that it is on the verge of selling off its server business entirely. That is a breathtaking possibility for a century-old hardware company, but today services and software provide the majority of its business. Watson, which is by all indications the harbinger of a new generation of computing, the first since the IBM 360 mainframe introduced programmable computing and made tabulating machine computing obsolete 50 years ago, is software, not hardware.

Indeed IBM may have signaled the end of its hardware business last August when it announced the Power Consortium and offered to provide the full specs of its Power RISC chip to consortium members so they could develop and, if they wished, manufacture their own versions of Power. Given that Watson is written to run on Power, this could be IBM’s way to ensure that it will have access to the hardware it needs when it stops making its own servers.

None of this will happen next month. But in a world where companies tend to focus on this quarter and worry about the next when it comes, IBM often takes the long view. Its evolution from a hardware vendor to a software and services provider has been going on for more than 15 years. And its sale of its PC division to Lenovo shows that it is capable of transformative actions. Certainly the day is coming when it will have to decide between remaining anchored to an increasingly commoditized and ultimately shrinking hardware business or cut the apron strings and become a cloud entity. Clearly its market strategy is preparing it for that day, and Rometty has the courage to cut it free.

Graphics courtesy IBM Corp.