Mad Dog 21/21: Jumping The Shark
January 16, 2012 Hesh Wiener
IBM just finished a banner year. It recorded the highest revenue in its history and impressive profit. But if you depend on IBM’s proprietary systems, the IBM i and the System z platforms, don’t let the headlines fool you. You (and IBM) have a strategic problem. When it comes to computer hardware, IBM jumped the shark years ago. Jumped the shark? When Fonzie jumped the shark, five years into Happy Days, it marked a zenith, an irreversible turning point although the show lasted another five years. IBM jumped the shark in 1981, as it introduced the PC.
In 1981, the company’s total revenue was $29 billion, less than a third of its current intake. Frank Cary was chairman, having just handed over his job as CEO because he had turned 60. The transition put John Opel at the executive helm, where he could, confidently and without hesitation, steer Big Blue into the PC iceberg. Here’s part of the story: At the time Opel became IBM’s CEO, he was on the board of the super charity United Way, where he worked with fellow board member Mary Gates. Gates got Opel to meet her kid, whom she knew was pretty excited about computers. The kid’s name was Bill. The rise of Microsoft on the seesaw of fortune, with IBM on the other end of the beam, began right there.
When Fonzie, played by Henry Winkler, jumped his shark, the writers of the television series wanted do to something different. It was 1977, and the people behind the hit show were anxious to kick off their fall season with a memorable episode. So they set the season opener not in the show’s native setting of Milwaukee but in Los Angeles. In the show, Fonzie gets on water skis (while wearing his trademark leather jacket) and jumps over a shark, several in fact. Fans of Happy Days may argue that this episode was so good that the show simply could never match it. Others might agree that this episode was memorable but also that the show peaked with the shark jump story and, slowly at first, declined from that point.
After four years as CEO, in 1985, Opel was succeeded by John Akers, who had gone to Yale and once served in the Navy as a carrier pilot. Apple had unveiled the Macintosh making it clear to anyone who was paying attention that the future of computing had escaped IBM. By 1993, after eight years at the top, Akers’s tenure had gone from challenge to disaster. That year IBM lost $8 billion, bringing losses for the first three years of the decade to more than $15 billion. IBM’s directors were understandably worried that their company might actually collapse. The board took an axe to Akers and hired Lou Gerstner, an outsider, to rescue IBM. Which he did.
Gerstner’s transformation of IBM led to the Big Blue we have today, the IBM run since 2003 and until the end of last year by Sam Palmisano, after which the reins were handed to Ginni Rometty.
Rometty is a rarity: a top IBM executive educated as an engineer. Also, she isn’t a lifer; Rometty began her business career with a brief stint at General Motors. Palmisano, by contrast, is more typical of IBM’s top executives. He majored in history as an undergraduate. IBM was his first and only employer. (Lou Gerstner studied engineering at college but he never had a career at IBM before coming in at the top.)
When Gerstner came in, IBM’s revenue was declining. Even so, hardware sales topped $30 billion of a total that exceeded $62 billion. In other words, hardware sales in 1993 approximated the total revenue of IBM in shark year 1981.
Today, IBM’s revenue has increased another 50 percent, to more than $100 billion. Nevertheless, hardware revenue has declined to roughly $18 billion a year, or less than two-thirds of IBM’s hardware intake in 1994 and less than a fifth of IBM’s sales volume. Of that $18 billion, only about $12 billion comes from the sales of servers, and that includes X86 industry standard products (which have the highest unit volume), AIX Unix variant Power boxes, IBM i versions of the Power boxes, and System z mainframes. An additional $3 billion or $4 billion comes from other hardware sold to end users, mainly storage subsystems. The rest comes from foundry work, making chips for other companies.
IBM’s software revenue is now above $25 billion a year; it was in the vicinity of $10 billion when the company jumped the shark. Moreover, software, while providing a quarter of IBM’s intake, appears to bring in about 40 percent of IBM’s profits. It is much harder to calculate IBM’s actual hardware profits because IBM commingles funds from systems software with money paid for servers. However it is unlikely that IBM is bringing a lot of its hardware revenue to the bottom line. Reported pre-tax margin is in the range of 8 percent (and that figure includes profit from operating systems). IBM’s pre-tax profit in software is just shy of 50 percent.
While sales of hardware, particularly proprietary systems like i and z, pave the way for sales of systems software, middleware, consulting services, maintenance and more, things that are uniquely IBM’s, the company could, arguably, completely stop manufacturing its own servers and come out in the same financial place or better by marketing boxes built by others, as it does with X86 hardware assembled by subcontractors. Whether to build or buy is a matter about which IBMers, investors, and outside observers may differ. Nevertheless, it is obvious that server manufacturing no longer lies at the heart of IBM’s business strategy. It is even possible that IBM’s continued devotion to the midrange and mainframe servers it invented may stand in the way of its ambitions in cloud computing.
In the market as a whole, cloud computing is most often based on X86 hardware and Linux software, although there is still plenty of opportunity for success involving diversity. To pick one obvious example, Microsoft productivity applications offered as cloud services might well produce the best results for users as well as the vendor when the code is run on Windows Server (or a cloud version of Windows). Rival productivity app services based on OpenOffice can be pretty slick, but organizations with vast experience using desktop Office are not going to sacrifice their culture and knowhow just because some cloud vendor claims to have a good alternative that is more or less compatible.
IBM’s most devoted customers might constitute a market for IBM i in the cloud or AIX in the cloud or z/OS in the cloud, but, like users of Microsoft Office, even the Bluest users of cloud computing might just focus entirely on apps in the cloud and not the remote platform if those apps are really compatible with the software used in-house.
So, sentimentality aside, that app focus (including insistence on compatibility with any apps that also run in data centers) seems to be what is driving not just cloud computing for business but the must larger opportunity of cloud computing for just about everyone, which to business users means their customers. And the leaders in cloud computing, such as Google, Amazon, and Apple–IBM isn’t a leader yet–aren’t sentimental about Linux or any other server related technology. They are mainly concerned about pleasing more app users, making apps faster and cheaper, and adding lots more attractive apps to their offerings.
IBM, going its own way, is trying to build its own apps or at least its own app framework using Lotus technology and other software that has largely been created by its own technologists or by software vendors that have gotten the attention of Big Blue’s suits. Moreover, IBM’s idea of a framework isn’t a structure as (comparatively) simple as the one provided by Facebook or Twitter. IBM thinks it knows what the world needs or at least what its customers and prospects need, and so far it’s just too early to say whether IBM is correct about this. Some answers ought to become visible during the next year, possibly a lot sooner, as IBM attempts to reinvent Lotus yet again and cheer on an ecosystem of sycophantic developers that live on a different planet from the one where you can find the street urchins who build hot Apple iOS and Google Android apps. The IBM cloud must be living on the smart planet. The iOS and Android riffraff have to settle for living on the successful planet.
As the cloud business evolves, with or without IBM in a prominent position, users of IBM servers or customers of IBM’s cloud services may have to reckon with some radically different additions to IBM’s product line. (Well, maybe. IBM is very likely to user new kinds of servers to provide cloud services, but it’s less certain that the company will also offer the same hardware to end users in the form of new products.)
IBM is going to get serious about servers built from the ARM chips used in smartphones and tablets or the MIPS chips used in WiFi hotspots, routers, and intelligent utility meters. Even if the company only builds or buys SoC based servers to test them out or to gain experience on its internal networks, IBM simply cannot ignore machines that might rewrite the IT cost models across the whole computer industry.
Very soon, perhaps as early as next year, X86 chips might not dominate the cloud server market the way they do today and Power chips, despite their superior performance, are becoming less likely to get cheap enough to run inside anyone’s cloud data centers, not even IBM’s. Only cloud users willing to pay a premium price can expect to get Power-based cloud services, and while IBM might be able to generate billions in annual revenue delivering IT power to such customers, those customers will have a hard time surviving in any business that has to live with thin margins. It would be financially impossible to run Amazon.com on mainframes or i systems or Power AIX (or any other high-test technology such as Itanium or Sparc). It’s hard enough for Amazon to turn a satisfactory profit using computers that only need regular gas, which is why Amazon might be one of the first large enterprises to use SoC servers, the hybrids of the computing world. (Google might beat Amazon to the punch but whether this would be for sport or to serve pressing business needs is likely to be a question that gets asked more than answered.)
None of the issues that threaten to overturn established server technologies are particularly important to corporate users with relatively small collections of servers. The electric bill for a small glass house is peanuts . . . compared to the value of the work the servers perform. Power and hardware acquisition costs might not even be all that important to big mainframe shops, although at some point a bank running a megawatt mainframe complex might get run off the competitive road by a rival that has found a way to get the same work done with kilowatt racks of servers.
Intel is hoping its current and future chips eliminate the differences in power and heat between the chips used in mobile gadgets and those found in glass house blades when throughput is considered. But as much as Intel is forging ahead, the companies making SoCs by the tens of millions seem to be learning all the time, too. And the bare handful of companies making exotic server chips, such as IBM, Oracle, and Fujitsu, are just as smart as their X86 and ARM rivals, but they simply don’t have the production volume to justify the dozens of engineering experiments that the high volume chip makers such as Intel, Samsung, Texas Instruments, Qualcomm, Nvidia, and their ilk routinely undertake.
Bascially, if the X86 chip is under threat, the Power chip is besieged. Here is some evidence along that line of thought: In order to compete with Apple and its wildly successful iPad, some companies building tablets are trying to find a way to give their products the pep needed to play games while still delivering long battery life. Their answer might not include X86 technology.
The other day I spotted a press release about what might be a milestone product, a software system sold as an app that turns a tablet into a credible gamer. You may know that gaming PCs are very similar to industrial workstations. Both have speedy engines and very effective graphics processors, although the gamer machines have more dramatic looking cabinets than the engineering workstations. Dell, which likes to sell computers to gamers and to engineers, routinely learns things from one of these market segments that it can use in the other. If Dell can’t build a competitive next-generation of gamer boxes (or engineering workstations) using Intel chips, it could mimic its Taiwanese rival, Asus, and add Tegra 3 machines to its product line. A lot will depend not on attractive demos that get customers’ attention but on actual machines that please users day in, day out. (The outfit that produced Splashtop for Tegra has also produced versions for Kindle, a TI ARM system; iBook, which uses Apple’s ARM chips; Intel X86 Sandy Bridge machines and would most likely build a version for any other architecture if it looked like there was a half-decent market opportunity. Disclosure: I have a bootable X86 Splashtop installation on a laptop; it’s there to provide emergency services if Windows gets trashed.)
It might not matter whether SoCs with roots in mobile computing knock out X86 laptops, desktops, or servers as well as other servers like Power machines. The rival technologies could coexist if it turns out that neither can please everyone. But however this plays out, it’s starting to become clear that client side presentation is going to improve quite a bit and that this will very substantially change the requirements of servers (and networks, too). This will be the case whether the servers are inside glass houses or out in the cloud.
Just take a good look at popular smartphone apps to get a clue: These apps involve quite a lot of fast interaction between client and server, and it’s not just about things the end user taps on a screen. Many mobile apps include interactive geolocation computing, camera capture and image processing, sound capture, and motion capture. These activities generate data that gets beamed to a server or server farm, data that may be much richer and more voluminous than anything churned up by a mouse and keyboard, and the GUI data flow is large compared to the green screen interactions that are typical of legacy systems.
Sure, whatever servers you have can probably work with anything your users can beam at you. If they can’t, it’s probably just a matter of buying some more software. But that won’t necessarily make the solution that required the least initial adjustment on your part into the solution that delivers the best results for your organization. And it might not make a future for IBM’s server business or services business, either. If the time comes when legacy systems no longer offer adequate value, IBM is going to run out of the cost offsets it obtained by moving software development to Asia. Once IBM runs out of rope in rope trick country, server technology will change quite a bit and very quickly, too. And, based on what seems to be happening in the mobile apps business, which so far is still truly worldwide and not driven to Asia, computing may have just about reached that turning point already. It might not seem so if you look at IBM, but that could well be because Big Blue is such a laggard, not a leader, when it comes to facing the new economics of computing.
For IBM, and for those who depend on IBM’s goods and services, this new year may include the moment to do what Fonzie did. It might soon be time to get brave and jump that shark again.
Rometty Kicks Off 2012 With Leadership Team Changes
Palmisano Hands The IBM Reins To Rometty
IBM Lays Out Plans for Future Growth and Profits
Palmisano Rakes in $9 Million for IBM’s 2010 Performance
Big Sam Is Worried About Oracle–And For Good Reason
IBM Reorganization Tucks Systems Under Software
Palmisano Says IBM Will Double Up Profits By 2015
IBM Taps New Server GM in Wake of Scandal
international Business (machines)