Occasionally I feel compelled to write down my thoughts, kinda like a blog. Except that I don't want to call it a blog. A blog brings back memories of me trying to keep a diary, something I was notoriously bad at. So just consider this a collection of personal ramblings.
Leendert van Doorn
Only the other week I was reminded of how careful you have to be in blog entries. Apparently my writeup below has been read by quite a few IBM executives and it annoyed them and they consider it ungrateful that I rejected their offer and feel that my blog is bitter.
What can I say? Part of this is no doubt a natural human reaction to rejection. I considered IBM's counter offer very seriously up to the last week before the deadline but in the end I asked myself the following simple question: If I want to do x86 virtualization and security how and why would I do that at IBM? Especially when you have an offer from a company that builds actual x86 CPUs.
My view of the state of IBM Research wasn't intended to be bitter. In fact, I think that from my description it is pretty clear what IBM Research has to do: Don't compete with startups and focus on projects that have a 5-10 year outlook. Anything shorter term is advanced development and should be done in the divisions themselves. Co-locating advanced development labs and Research facilities and migrating people back and forth to ensure cross fertilization of ideas makes a lot of sense to me. I also think it will make this reinvention more palatable to its employees.
Will IBM Research implement a plan like this? I doubt it. It is serious organizational overhaul and most institutions tend to resist that. It will be next to impossible for the current executives to change the Research organization because it will resist and drag things out. Perhaps bringing in an outsider, someone with the stature of Lou, is what is needed.
Of course, this isn't the only solution to this challenge. But the fact that IBM Research needs to reinvent itself is, in my opinion, inevitable.
Austin, Texas.
Many people have asked me why I left IBM Research and what my thoughts are about IBM and especially IBM Research's long term future.
First of all, IBM is a great company. It has its challenges but the overall path IBM is on is a strong one. I personally think that the Research division has some serious long-term challenges but more about that later.
Why did I leave IBM? The short answer is that I wanted to take the next step in my career and that despite increasing indications from my side (more than a year) IBM failed to respond adequately. The reasons for this were largely external: The organization I was in was in turmoil because of reorganizations and the Systems and Technology Group (STG) was drastically reducing its expenses. Not a good time to look for another job in IBM. In the end when I announced that I was leaving, IBM did make me a Distinguished Engineer and created a CTO position for me in STG but at that point I was no longer interested. As a former IBM executive pointed out to me, the future is pretty bright outside.
To be honest, this is only half of the story. Underlying all this were my ever growing concerns about the long-term role of Research in IBM. Let me explain.
I think that in the end IBM's innovation strategy is doomed to fail. This isn't because IBMers don't innovate but they don't do it any faster than the rest of the industry and this is not good enough in a competitive environment. In today's world the most efficient technology innovators are startups and for Goliaths like IBM it is hard to compete with these David's. Like many other multinationals, IBM's response to this is to acquire these young companies and their intellectual property and technologies. Especially IBM's software group has had a very active acquisition strategy and the services and systems divisions are rapidly ramping up theirs. As far as I am concerned this is the right strategy but it has some very profound long-term implications for the Research division.
The benefits of having a Research division are many. They range from the intangible such as luster and recruiting to the tangible such as intellectual property and advanced product development. IBM Research tends to do a lot of advanced product development, this is encouraged by the internal reward system and it ensures that Research continues to be relevant to the rest of the company. Still product impact isn't easy and many projects fail because of the typically cross-division collaboration challenges. The net result is that using Research to do advanced development is not very cost effective and thus not sustainable when compared to a well thought out acquisition strategy.
The other benefit of a Research division is the generation of intellectual property (primarily patents). Research generates a lot of patents but so do the product divisions. However, the role of Research is to be more forward looking and act as insurance for the company. This is very important. IBM is a conservative company (which I like as a stock holder) and will rarely join new technology in the first wave but by the time IBM is ready it can rely on its Research division for the know-how and intellectual property. However, with the acquisition of a startup you get the intellectual property and it is even somewhat product and customer tested.
In this new reality with each division having its own active acquisition strategy what should the role of Research be? Some wishful thinkers hope that Research will become more "science" focused, which may well happen but it is hard to imagine that IBM will continue to invest in 3000 researchers in that case. For people like me, who like to have impact on the business and drive technology, Research may no longer be the best place for this.
So what is a right place? Well, for example, a CTO job in a product group might be such a position. The problem with this is that IBM has many CTO's who rarely have budget control. That means you have everyone's ear but other than persuasion no stick to make things happen. This is not necessarily a bad thing after all this is your typical role as a researcher who is trying to make something happen in a product group.
The kicker for me personally is IBM's myopic focus on PowerPC. As far as I'm concerned, for better or worse, the market has picked the x86 as the defacto universal instruction set architecture (ISA) and IBM does not have a stake in this high-volume architecture. They are letting AMD and Intel define it. What's more, AMD and Intel are in such a competitive race for features that they are rapidly commoditizing technologies that IBM considers to be their key differentiators. Examples of this are virtualization, scalability, management and to some extent reliability.
In the end, I think, the big differentiator for IBM will be reliability. It may not make economic sense for AMD and Intel to provide the same reliability guarantees as PowerPC. However, these are niche application areas that IBM is already supporting with their mainframe business.
Of course, PowerPC does have its merits but my point is that as its differentiating features are being commoditized at such a rapid pace that its potential market is becoming smaller. The question is, is IBM in the end willing to maintain and support two niche architectures? This isn't just a hardware design question but a full horizontal software stack question.
Hang on are no doubt some of you now saying. What about all the game consoles? Yes, those are a big (well deserved) boon for IBM. However, it is far from a secure market, is it? The switch to PowerPC shows that that market is willing to change instruction set architectures overnight and while I don't have any inside information, I don't doubt that the x86 vendors are working hard on recapturing that market.
This is a gist of the reasons why I left IBM. I'll be starting at AMD in another week and I'm excited about working on an architecture I'm passionate about.
Valhalla, New York.