Special Feature: So You Wanna’ Start an HPC Business?

Print Friendly, PDF & Email

Whether you’re a startup or an industry giant, one day you may catch the HPC bug. You may ask yourself “why not start an HPC business?” After all you’re visionary, a risk-taker, smarter than the average bear. Besides, success at the apex of technology – worthwhile in its own right – can also translate into untold opportunity and advantage in less demanding downstream markets.

While the lure of HPC can be rightfully strong, a growing body of empirical research suggests some counterintuitive keys to success for those who dare dip a toe into HPC waters for the first time. This research stems from “disruption theory,” a body of innovation theory pioneered by Clayton Christensen, the legendary Harvard Business School professor who delivered the keynote at SC10 in New Orleans this month.

Here are 3 disruption-minded tips for starting an HPC business:

1. Avoid high performance. According to disruption research, new entrants tend to fail if they position their offerings as having higher performance than pre-existing incumbents. In other words, if your market entry strategy is to out-perform existing solutions, think again. Such businesses disproportionately fail.

This is counterintuitive, especially in HPC where firms tend to target the high end almost by definition (i.e. if they weren’t positioning themselves as higher performance, they probably wouldn’t be in HPC to begin with). However the results are in; such firms have alarmingly higher probabilities of failure. Avoid the high-end when you first enter an industry, even if that industry happens to be HPC.
The reason new entrants in this situation tend to fail is simple. By out-performing competitors right out of the gate, incumbents are forced to respond aggressively (or risk losing their best customers). Pitted head to head in this manner against, larger, more powerful, better funded incumbents, startups tend to fail.

2. Target low performance or invisible markets. Rather than targeting better performance, disruption research finds far higher probabilities of survival for new entrants that start with lower performance, lower cost offerings. Instead of targeting the high-end (like the now defunct networking and storage firm Entrada Networks did the late 1990s), new entrants should target the low-end (like NetApp did during the same timeframe).

Alternatively, startups also tend to survive if they begin in entirely new, uncharted, invisible market segments. Rather than being lower cost and lower performance than the incumbents, they simply go where no incumbents exist. For example, Linux allowed new generations of less sophisticated developers with smaller budgets to enter the market for the first time and begin writing enterprise/HPC code. To some extent Linux did, in fact, compete with Unix and other proprietary tools. However the bulk of its growth didn’t come from direct Unix converts, but from the larger population of new developers that could finally enter the market due to Linux’s accessibility and convenience.

3. Don’t have vision. While it seems like every 3rd title in the airport bookstore touts the virtues of “visionary” leadership, a growing body of data-centric research is painting a very different picture of what kind of leader helps a startup survive. As it turns out, startups tend to do better without “O-Captain-my-Captain” leaders who guide their firms in one steadfast direction towards an unwavering future vision.
Instead, since the business of startups is highly iterative and uncertain (more is unknown than known) startups tend to be more successful with leaders who focus on learning quickly and making rapid, informed course corrections using real-world data. Apparently it’s less about vision and more about learning. Victory goes not to the proud, but to those who most quickly uncover and respond to the winning strategy… that is, before running out of money.


While it’s never been easy for startups to succeed in HPC (or any highly competitive industry for that matter), some bodies of research are beginning to make progress. It is possible to learn from those who have gone before and not every HPC startup has to make the same mistakes as its predecessors. By studying business more scientifically and empirically (rather than merely philosophically and subjectively) we can materially improve our odds of success; and that’s what high performance is really all about.

About the author: Thomas Thurston is President and Managing Director of Growth Science International.


  1. This is very reconfirming of my experiences with startups. Thanks for the nice article.

  2. I don’t think a lot of startups have the patience for a disruptive approach, especially since they tend to be so technology centric in HPC. You have a lot of really smart people who want to flex their tech muscle, not build a business from the low end. While I appreciate this article and think it’s right on, I still think it’s going to be a hard pill for HPC to swallow. I guess that will be the difference between startups that make it and those that don’t.

  3. Yes nice post, thanks.

  4. I read this article and had to think for awhile about why it was valid and sensible, yet so different from the advice and discussion for start-up’s/ entrepreneurs that I usually read. Maybe this is why: An HPC start-up is mostly about technological innovation with a small amount of marketing insight. It also requires lots of funding. Start-ups of the Web 2.0 sort, and dot net ventures e.g. in the fold at Y-Combinator, or covered by Tech Crunch, are the sort that stir up a lot of interest, and are most visible lately. They have very different profiles than an HPC start-up. Although some are well-funded enough to do basic research, rather than social network applications…!

    Your bullet points one and two make sense for any business where success is easily quantified by an objective measure of performance. One becomes very vulnerable to ranking. There is no wiggle room, in comparison to endeavors whose success is driven by subjective aspects too.

    Bullet point three resonated with my personal beliefs about “the cult of personality”. The visionary leadership thing rankles for me. I grudgingly acknowledge that it has a place in certain contexts. Yet I dislike the rigidity and “tow the line” ethic it tends to foster.

    “Victory goes not to the proud”…. Very nicely said. Flexibility and the willingness to acknowledge mistakes without a crippling crisis of confidence will most often win in the long-term. Or so I would like to believe.