After hemorrhaging cash for the better part of a decade and filing for Chapter 11 bankruptcy twice in three years, it looks like storied server and supercomputer maker SGI (s sgi) might actually turn a profit again. In April 2009, simultaneously with its second Chapter 11 filing, SGI sold itself for peanuts ($25 million) to webscale server maker Rackable Systems, which — surprisingly — decided not only to keep selling SGI’s products, but actually to take on the somewhat-tarnished name of the then-NASDAQ-delisted entity. Now, after an initially bumpy ride that looks like the result of having to iron out some wrinkles and absorb SGI’s high-end product lines, the new SGI reported impressive revenue during the second quarter and raised its fiscal year 2011 guidance from “break-even” to “profitable.”
The methods the new SGI used to turn a profit from the old SGI’s money-losing business mirror some basic tips that your stock adviser, or your father, might have told you about how to make money with your investments. You want to buy cheap, put some money in blue chips and diversify. Of course, buying a large, unprofitable company is more than just a monetary investment — it’s also a business — and that means getting operational costs in check by any means necessary.
Rackable definitely bought SGI cheap. The $25 million price tag (technically higher because Rackable assumed certain SGI liabilities that weren’t discharged) is a mere fraction of the hundreds of millions in revenue that the old SGI used to bring in quarterly. It’s noteworthy, however, that the new SGI still isn’t reaping old-level-revenues from its meager investment — but its $177.5 million in second-quarter revenue represent an 88-percent year-over-year increase and a record quarter for the new SGI.
There’s a lesson to be learned here, which seems to be the virtues of patience. Technology companies such as SGI — or even Sun Microsystems, which Oracle arguably got for a steal at $7.4 billion — didn’t reach the peaks they did without great technology and even greater minds. But when their innovation-at-all-costs business models and poor decisions start dragging them down financially, there are rewards to be reaped for vendors willing to put forth the effort to stop the bleeding. Not everyone is headed for bankruptcy — which certainly eases a lot of the debt pain and takes shareholders out of the picture — but when they’re at their lowest, even shareholders might accept a low offer to get something from their investments.
Don’t Discount Blue Chips
High-performance computing is somewhat of a blue-chip market in the technology world, as there is a solid core of organizations and companies that need rely on it and will always be willing to shell out the big bucks required to get super-powerful systems. So, rather than try to just incorporate SGI’s technologies into the existing Rackable business of selling webscale servers, the new SGI embraced HPC and is once again a serious player in that space. According to CEO Mark Barrenechea, technical computing (SGI’s preferred term for the space, which Barrenechea says encompasses HPC and a whole lot more) is a $9 billion market with Dell (s dell) operating at the low end, IBM (s ibm) operating at the very high end with its proprietary Power7-based systems, and with the middle wide open for the taking by SGI and its lineup of Altix systems and high-performance storage gear.
In fact, Barrenechea told me about half of SGI’s business comes still comes from the public sector, including the Department of Energy laboratories, intelligence agencies and NSF-funded universities. “It’s a great market for us,” he noted, adding that SGI systems for these types of customers, as well as for large enterprise customers, typically pack between 50 and 100 teraflops per cabinet. SGI Altix systems accounted for 22 of the 500 fastest computers in the world according to November’s Top500 list, and, going forward, SGI also expects to be “squarely in the middle” of the race to exascale computers. The Obama administration is proposing $126 million for exascale research in its congressional budget.
The new SGI isn’t wholly dependent on selling huge shared-memory systems to national labs, though. According to Barrenechea, the company doesn’t break down revenue by divisions, but it expects only about one-third of annual revenue to come from the traditional SGI products, with the remaining two-thirds split equally between services and the Rackable business line, which focuses on dense, energy-efficient servers for webscale data centers. The company has never really moved the needle in terms of overall server sales, but it does have some high-profile customers — including eBay (s ebay), Carbonite, Microsoft (s msft), BT (s bt) and Amazon (s msft) — that contribute enough to the bottom line as they fill their data centers with Rackable gear. During a data-center build-out in 2008, Amazon (s amzn) bought $86 million worth of Rackable servers.
The Ugly Side
However, making an unprofitable company profitable isn’t always pretty. Much like Oracle (s orcl) axed numerous Sun products and employees upon taking over, on top of putting price tags on once-free products such as MySQL, the new SGI has had to trim some fat, too. For one, Barrenechea explained that the old SGI spent lots of money on projects that weren’t always tied to any direct revenue, whereas the new SGI tries to focus on customer- and industry-focused projects that it expects will pay off financially. A cynic might view this as killing true innovation in the name of making a buck, a stance that has some merit.
And just this morning, news broke that SGI is laying off 55 employees, roughly 4 percent of its global workforce. I inquired whether the layoffs are part SGI’s commitment to become profitable in this fiscal year, to which a company spokesperson confirmed via e-mail that “[s]treamlining operations and prioritizing efficiency is directly aligned with achieving SGI’s [fiscal year 2011] financial goals.”
Image courtesy of Flickr user YuviPanda.
Related content from GigaOM Pro (subscription req’d):
- Supercomputers and the Search for the Exascale Grail
- Pushing Processors Past Moore’s Law
- Think Converged Infrastructure Means Lock In? Think Again.