by Brenon Daly
For all of its status as the hot new thing, machine learning (ML) has actually been around for some time. That’s true for the technology itself, which is basically an evolution of the data mining and analytics software that have been running for decades. Likewise, it’s true for dealmaking, where companies have looked to make their businesses and products smarter by acquiring ML startups for years.
In fact, the first record in 451 Research’s M&A KnowledgeBase in which an acquirer cited the term ‘machine learning’ as part of the rationale for buying a startup goes all the way back to 2003. For history buffs, the inaugural ML print – at least in our record-keeping – belongs to Nokia, which spent $21m for Eizel Technologies in April 2003. (The then-dominant phone maker was looking to Eizel to help smooth the rendering of email and web content on its phones, which were fairly limited at the time. The iPhone wouldn’t be introduced for another four years.)
More than just a historical artifact, however, the initial ML transaction we captured is worth revisiting because the reasoning behind the decade-and-a-half-old deal getting done could very well have been on a page ripped from a pitchbook making the rounds today. Even Eizel’s valuation, which works out to roughly $1m per employee, wouldn’t be out-of-whack in an ML transaction printed now, particularly if we consider the inflation-adjusted figure of $1.4m per employee.
As exemplified in that inaugural Nokia-Eizel pairing, the strategy and structure of ML transactions haven’t changed all that much over the intervening years. What do we mean? Essentially, the deal boiled down to an established technology vendor picking up some ML technology to optimize an existing product. Further, as is often the case for these ML startups, the technology had its roots in the computer science department of a research-intensive university (Carnegie Mellon University for Pittsburgh-based Eizel).
Those two trends came together more recently, for example, in Microsoft’s reach for Semantic Machines last summer. In that deal, the software giant was seeking some smarts around ‘conversational technology’ that it could deploy on Cortana, Azure and other products, so it picked up Berkeley, California-based Semantic Machines, which had ties to UC Berkeley and Stanford.
It’s almost as if that pair of transactions – separated by half a generation, involving vastly different buyers and technology applications – were nonetheless produced by the same algorithm. For more on how the emerging trend of ML is shaping both the software industry and the M&A market, 451 Research will host a special ML-themed edition of its M&A Summit next Thursday. For more details, see our website.
For more real-time information on tech M&A, follow us on Twitter @451TechMnA.