News broke this morning in Calcalist that Primesense, the Israeli semiconductor company behind Microsoft’s Kinect 3D sensor, was acquired by Apple for $345M. It’s still a rumor, and I don’t know if the story is true. In a year full of massive Israeli VC-backed exits, this one is not nearly the biggest, the most widely anticipated, or the most breathlessly covered by the press and blogosphere, Primesense was in no way a deal that I led, but I was close enough to watch the company’s story unfold. So this story hits home for me in a number of ways:
First meeting, first impressions. I was a principal at Genesis Partners when the company was raising its seed round. They had a commitment from Gemini for around $5M, but were out raising more. Together with Eddy Shalev, Eyal Kishon, and Ron Yachini, I attended a meeting with Aviad Maizels and the rest of the PrimeSense founding team. They demonstrated the early prototypes of their technology, and showed how they intended to revolutionize 3D computer vision. They had a demonstrator unit wired together with duct-tape. Eddy, Eyal, Ron, and I left the meeting and looked at one another. This was why we were all in the VC business to begin with – to find and back true innovation. I don’t remember who, but one of us said this was a “no brainer,” and we all agreed. A few weeks later, Genesis joined Gemini with a $5M check, bringing PrimeSense’s seed financing round to $10M.
Real technology is hard, and takes a lot of money. When we met them, PrimeSense was attempting to do something pretty amazing: they wanted to revolutionize the 3D computer vision market by making it far cheaper and more accurate than ever before. Before PrimeSense, other 3D vision technologies existed, mostly involving one of two approaches. The first approach (stereoscopic) used two cameras and looked for correlations between the two images in order to infer depth – much like our eyes do. The trouble was that these systems were very expensive, didn’t work in low light, and were unreliable especially when presented with things like patterned clothing that completely messed up the algorithms. The other approach involved a constantly moving camera – great for some military applications but otherwise impractical. Other software-only approaches that attempted to gather depth data from a 2D image were hopelessly inadequate. PrimeSense came up with a novel approach: a beam of patterned invisible light was projected outwards from the sensor unit, it was reflected back by the objects in the field of view, and the reflection was matched against the original projected pattern. This approach worked in low-light, wasn’t thrown off by weird colors or patterns on the target objects, was super fast – and could be implemented on a low-cost chipset – which was exactly what PrimeSense did – eventually bringing the cost of such a unit down by two orders of magnitude – or enough to get it into everyone’s living room.
Protection via the semiconductor. PrimeSense’s innovation consisted primarily of their approach to the problem and the software algorithms they were using to infer the depth information from the reflected pattern of structured light. They could have attempted to license this IP as a software-only solution. Instead, they chose to become a semiconductor company. This meant (1) taking their algorithms and embedding them onto a chip and (2) seeking out consumer electronics companies such as Microsoft and helping them to build systems that would require the chips that only PrimeSense was building. Needless to say, this was a bold move in a high-stakes game. Expressing their technology in a chip meant that it couldn’t be hacked or easily reverse engineered and it meant that the company could ensure performance would be optimal. On the other hand, it meant a long and painful route to market.
Consumer electronics are the hardest nut to crack. So PrimeSense had developed an entirely new capability for consumer electronics – 3D vision. But getting that into the hands of consumers would require a deep partnership with a consumer electronics company. This proved incredibly difficult. The standards of performance, reliability, and usability demanded by the world’s leading consumer electronics companies are exacting. Whereas enterprise customers will often tolerate some bumps on the road when deploying new technologies, consumers are far less forgiving. If a technologist wants to get his chip or software into a Microsoft Xbox or Apple system – that’s probably just about the highest bar one can set for one’s self. Getting the PrimeSense chip into the Kinect was an odyssey to say the least.
Dependence on a single partner is a scary place to be. After PrimeSense secured the Microsoft relationship there was, of course, a long waiting period before the chip actually made it into units that got onto store shelves and into consumers’ hands. The stakes for the company were extremely high. Naturally, exclusivity agreements prevented PrimeSense from selling the same sort of solution for a similar application to someone else, and confidentially agreements had the whole situation deep under wraps. For the management team and the VCs involved, all I can say is “nerves of steel.”
Looking for other applications. Partly because of exclusivity agreements – and partly because they wanted to build a really big company – PrimeSense looked far and wide for other applications for their chip and its unique set of capabilities. I haven’t been involved with the company for a long time, but it appears that those activities never really yielded the results they hoped for. Millions of people played with the Kinect and quite enjoyed it. A developer eco-system sprang up around 3D sensing – once in San Diego, I ran into a team of Indian entrepreneurs that was building a virtual fitting room technology based on a hacked Kinect unit (“Israeli technology inside”). But the idea that people would control televisions with “Minority Report-style” gestures never really seemed to take off. It will be interesting to see what Apple is planning to do with this technology…more on that below.
Second source and the first law of semiconductors. In 2009, Microsoft bought 3DV, another Israeli 3D vision company, for around $30M. In 2010, Microsoft bought Canesta, a Californian company in the same space. The writing was on the wall. About a year ago, I met a team of entrepreneurs in Israel. Among other things, they told me that Microsoft had managed to build its own version of the “PrimeSense” chip internally and would be shifting away from dependence on PrimeSense for 3D vision going forward. In addition, they told me that other semiconductor companies had developed similar technology and were going to be soon bringing it on market at much lower prices that the ones that PrimeSense had been charging for the original Kinect chipset. The first law of semiconductors (far more important than Moore’s law) is that price erosion always happens – and PrimeSense was not exempt from that reality. The company had moved faster than anyone else, had a better product than anyone else, and had expertly managed to partner with a massive company to get that into the hands of consumer. Perhaps they kept the price too high and, thereby, drove Microsoft to seek a second source – we’ll never know. But ultimately, the outcome would have been the same – other supplier would have come online, the price of the system would have started to drop, and – in order to stay independent – PrimeSense would have had to either (1) grow rapidly into new markets or (2) maintain a far superior technological edge to keep the competition at bay. Others closer to the story will one day analyze the internal strategic and operational decisions of the company, but the external factors were – in many ways – beyond their control.
Great expectations. Let me be clear: PrimeSense is a massive success story for Israeli technology and for the Israeli VC community. It is a deeply innovative technology that was seeded by two local Israeli VC funds in the grand tradition of venture capital. But there is also no doubt that the expectations for the company were greater than the size of the ultimate exit. At one time, there were widely shared hopes that PrimeSense could expand into other markets, go public, and remain independent. At the peak, revenues from Microsoft alone were rumored to be north of $100M a year, and the Xbox Kinect was noted as the fastest selling consumer device on record. Ultimately, however, the company didn’t find a way to escape gravity. We must not gloss over the incredible success that PrimeSense represents, but there are lessons to be learned here as well.
Apple appears to be in Israel to stay – and to stay at the forefront of hardware innovation. Apple’s acquisition of PrimeSense comes only a year after its acquisition of Anobit, for a similar price tag in the hundreds of millions. It’s clear that Apple is expanding into Israel for the long run – in search of talent, know-how, and patents. But it’s also clear that there is a theme here: Apple has built a name as the builder of the best hardware in the world – and two recent hardware acquisitions in Israel confirm that the Silicon Wadi is a world-leading center of innovation in that arena.
Not because it is easy, but because it is hard. It is stories like PrimeSense that make me want to be a VC – and it is teams like Aviad, Inon, and the PrimeSense team that inspire me to look hard for world-beating technologies. PrimeSense is the story of how four friends from the Israeli army reinvented an entire industry and took a company to hundreds of millions of value and brought a benefit to millions of people around the world. Some recent exits and valuations have been widely discussed lately on company with negligible (or non-existent) revenues and, often, limited (or non-existent) technology. Let there be no mistake. PrimeSense is the real deal. Real technology. Hard technology. Massive value creation. And it started with four kids and an idea. Well played! Mabruk!