Is it a feature, product, or company?

This is one of the key questions a VC will ask themselves after hearing your pitch. Inside Intel (as opposed to Intel Inside), they talk about the Intel funnel--it’s a nasty piece of business engineering that takes what was a product, and turns it into a feature on a chip. In fact it can often take a company and turn it into a feature on an Intel chip (not so good for the company).

Remember when WiFi first came out, and we had those little PCMCIA cards that plugged into our laptops? Then we got USB plug ins. Then within about 28 months, WiFi was embedded on the Centrino chip. This is the way of the semiconductor--its why you never want to build a business in the way of CMOS, it will crush you--silicon always finds a way.

A corollary in lasers is, if there is any other way to solve a problem without using a laser ... then do it that way. Lasers generally only win in applications where there is no other option: ophthalmic lasers win because only they can get inside the eye non-invasively. Radio frequency (RF) or intense pulsed light (lamps) tend to beat lasers in most other applications because they are easier, lower tech, and less expensive.

Ironically, I’ve been run over by CMOS a couple of times. Worst case was EDC (electronic dispersion compensation) in telecom or optical silicon. By throwing DSP (digital signal processing) at dispersion compensation, a little Intel company called Intersymbol was able to negate the effects of dispersion and extend the reach of a variety of transponders, while the rest of us were selling optical solutions firmly convinced that this was the only and most elegant way to correct for dispersion. But take heart, the EDC company also turned out to be just another feature, and needed to use that to create a unique transponder to become a company ... but wait! That company needed to be acquired to be part of the larger ecosystem before the entire entourage of technologies could become a real company…..ugg…..

With optical silicon, Translucent created an optical gain medium trying to piggyback on CMOS and help solve the pin problem limiting speed of chip communications--again, the problem, even when you are trying to be part of the CMOS ecosystem, is it will always find a way to do it in silicon anyway. So as a startup, you may have a unique software algorithm that can accelerate chip communications, you may be able to do it faster than anyone else, and you’ve patented the hell out of it, but you are still at the mercy of a chip supplier to use your product and when they do it becomes a feature on their product and you have no company.

In contrast, you may be able to use this unique acceleration algorithm to create a new type of chip that you can supply to, for example, Cisco to make faster routers, and Apple to make a faster web interface on its I-phone. In this case you may just have a company (and some nice acquirers to boot ;-) ). The femtocell BLOG I wrote a while back is another good example. A large ecosystem has developed around femto, from carriers doing trials, to investors, femto vendors, and suppliers to those vendors. The femtocell makers are trying to sell to the carriers, the chip makers are trying to give the femto makers an unfair advantage, and everyone is trying to guess when the carriers will deploy, and the carriers are trying to drive down the ASP from $300 to $50 ... but promising massive volumes ... sound familiar?

Someone making a chip controls the heart of the femtocell, enabling the key characteristic valued by the carriers. That vendor is in a position of power while they can lead the adoption curve. Some cell maker recognized this early and leveraged that characteristic and is getting design wins--very likely, everyone else in the ecosystem is trying to tag along for the ride but will get crushed by the price erosion---sound like fun?


And by the way ... I’ve been getting quite a few emails with Q&As for the Blog, instead of postings here. So to make it fair for all feel free to send questions/comments to lmarshall@sxvp.com BUT, please use the subject header BLOG so I can filter them ;-)