Tim Oren's Due Diligence

Brobeck, when your future is

Brobeck, when your future is at stake.

High profile Internet boom legal firm Brobeck, Phleger & Harrison closes its doors. One of the more visible examples of the knock-on effects of the bust on the service firms. It's also another example of the famous Valley 'edifice complex' curse. The surest way to blow up a company is to get heavily exposed on real estate. In Brobeck's case, they took a 12 year lease on a new building in (East) Palo Alto, just at the peak of the real estate market.

Sweet! Latest sha-mail numbers from

Sweet!

Latest sha-mail numbers from J-Phone. Sure would enjoy being the general manager of that line of business. Any questions about where the scale economies advantage for photo-phones lies?

Grey ships, sailing Way back

Grey ships, sailing

Way back last April, in the first post of this version of the blog, I suggested watching the reserve fleet ships berthed up in SF to see when events were really starting to move.

Some things take a while. They were
activated Monday.

Now that is a big

Now that is a big crater...

$99 billion worth of 'synergy'. Impressive. I guess letting Ted Turner go will help the cash flow a tad.

Is Apple breaking out of

Is Apple breaking out of the box?

(I haven't blogged about Apple at all yet, so time for a nod to my heritage.)

You know the box I'm talking about. The

• minority platform forever
  
• oddball chip falling behind on performance
  
• dependent on Microsoft Office and Explorer
  
• shrinking developer support
  
• channel dead as a doornail
  
• pushed back into the creatives' ghetto
  

box.

I went on record that Apple would not last the century independent because of these accumulated debits. I had to eat the big black bird on that one. Now I think there are accumulating signs that Apple is at least creating room for strategic maneuver. Bits of evidence:

• iPod and Airport show Apple's ability to create real differentiation through packaging and interface in intensely commoditized markets.
  
• iTunes, iVideo et. al. expand the notion of a basic software suite, and play right into the growth of consumer digital media.
  
• Apple takes from the open source movement (KHTML->Safari), and gives back with Darwin. (I haven't used Darwin, but my neighbor has it on an old P4 box and says it's a real kick to see 'Welcome to Mac' on an x86.)
  
• Rendezvous is made open, provides a real usability benefit, and breaks down complexity barriers created by directory and other name space servers.
  
• Persistent rumors of a parallel x86 version of OS X that is tracking the PPC releases.
  

So how about some rank speculations on what pattern is being formed here...

One of the raps on open source has been that it's hard to create solid and consistent UIs through the community process. Safari and Darwin/OS X both position Apple as the UI value adder above an open infrastructure.

Apple is also positioning itself as a technology packager and brander in the large, getting set to move off of the computer per se on to peripheral and non-PC devices.

Apple is using the open source community to fill in some of its lack of commercial developer support, and as a Trojan horse to sneak into the 'x86' world.

Apple is pushing its dependence on Microsoft into parts of the market that matter less to the consumer user.

Apple is acting to undermine the value of the 'directory layer' in the Internet that is partly controlled by Microsoft, and partially removing a technical (and legal) single point of failure for distributed non-PC devices.

I'm not saying that all (or any) of these are received wisdom and strategy over in Cupertino. But the first stage in strategic maneuver is restoring freedom of movement. Apple could be wiggling out of the box.

There's more to Silicon Valley

There's more to Silicon Valley than chips and VCs

It's 9PM at the homestead, and the coyotes are yapping their heads off not far up the canyon, driving the neighborhood dogs nuts. Bet that doesn't fit most folks' image of this place in the world...

Care for some EvDO with

Care for some EvDO with that?

So the WaPo discovers a 'new' cellular data technology and gets the vapors, and the story rattles around the tech blogosphere in a rather uncritical way. Congrats are undoubtedly due to the flacks at Lucent, Verizon and Monet for scoring the placement, but it's actually called 1xEvDO (evolutionary data only) and it's not really news unless you've ignored everything happening in East Asia, the current home of cellular systems innovation. TechDirt Wireless gets it right, though.

1xEVDO is a Qualcomm CDMA flavor building on their 1xRTT (CDMA2000) standards. It's already widely deployed in Korea - hence the Samsung pocket PC/phone blogged by Gizmodo and is the basis of the J-Phone systems driving DoCoMo crazy in Japan.

Other than the WaPo's shortcomings as a tech rag, there are a few things to gather here:

• Qualcomm CDMA flavors and carriers using them seem to be quite a ways ahead of nominal '3G' standards and carriers in successful deployments, and figuring out what real customers will accept in services and pricing.
  
• The acceptence seems to be biased toward the best-efforts 'data only' variant as opposed to the more voice-oriented 1xEvVD voice + data version.
  
• US carriers lag as usual. Given their costs of spectrum, the notion that carrier priced 1xEvDO is going to seriously impact the deployment of 802.11 with its expectations of free or flat rate megabit service to laptops is seriously offbase.
  

In spite of dishing Verizon above, the CDMA carriers do seem better positioned for incremental exploration of data services (if they actually carry through) since they can incrementally add 1xRTT derived standards without a network rebuild.

Don't trash that old inkjet

Don't trash that old inkjet printer!

It turns out that you could be printing out living tissues, if you reload the cartridges with the right kind of cellular mash!

This may be the most dramatic reapplication of inkjet technology for creating precision structures, but it's far from the only one. Experiments are busting out everywhere, from polymer circuits to displays
to biosensor arrays, even metalizing photovoltaic cells (PDF).

The secret sauce is the goop that goes into the inkjet of course, but it's still remarkable what can be attempted with a scrapped out $100 printer and an empty cartridge. Best if it's an old Epson, though, because their piezoelectric technology poses fewer chemistry problems than the heat-driven approach uses by HP and others. Look
here (PDF) for a survey of the best inkjet to use in your precision fabrication experiment.

UNE-P redux The latest rumor

UNE-P redux

The
latest rumor is that the FCC will only partially back away from the reg requiring ILECs to unbundle copper circuits for competitors' use. The trial balloon is to continue to require unbundling on existing plant, but allow sole use of newly installed equipment and lines to encourage investment in broadband infrastructure. OK, could work, with a few caveats:

• 
  Continued oversight of performance on unbundling.
• 
  New sole use builds cannot be put into the regulated rate base.
• 
  An aggressive policy on licensed BWA spectrum suitable for 802.16 usage, cutting it up into small geographic areas to encourage overbuilds, with the incumbent wireline carrier forbidden to own it.
• 
  Incumbent wireline operator also forbidden to operate 802.16a in the U-NII band in its territory.

Then let 'em all duke it out.

Moore's Law considered harmful? The

Moore's Law considered harmful?

The newest Red Herring arrives with an article by Michael Malone (not yet online) wondering whether the celebrated Moore's Law has become counterproductive for the semiconductor industry and for IT as a whole. Update: It's here

While Gordon Moore's observation about the doubling time of semiconductor performance has been highly reliable, it's not a statement of natural law. Rather its an observation about the aggregate behavior of an industry over time. A collective hallucination, if you would. This particular dance is so well planned that you can download the script. Everyone reads the same time tables, and all head out to make their respective investments, from EDA software to mask tools to next generation crystal growers and fabs. Everyone knows their deadlines, and the whole industry moves ahead in unison - ker-chunk - down to the next smaller geometry and/or up to the next larger wafer size. Miss the beat and you're road kill. Keep it and profits are yours. Right?

That's the rub, of course. The game keeps getting more expensive. A next generation fab is over $2 billion and counting, for example. The total capital exposure at each generation is likely somewhere north of $100 billion (anyone got a better number to share?). If that money doesn't come back in profits, the whole game can come to a sticky end, and that's Malone's valid worry.

A lot of the ROI for each generation has typically been returned in a 6-9 months window when a premium can be charged for the latest and greatest microprocessor, from OEMs and others who just have to have the most cycles to stay competitive. After you've recouped a good chunk of your capital costs you can knock down the price and margins, drive the new generation core out to a wider market and put more commodity chips into the fab, and prepare to repeat the trick. Everyone back along the food chain eats because of this cycle.

It has been getting harder to get the ROI, particularly using the MPU as the main driver. It's been awhile since most end users have needed the latest processor on their desktop, unless they do high end graphics and video authoring, or happen to have a taste for 3D action gaming. Gilding the lily on the GUI does not create high margin sales.

The server side has been a savior to the last few generations, aided by the shift to Internet transactions and clustered server architectures. Running out of cycles is not an option when the customer is waiting. What's got the wind up Mr. Malone is that Eric Schmidt, CEO over at Google, isn't buying the lastest and greatest chip generation for his server farm. Mr. Schmidt is one of the smartest technologists in the Valley. His heritage is from Sun, mother home of the non-commodity, high end server. Google prides itself on wicked fast search. So how is he creating his server farm? His staff is building it. From generic Intel architecture motherboards. Using the last generation of chips. Google is not the first to figure out that clustering generic 'white boxes' is a cheap way to get cycles, but this may be a very public watershed moment, when the other path for recouping Moore's Law investments breaks down. Well worth some angst.

With the MPU route in doubt (and the NPU market toast, for those who follow it), now what? Moore's Law is a specific instance of something called an experience curve, or learning curve. A generic learning curve has units produced on the X axis, and unit costs per equivalent capability on the Y curve and describes a nice negative exponential drop with a coefficient typical of the technology. (Remember, it's just a collective hallucination that the X axis is defined by time.) You can draw one of these for anything from aircraft to CRTs.

What we're going to need are some new learning curves, but with something other than clock speeds and chip geometries on the Y axis. Something we can get paid for again. Maybe it's lower power. Maybe it's a cheaper to program architecture. Maybe it's just dropping unit costs relentlessly and driving processors and circuits into ubiquity. We're just starting to find out. Until we do, everyone is at risk. Dance the old dance, keep the beat, and you can still lose. Or bet on a new learning curve and try and find your place in a food chain whose existence is still an informed guess.