^{Democracy & Technology Blog} Telecosm Tidbits

We often forget about the secondary and tertiary effects of massive capital infrastructure investments. Two panel discussions at George Gilder’s 10th annual Telecosm, just concluded in Squaw Creek, California, indirectly reminded me of these virtuous side-effects.
On the first night of the conference, Terry Turpin of communications equipment vender and hyper-tech defense contractor Essex Corp., was listing some of the pluses and minuses of the telecom boom (bubble?). Essex makes optical processors capable of the most demanding computational tasks ever known, from breaking codes to looking through the walls of caves in Afghanistan. Can you do 5 petaflops on just 10 watts? Didn’t think so.

Anyway, Turpin mentioned that some of the machines he’s built in the last half-decade would never have been possible without the late-90s optical network buildout boom. Waves of new optical components — and old components at rock-bottom prices — flooded the market. Lenses, prisms, coatings, gratings, free-space interferometers, and assortments of lasers with more power and better precision. Pieces that Turpin had previously hand-made at prohibitive cost and pieces that previously did not exist were now available in bulk at reasonable prices. The super-secret machines Turpin has turned out, one by one, for decades could now be supercharged with the volumes of high-end technology churning out of the telecom industry.
The next morning of the conference, Gilder had engaged Google, Equinix, Mozy, and others in a discussion about data centers and the massively parallel computing architectures that deliver your Gmail, YouTube clips, and most other web services. Bottlenecks abound in these giant, bandwidth hogging, power sucking, behemoth hardwarehouses. But one key bottleneck that emerged from the discussion was the dearth of cheap 10 Gigabit Ethernet components needed to connect the tens or hundreds of thousands of processors and disks.
In the past, the motherboard bus and other computer system communications links had been faster than the network to which they would one day connect. Today network bandwidth is beginning to outstrip PC-level communications. But 10 GigE links are not yet made in the volumes needed to make economical the connection of these server farms with wires 10 or 100 times faster than the 100 base T Ethernet or 1 Gig E connections used today.
The fiber optic metro and access networks being built by the telcos today, however, are making abundant use of 10 GigE and 1 GigE links. Should government regulators in Washington and the states allow these gigantic fiber optic buildouts to go forward, a new wave of optical and electronic components will flood the market, trickling down into the storewidth data centers and across the network. One part of the ecosystem feeds the others.
Telecom investment is thus not important only for our entertainment and communications capabilities today, not just for our productivity next year, but for the long-term health of a range of military, security, high-tech, high-growth industries as well.
-Bret Swanson