Less certain is whether "ubiquitous wireless networking" will be the cash cow that covers the staggering investments in 65- and 45-nm CMOS.

One panelist, marketer Robert Glidden of Impinj (Seattle), stressed that his company's 902- to 928-MHz RFID technology did not depend on advanced CMOS, and could be readily implemented in conventional 0.13- or 0.18-micron CMOS. Despite Glidden's projections for 50 billion chips per year, the question of whether this 10-cent item pays for tomorrow's fabs was left unanswered.

Panelist Ryo Imura of Hitachi Ltd. (Tokyo) said the "beneficiary" and the "cost-payer" for an RFID support structure are not always the same. Hitachi's Mu-Solutions RFIDs will support applications like inventory tracking and control to the tune of 10 billion chips per year.

"There are no real dollars in RFID," Imura said. The real money associated with RFID isn't in the 200,000 8-inch wafers his company produces, but comes from networked readers and data accumulators sold by companies like Cisco and IBM.

Gerard Beenk, senior vice president at Philips Research (Eindhoven, Netherlands), said "ambient intelligence" could perform many of the handholding functions now associated with anthropomorphic robots. A sensitive, embedded network would sense the mood of the user and use it to control the lighting, music and entertainment, Beenk added.

In terms of the intelligence required, Beenk said 10 (GOPS) billions operations per second is currently required to synthesize movies from digital source material. At least an order of magnitude more (100 GOPS) would be required to recognize faces or a user's emotional states, while another order of magnitude (1,000 GOPS) would be required to provoke an emotional reaction from machinery.

The manufacturing potential of such devices would in hundreds of millions of pieces, Beenk projected. This embedded sensor technology would be added to existing consumer devices: the 180 million TV sets produced each year, the 60 million cars, and the 600 million cellphones. And this count does not include the projections for future wearable computers and e-books. The potential for intelligent communicating devices would easily exceed the number of people on this planet, Beenk said.

Bert Gyselinckx of IMEC (Leuven, Belgium) described the potential of ambient intelligence for human fitness and health monitoring. Like Hitachi's Imura, Gyselinckx saw the real development issue not in the monitors themselves, but in the infrastructure required to accumulate and make sense out of the data. For every remote sensor, there needs to be a series of receivers and repeaters.

One suggestion is to scatter the sensing devices among applications like heart rate monitors, calorie counters and blood sugar monitors, using the patient's cellphone as the data-accumulation device. The collected information never leaves the individual's hands until, for example, he dials his doctor's computer terminal with the cellphone.

Thus, the proliferation of these sensors is attached to the proliferation of cell phones, which Gyselinckx pegged at 60 billion.

Panelist Kristofer Pister built a startup on a proposal he made the Defense Advanced Research Projects Agency on "smart dust" as a University of California- Berkeley professor. The so-called smart houses, smart cars, smart highways, even smart bombs of the future would all depend on autonomous wireless sensors.

The original "smart dust" project packaged a 1V, 1 MIPS processor, a 100 Ksamples/s 8-bit A/D converter and radio transmitter into a cube 1 mm per side, consuming roughly 304 microamps from a 3V supply.

"Low-power design is a bear," Pister admitted.

The mission of Pister's Berkeley startup, Dust Networks, is to penetrate applications not currently covered by conventional sensors. The $42 billion in revenue associated with the sensor industry is not all devices, he argued. The cost of installing a sensor is roughly ten times the cost of the sensor itself. The advantage of low-data rate wireless mesh networks is that they don't need a star topology, and can bypass traditional conduits and wiring harnesses.

"This market is the proverbial hockey stick," Pister said.