Do Sensor Investments Make Sense? –

Kris Pister coined the term “smart dust” in 1994. It was a tongue-in-cheek thing, he says, a smart aleck way of describing the bite-sized sensors he was building in an electrical engineering lab at UC Berkeley. He never imagined that he’d soon have venture capitalists chasing after motes – the Middle English term for a speck of dust that now refers to a single node along a wireless network of sensors.

Pister picked up a $7 million slice of that enthusiasm late in February, when his startup, Berkeley, Calif.-based Dust Inc., closed a Series A round with commitments from In-Q-Tel, the venture arm of the Central Intelligence Agency; Foundation Capital; and Institutional Venture Partners.

Dust Inc. is one of four smart dust startups that have sprouted in the last five years to try to commercialize academic research like Pister’s by using wireless technology to build a network of sensors. Together, they’ve attracted in excess of $53 million from venture capitalists.

Sensors linked together in a wireless network represent the convergence of next-generation sensor technology, wireless networking capabilities and the automation of business processes across nearly all vertical markets. A wireless sensor network can be used to measure a skyscraper’s energy efficiency, monitor chemical leaks in a fertilizer factory, or record soil moisture in a vineyard.

To make a wireless sensor network work, each mote needs to have a sensor, a battery and a radio chip in order to collect data and pass it on. It converts the data into a wireless signal, which jumps from mote to mote in a mesh network until it reaches a bay station, which, in turn, sends the data back through the Internet for delivery on a laptop, cell phone or a handheld device.

Smart dust companies, the ones like Dust Inc., are part of a bigger market for wireless sensors. Navtrak, a Salisbury, Md.-based startup, has developed a system to monitor delivery trucks. It might place a temperature sensor on the trailer of a refrigerated truck, take a read, then send that information over a cellular network back to an operations center to make sure poultry deliveries stay as cold as they need to be. What makes smart dust companies different is that they communicate along an unregulated part of the radio spectrum and the particular markets they’re going after, namely, defense, manufacturing and inventory control.

No Wires

Proponents of the technology say that wireless sensor networks have a broader reach than wired networks. If a mote can be dropped into place, use mesh networking technology to communicate with other nodes, the network is cheaper to install and maintain than wired networks. Each wireless mote costs about $100 to install, while each mote in a wired network costs about $600 to install, Pister says.

“We need to convince the market that this is real technology that’s here today with real ROI and clear value to existing businesses,” Pister says. “The question is how quickly it happens.”

What smart dust companies are asking customers to do is to replace legacy systems connected together with wires with a wireless one. To do that, they need to prove their technology is not only cheaper but also better than the wired systems customers have in place. Better as in faster, more sensitive, more responsive and more durable. And there’s the rub.

Not one of the four venture-backed smart dust companies has yet been able to secure a position as a leader in this emerging market. They’re going after federal and defense contracts, and have started partnering with universities and corporate research groups to think up new applications. Others have already come up with novel ones. Digital Sun, a San Jose, Calif.-based company, began selling sprinkler systems last February that turn on when they sense the lawn has gotten too dry. Pister, though, is certain there is near-term demand. He’s readying his company to begin full-scale production of sensor networks by the end of the year. He faces competition from at least three other startups: Crossbow Technology, Ember Corp. and Millenial Net.

Crossbow Technology, a San Jose-based company founded in 1998 with technology that’s based on Pister’s earlier research, has raised $12 million from venture capitalists. Its backers include The Cambria Group, Intel Capital and Morgenthaler Ventures.

Ember Corp. of Boston has pulled down the most money. The company, founded in 2001 by two MIT grads, has piled up $28.3 million. Its backers include Mr. Ethernet, Bob Metcalfe, Draper Fisher Jurvetson, DFJ New England, GrandBanks Capital, Polaris Venture Partners, RRE Ventures, and Stata Venture Partners.

Millennial Net was spun out of MIT’s labs in 2000 and has secured $6 million in two rounds from the likes of General Catalyst Partners, Globespan Capital Partners and Kodiak Venture Partners.

These four are building sensor platforms, not sensors. The company provides a matchbook-sized piece of hardware equipped with a battery that can last up to seven years, plus an operating system and a wireless communications device. The sensor itself and the software needed to decode the information it transmits are plug-ins.

Drum Roll, Please

Right now, smart dust companies are showing off the technology in a whiz-bang pageantry of one-off examples of how the technology can be used, but none has been able to deploy large-scale commercial networks.

In the mountains east of Los Angeles, at the University of California’s James Reserve, Crossbow has set up a sensor network to monitor complex weather patterns and the movement of nutrients and pollutants through groundwater. The company’s sensors monitor and record environmental factors like temperature, soil moisture levels, leaf wetness, wind speed and wind direction. Back at UCLA, computer science professor Deborah Estrin downloads the data off the Internet. Crossbow has licensed technology from Pister’s lab to build a sensor network and has secured a licensing agreement with Intel to provide a gateway chip that will turn the data into a readable and useful form.

At Lake Okanagan, in Canada’s British Columbia, vintner Don King of King Family Farms says his “ice wine” depends precision farming – knowing exactly when and how much to water the grapes. His vineyard serves as an Intel Research project. A system of 70 motes monitors all the environmental factors that build a grape’s sweetness and taste and then translate that into data King can use to manage his vineyard.

Boston-based Ember has focused its energies on developing sensor networks for building monitoring, inlcuding HVAC, lighting and refrigeration control. It has developed a prototype to detect weapons materials in shipping containers, and began collaborating with the U.S. Department of Energy in December on a $10 million project to minimize power consumption in heavy industrial facilities. Energy consultants are testing Dust Inc.’s system. They’re using the sensor networks to gauge energy loss at various points in a building, making the sensor network a tool for cutting energy costs and alerting the landlord when the system needs maintenance, says Eric Kaufmann, a vice president and partner in In-Q-Tel’s Menlo Park office.

While Dust is in the second phase of beta trials with customers, it has already secured a contract manufacturer in Silicon Valley and plans to ship as many of the devices as it can sell by the summer of this year. It plans to use its recent infusion of cash to build a new generation of products that consume less power and cost less.

Pister is a pioneer in smart dust research. He began working in the field in 1992 as an assistant professor at UCLA. There he made his first attempt to build a sensor that was no bigger than 1 cubic millimeter, a goal that remains elusive. He left for UC Berkeley in 1997, but by then he had secured DARPA funding to build a wireless network of sensors. He was able to make sensors only 4.8 mm long, the size of a grain of rice, that could communicate with each other and relay information back to a gateway using wireless mesh networking technology.

That’s the technology Dust Inc. is founded on. It was enough to attract the attention of angel investors like Gordon Bell of Microsoft Research, Dick Atkinson, the former president of the University of California, Pierre Omidyar, eBay’s founder, and Joel Birnbaum, Hewlett-Packard’s senior technical advisor. They put $1 million into the company when Pister left his faculty position to found the company with four others.

Pister is not the only one working on smart dust technology. There’s research going on at other UC Berkeley labs, at MIT and at UCLA. Corporations like Intel, the ones that will do the number crunching and data transmission to build wireless sensor networks, are also want in on the action.

For Intel, sponsoring smart dust research is about getting in early on whatever shape next-generation computing will take. If sensor networks do take off, Intel wants to be there, with specially designed microprocessors and software to monitor the networks. “We want to be there to shape it any way we can,” says Robert Manetta, an Intel spokesman. It’s sitting on the sidelines, though, waiting to see how big the market for wireless sensor networks is, if it exists at all.

Venture capitalists have taken a riskier strategy in financing smart dust companies. The technology is cool, no doubt, but none of the startups has made a convincing argument as to why factories and assembly lines and the rest already wired with sensors should replace their existing technology with smart dust. Until the startups prove that smart dust is better, faster and cheaper, the technology will remain in the fantasy labs at DARPA-or watering the lawns of Silicon Valley’s gadget geeks.