Published by CilderGroup
INAUGURAL ISSUE
We are entering the century of the electron. Not the information century? The communications age? The bit era?
Well bits are electrons: small buckets of them stored in silicon capacitors, or propelled through metal wire, or (transformed into photons) oscillating through glass or air. Bits are, at bottom, packets of energy that have to be sifted, herded, and propelled across planes of silicon and through tunnels of copper, coax, and glass. This takes electricity: flows of electrons.
Not much electricity-not for just one hit. But the number of bits in motion is growing at big-bang speeds. One result is that while electricity accounted for 25% of our energy consumption 25 years ago, it accounts for 37% today. It will account for more than half U.S. energy use by early in the next century. Most of those additional electrons will flow into information devices. But far more important than the sheer increase in the volume of electrons demanded by information technology is the type of electricity the information economy requires. Bits demand unusually clean, stable, reliable electrons. Electrons for bits cannot he reliably provided by the same old technologies on the same old power grid that powers our light bulbs, electric motors, or air conditioners, or at the same old price.
To accommodate this great energy shift, much of the sprawling infrastructure of the U.S. power grid will have to be rebuilt. Unlikely though it may seem, your century-old power company-stolid and plodding, funded by ratepayer bonds and entangled in sclerotic commissions - is now hitched to the dot-corns. It may prosper with them, or it may end up as road kill, but either way its destiny is now linked to theirs.
Hundreds of billions of dollars per year are going to
he invested in new technologies to move, condition, store, and distribute electrons. The
companies that will do this range from the familiar to the unknown: Caterpillar, Cummins,
GE, ABB, American Superconductor, Siemens Westinghouse, Silicon Power, American Power
Conversion, Generac, Hitachi, Active Power, Cutler Hammer, Williams, Capstone, Allied
Signal, and dozens - or if we include component makers and utilities themselves,
hundreds-more.
The New Chip
Compare two silicon chips, side by side. One-call it SmartChip-contains 100 million gates.
Each gate operates at one microwatt of power; the entire chip consumes 100 watts. The
other-cAll it Powerchip - contains just one, mammoth gate. But it is big enough to switch
a megawatt.
SmartChip is, of course, the building block of the lnternet
Economy, spanning companies like Cisco, lntel, AOL, and Microsoft; Sun and IBM; E-trade,
Yahoo!, and Cnet; Amazon, and eToys and WSJ.com. SmartChip has taken apart and put back
together mainframes and micros, switches and routers, banks and brokerage houses,
book-stores and newspapers, radio stations and televisions.
And PowerChip? It is now poised to take apart and put back together the trillion-dollar
U.S. network of central power stations and transmission distribution lines, and the $500
billion-a-year kilowatt-hour economy.
PowerChip is in fact the older sibling in the solid state family. Selenium diodes, used as
switches in power supplies and amplifiers, entered commercial production in the 1950s.
lnternational Rectifier - a founding company of solid-state technology - went public in
1958. As IR's founder observes, "selenium diodes begot germanium diodes, which in
turn begot silicon diodes, which then resulted in commercial transistors and thyristors,
and then begot ICs which begot memory and microprocessors."
Though PowerChip was the first born, SmartChip grew up a lot faster, propelled up the
steep curve of Moore's Law by technological advances that etch ever-finer gates, pushing
circuits ever closer together, ramping up processing speeds even while pushing down power
requirements. As our colleague George Gilder summarizes in the law of the Microcosm,
"the less the space, the more the room."
PowerChip lives in a different world altogether. For PowerChip the goal is not to make
ever smaller gates requiring less and less power, but ever bigger ones that can switch
more and more power, faster and faster. Denied all the rich paradoxes of miniaturization
that drove the silicon learning curve for SmartChip, PowerChip technology has proved even
more challenging. PowerChip does exactly what a transistor does, uses a smaller current to
switch a larger one. But in this case, a much larger one. That turns out to be a very big
difference indeed.
The millions of tiny gates in a microprocessor perform their switching at micro-watt power
levels. A PowerChip switching "gate,"which can switch an entire office building
from the utility grid to a backup power source, has to handle kilowatts, or even
megawatts. Power-wise, that's about the difference between a hang glider and the space
shuttle. At that power density the electrons literally push the atoms of the conductor
around and create their own circuits. If you let them.
With a PowerChip, speed is everything. The slower your PowerChip switches, the more heat
gets generated while it's flipping from open to shut. Ten milliseconds is no big deal when
you're switching 10 watts. In a 10 MW switch, however, a 10 ms switching time means your
chip is toast. Today's high-power thyristors meet 1,000 V/microsecond and 400
Amps/microsecond standards. That's fine for switching 6 MW, but still chump change
compared to the power levels in the backbone of the electric power grid.
To handle higher power levels, you pile 8 kV/8 MW silicon thyristors into a very tall
stack-up to 50 feet high, with lots of extra space for heat sink and cooling. Now, you're
set to switch 80 MW - if you can fire an those thyristors exactly simultaneously. But if
one of those stacked thyristors goes off even microseconds early, it takes all the power
by itself, and explodes.
The pre-history of the PowerChip spanned the two decades between 1975 and 1995. During
that period, power ratings of individual PowerChips barely doubled. Nevertheless over the
past ten years, the market has grown at a compound annual rate of better than 15%, until
by 1997 power conversion semiconductors comprised an $8 billion market.
In the late 1990s, advances in PowerChip technology began to accelerate very rapidly.
Today, PowerChip technology is crouched at the sweet spot of its learning curve, at about
the same point lntel occupied in its business in 1979: poised for three-digit annual
growth, with production volumes doubling or redoubling inside a year, yielding
Moore's-Law-like accelerations in switching capacity per dollar. As that happens,
PowerChips will seize control of the MWs, as inevitably as SrnartChips seized control of
Mips.
Today only about 12% of the world's electricity is switched by PowerChip. But 100%
penetration is inevitable, as PowerChip accelerates up the learning curve. And the most
crucial, and profitable, market will be satisfying the power requirements of the
SmartChip. The PowerChip is the one technology that can satisfy the surging demand for
something altogether new in the power business: "Ten Nines" of reliability.
The Tenth Nine
A remarkable number of bad things can happen to a power svstem woven out of tens of
thousands of miles of lines: "car-tree interactions," and Mother Nature being
just two. Solar electric storms induce huge currents in the grid's long wires. Such storms
follow a predicuble 11 year sunspot cycle. The last peak, which began at the end of
1989-back in pre-Cambrian lnternet tirne - put six million people in the dark on Quebec
Hydro's systems. The next peak starts in 2001.
Worse yet, the network is as much a part of the problem as part of the solution. Every
time a big motor starts up at a water plant, or an electric welder fires up, power spikes
surge and ripple up the grid. A few years. back a Stanford computer center found its power
fatally