Unbounding the Future

The Nanotechnology Revolution

Eric Drexler and Chris Peterson, with Gayle Pergamit William Morrow and Company, Inc., New York

© 1991 by K. Eric Drexler, Chris Peterson, and Gayle Pergamit. All rights reserved.

Chapter 7: The Spiral of Capability

In earlier chapters, we have stepped forward and backward through time. The last step was a big one, leaping from small laboratory devices to the high-capacity industrial facility of the Desert Rose scenario. Our narrative crossed this gap in a single leap, but the world won’t. To understand how nanotechnology might unfold, it makes sense to look at some of its easier and more difficult applications. The result won’t be a timetable, or even a series of milestones, but it should give a better picture of what we can expect as nanotechnology develops from simple, crude, costly beginnings to a state of greater sophistication and lower cost.

Improving Quality

Molecular manufacturing will make better products possible. We’re likely to see some early applications in at least two areas: stronger materials and faster computers. Strong materials are simple, and will be hard to pass up. Computers are more complex, but the payoff will be enormous.

Computers

The computer industry has been under steady pressure to make computer chips ever smaller. As sizes have shrunk, costs have fallen while efficiency and capabilities have increased. The pressure to continue this process pushes in the direction of nanotechnology; it may even be one of the major motivations behind developing the technology.

John Walker, a founder of Autodesk, explains: “Even technologies with enormous potential can lie dormant unless there are significant payoffs along the way to reward those who pioneer them. That’s one of the reasons integrated circuits developed so rapidly; each advance found an immediate market willing to apply it and enrich the innovator that created it.

“Does molecular engineering have this kind of payoff? I think it does. Remembering that we may be less than ten years away from ‘hitting the wall’ as far as scaling our existing electronics goes, a great deal of research is presently going on in the area of molecular and quantum electronics. The payoff is easy to calculate: You can build devices one thousand times faster, more energy-efficient, and cheaper than those we’re currently using—at least one hundred times better than exotic materials being considered to replace silicon when it reaches its limits.”

Federico Capasso, head of the Quantum Phenomena and Device Research Department at AT&T Bell Labs, agrees that electronics researchers will keep pushing for smaller devices once silicon’s potential has been reached. He explains that “at some point we will reach difficulties: some people say at a hundred fifty nanometers, others think it’s beyond that. What will happen then? It’s hard to think that the electronics industry will say, ‘Stop here. We’ll stop evolving because we can’t shrink the device.’ From an economic point of view, in order to survive, an industry has to innovate continuously.”

The computer industry’s push toward devices of molecular size has an air of inevitability. Today’s researchers struggle to build molecular electronics using bulk techniques, with no products yet in sight; with molecular manipulators, they will finally have the tools they need for fast and accurate experimentation. Once successful designs are developed, packaged, and tested, the pressure will be on to learn to make them in quantity at low cost. The competitive pressures will be fierce, because advanced molecular electronics will be orders of magnitude better than today’s integrated circuits, ultimately enabling the construction of computers with trillionfold greater capability.

Strong, Lightweight Structures

At the opposite extreme from molecular electronics—complex and at first worth billions of dollars per gram—are structural materials: worth only dollars per kilogram in most applications, but much simpler in structure. Once molecular manufacturing becomes inexpensive, structural materials will be important products.

These materials play a central role in almost everything around us, from cars and aircraft to furniture and houses. All of these objects get their size, shape, and strength from a structural skeleton of some sort. This makes structural materials a natural place to begin in understanding how nanotechnology can improve products.

Cars today are mostly made of steel, aircraft of aluminum, and buildings and furniture largely of steel and wood. These materials have a certain “strength-to-weight ratio” (more properly, a strength-to-density ratio). To make cars stronger, they’d have to be heavier; to make them lighter, they’d have to be weaker. Clever design can change this relationship a little, but to change it a lot requires a change of materials.

Making something heavy is easy: just leave a hollow space, then fill it with water, sand, or lead shot. Making something light and strong is harder, but often important. Automakers try to make cars lightweight, aircraft manufacturers try harder, and with spacecraft manufacturers it is an obsession. Reducing mass saves materials and energy.

The strongest materials in use today are mostly made of carbon. Kevlar, used in racing sails and bulletproof vests, is made of carbon-rich molecular fibers. Expensive graphite composites, used in tennis rackets and jet aircraft, are made using pure-carbon fibers. Perfect fibers of carbon—both graphite and diamond—would be even better, but can’t be made with today’s technology. Once molecular manufacturing gets rolling, though, such materials will be commonplace and inexpensive.

What will these materials be like? To picture them, a good place to start is wood. The structure of wood can vary from extremely light and porous, like balsa wood, to denser structures like oak. Wood is made by molecular machinery in plants from carbon-rich polymers, mostly cellulose. Molecular manufacturing will be able to make materials like these, but with a strength-to-weight ratio about a hundred times that of mediocre steel, and tens of times better than the best steel. Instead of being made of cellulose, these materials will be made of carbon in forms like diamond.

Diamond is emphasized here not because it is shiny and expensive, but because it is strong and potentially cheap. Diamond is just carbon with properly arranged atoms. Companies are already learning to make it from natural gas at low pressure. Molecular manufacturing will be able to make complex objects of the stuff, built lighter than balsa wood but stronger than steel.

Products made of such materials could be startling by our present standards. Objects could be made that are identical in size and shape to those we make today, but simultaneously stronger and 90 percent lighter. This is something to keep in mind next time you’re lugging a heavy object around. (If something needs weight to hold it in place, it would be more convenient to add this ballast when the thing is in its proper location than to build in the extra weight permanently.)

Better structural materials will make aircraft lighter, stronger, and more efficient, but will have the greatest effect on spacecraft. Today, spacecraft can barely reach orbit with both a safety margin and a cargo. To get there at all, they have to drop off parts like boosters and tanks along the way, shedding weight. With strong materials, this will change: as in the space-travel-for-business scenario in Chapter 1, spacecraft will become more like aircraft are today. They will be rugged and reliable, and strong enough and light enough to reach space in one piece.

Quickening Development

In some areas of high technology—spaceflight has been a notorious example—it takes years, even decades, to try a new idea. This makes progress slow to a crawl. In other areas—software has been a shining example—new ideas can be tested in minutes or hours. Since the Space Shuttle design was frozen, personal computer software has come into existence and gone through several generations of commercial development, each with many cycles of building and testing.

Fast, Inexpensive Testing

Even in the days of the first operational molecular manipulators, experimentation is likely to be reasonably fast. Individual chemical steps can take seconds or less. Complex molecular objects could be built in a matter of hours. This will let new ideas be put into practice almost as fast as they can be designed.

Later assemblers will be even faster. At a millionth of a second per step, they will approach the speed of computers. And, as nanotechnology matures, experimenters will have more and more molecular instruments available to help them find out whether their devices work or not. Fast construction and fast testing will encourage fast progress.

At this point, the cost of materials and equipment for experiments will be trivial. No one today can afford to build Moon rockets on a hobby budget, but they can afford to build software, and many useful programs have been the result. There is no economic reason why nanomachines couldn’t eventually be built with a hobby-size budget, though there are reasons—to be discussed in later chapters—for wanting to place limits on what can be built.

Early Simplicity

Finally, established technologies are always pushing up against some limit; the easy opportunities have generally been exploited. In many fields, the limits are those of the properties of the materials used and the cost and precision of manufacturing. This is true for computers, for spacecraft, for cars, blenders, and shoes. For software, the limits are those of computer capacity and of sheer complexity (which is to say, of human intelligence). After molecular manufacturing develops certain basic abilities, a whole set of limits will fall, and a whole range of developments will become possible. Limits set by materials properties, and by the cost and precision of manufacturing, will be pushed way back. Competition, easy opportunities, and fast, low-cost experimentation should combine to yield an explosion of new products.

	Space	Computers	Nanotechnology
Precursor science and technologies	Physics Sounding rockets	Mathematics Electronics	Theoretical chemistry Chemical synthesis
Crucial advance	Teams combine and improve technologies	Teams combine and improve technologies	Teams combine and improve technologies
Threshold capability	First satellite	First computer	First assembler
Early practical applications	Weather, spy and communication satellites	Scientific calculations Payroll calculations	Molecular sensors Molecular computing
Breakthrough capability	Routine, inexpensive spaceflight	Powerful mass-market desktop computers	Powerful inexpensive molecular manufacturing
Further projected developments	Lunar base, Mars exploration	Widespread electronic publishing	New medical abilities New, inexpensive products
More advanced developments	Mining, development, settlement of solar system	Major automation of engineering design	Help with computer goals Environmental cleanup
Yet more advanced developments	Interstellar flight and settlement feasible	Trillionfold computer power	Help with computer goals General tissue repair

This does not mean immediately, and it does not apply to all imaginable nanotechnologies. Some technologies are imaginable and clearly feasible, yet dauntingly complex. Still, the above considerations suggest that a wide range of advances could happen at a brisk pace. The main bottleneck might seem to be a shortage of knowledgeable designers—hardly anyone knows both chemistry and mechanical design—but improving computer simulations will help. These simulations will let engineers tinker with molecular-machinery designs, absorbing knowledge of chemical rules without learning chemistry in the usual sense.

Climbing Complexity

Making familiar products from improved materials will increase their safety, performance, and usefulness. It will also present the simplest engineering task. A greater change, though, will result from unfamiliar products made possible by new manufacturing methods. In talking about unfamiliar products, a hard-to-answer question arises: What will people want?

Products are typically made because their recipients want them. In our discussions here, if we describe something that people won’t want, then it probably won’t get built, and if it does get built, it will soon disappear. (The exceptions—fraud, coercion, persistent mistakes—are important, but in other contexts.) To anchor our discussion, it makes sense to look not at totally new products, but instead at new features for old products, or new ways to provide old services. This approach won’t cover more than a fraction of what is possible, but will start from something sensible and provide a springboard for the imagination.

As usual, we are describing possibilities, not making predictions. The possibilities focused on here arise from more complex applications of molecular manufacturing—nanotechnological products that contain nanomachines when they are finished. Earlier, we discussed strong materials. Now, we discuss some smart materials.

Smart Materials

The goal of making materials and objects smart isn’t new: researchers are already struggling to build structures that can sense internal and environmental conditions and adapt themselves appropriately. There is even a Journal of Intelligent Material Systems and Structures. By using materials that can adapt their shapes, sometimes hooked up to sensors and computers, engineers are starting to make objects they call “smart.” These are the early ancestors of the smart materials that molecular manufacturing will make possible.

Today, we are used to having machines with a few visible moving parts. In cars, the wheels go around, the windshield wipers go back and forth, the antenna may go up and down, the seat belts, mirrors, and steering wheel may be motor-driven. Electric motors are fairly small, fairly inexpensive, and fairly reliable, so they are fairly common. The result is machines that are fairly smart and flexible, in a clumsy, expensive way.

In the Desert Rose scenario, we saw “tents” being assembled from trillions of submicroscopically small parts, including motors, computers, fibers, and struts. To the naked eye, materials made from these parts could seem as smooth and uniform as a piece of plastic, or as richly textured as wood or cloth—it is all a matter of the arrangement and appearance of the submicroscopic parts. These motors and other parts cost less than a trillionth of a dollar apiece. They can be quite reliable, and good design can make systems work smoothly even if 10 percent of a trillion motors burn out. Likewise for motor-controlling computers and the rest. The resulting machines can be very smart and flexible, compared to those of today, and inexpensive, too.

When materials can be full of motors and controllers, whole chunks of material can be made flexible and controllable. The applications should be broad.

Scenario: Smart Paint

Surfaces surround us, and human-made surfaces—walls, roofs, and pavement—cover huge areas that matter to people. How can smart materials make a difference here?

The revolution in technology has come and gone, and you want to repaint your walls. Breathing toxic solvents and polluting water by washing brushes have passed into history, because paint has been replaced with smarter stuff. The mid-twentieth century had seen considerable progress in paints, especially the development of liquids that weren’t quite liquid—they would spread with a brush, but didn’t (stupidly) run and drip under their own weight. This was an improvement, but the new material, “paperpaint,” is even more cooperative.

Paperpaint comes in a box with a special trowel and pen. The paperpaint itself is a dry block that feels a lot like a block of wood. Following the instructions, you use the pen to draw a line around the edge of the area you want to paint, putting an X in the middle to show where you want the paint to go on; the line is made of nontoxic disappearing ink, so you can slop it around without staining anything. Using the trowel, you slice off a hunk of paperpaint—which is easy, because it parts like soft butter to the trowel, even though it behaves like a solid to everything else. Very high IQ stuff, that.

Next, you press the hunk against the X and start smoothing it out with the trowel. Each stroke spreads a wide swath of paperpaint, much wider than the trowel, but always staying within the inked line. A few swipes spreads it precisely to the edges, whereupon it smooths out into a uniform layer. Why doesn’t it just spread itself? Experience showed that customers didn’t mind the effort of making a few swipes and preferred the added control.

The paperpaint consists of a huge number of nanomachines with little wheels for rolling over one another and little sticky pads for clinging to surfaces. Each has a simple, stupid computer on board. Each can signal its neighbors. The whole mass of them clings together like an ordinary solid, but they can slip and slide in a controlled way when signaled. When you smooth the trowel over them, this contact tells them to get moving and spread out. When they hit the line, this tells them to stop. If they don’t hit a line, they go a few handbreadths, then stop anyway until you trowel them again. When they encounter a line on all sides, word gets around, and they jostle around to form a smooth, uniform layer. Any that get scraped off are just so much loose dust, but they stick together quite well.

This paint-stuff doesn’t get anything wet, doesn’t stain, and clings to surfaces just tightly enough to keep it from peeling off accidentally. If some experimentally minded child starts digging with a stick, makes a tear, and peels some off, it can be smoothed back again and will rejoin as good as new. The child may eat a piece, but careful regulation and testing has ensured that this is no worse than eating plain paper, and safer than eating a colorful Sunday newspaper page.

Many refinements are possible. Swipes and pats of the trowel could make areas thicken or thin, or bridge small holes (no more Spackling!). With sufficiently smart paperpaint, and some way to indicate what it should do, you can have your choice of textures. Any good design will be washable, and a better design would shed dirt automatically using microscopic brushes.

Removal, of course, is easy: either you rip and peel (no scraping needed), or find that trowel, set the dial on the handle to “strip,” and poke the surface a few times. Either way, you end up with a lump ready to pitch into the recycling bin and the same old wall you started with, bared to sight again.

Power Paint

Perhaps no product will ever be made exactly like the smart paint just described. It would be disappointing if something better couldn’t be made by the time smart paint is technologically possible. Still, paperpaint gives a feel for some of the features to expect in the new smart products, features such as increased flexibility and better control. Without loading yet more capability into our paint (though there is no reason why one couldn’t), let’s take a look at some other smart properties one might want in a surface.

External walls, roofs, and paving surfaces are exposed to sunlight, and sunlight carries energy. A proven ability of molecular machinery is the conversion of sunlight to stored energy: plants do it every day. Even now, we can make solar cells that convert sunlight into electricity at efficiencies of 30 percent or so. Molecular manufacturing could not only make solar cells much cheaper, but could also make them tiny enough to be incorporated into the mobile building blocks of a smart paint.

To be efficient, this paint would have to be dark—that is, would have to absorb a lot of light. Black would be best, but even light colors could generate some power, and efficiency isn’t everything. Once the paint was applied, its building blocks would plug together to pool their electrical power and deliver it through some standard plug. A thicker, tougher form of this sort of material could be used to resurface pavement, generate power, and transmit it over large distances. Since smart solar-cell pavement could be designed for improved traction and a similar roofing material could be designed for amazing leak-resistance, the stuff should be popular.

On a sunny day, an area just a few paces on a side would generate a kilowatt of electrical power. With good batteries (and enough repaved roads and solar-cell roofing), present demands for electrical power could be met with no coal burning, no oil imports, no nuclear power, no hydroelectric dams, and no land taken over for solar power generation plants.

Pretty Paint, Acoustic Paint

The glow of fireflies and deep sea fish shows that molecular devices can convert stored chemical energy into light. All sorts of common devices show that electricity can be converted to light. With molecular manufacturing, this conversion can be done in thin films, with control over the brightness and color of each microscopic spot. This could be used for diffuse lighting—ceiling paperpaint that glows. With more elaborate control, this would yield the marvel (horror?) of video wallpaper.

With today’s technology, we are used to displays that glow. With molecular manufacturing, it will be equally easy to make displays that just change color, like a printed page with mobile ink. Chameleons and flatfish change color by moving colored particles around, and nanomachines could do likewise. On a more molecular level, they could use tunable dyes. Live lobsters are a dark grayish green, but when cooked turn bright red. Much of this change results from the “retuning” of a dye molecule that is bound in a protein in the live lobster but released by heat. This basically mechanical change alters its color; the same principle can be used in nanomachines, but reversibly.

How a surface appears depends on how it reflects or emits light. Nanomachines and nanoelectronics will be able to control this within wide limits. They will be able to do likewise for sound, by controlling how a surface moves. In a stereo system, a speaker is a movable surface, and nanomachines are great for making things move as desired. Making a surface emit high-quality sound will be easy. Almost as easy will be surfaces that actively flex to absorb sound, so that the barking dog across the street seems to fade away.

Smart Cloth

Looking further at the human environment we find a lot of cloth and related materials, such as carpeting and shoes. The textile industry was at the cutting edge of the first industrial revolution, and the next industrial revolution will have its effects on textiles.

With nanotechnology, even the finest textile fibers could have sensors, computers, and motors in their core at little extra cost. Fabrics could include sensors able to detect light, heat, pressure, moisture, stress, and wear, networks of simple computers to integrate this data, and motors and other nanomechanisms to respond to it. Ordinary, everyday things like fabric and padding could be made responsive to a person’s needs—changing shape, color, texture, fit, and so forth—with the weather and a person’s posture or situation. This process could be slow, or it could be fast enough to respond to a gesture. One result would be genuine one-size-fits-all clothing (give or take child sizes), perfectly tailored off the rack, warm in winter, cool and dry in summer; in short, nanotechnology could provide what advertisers have only promised. Even bogus advertising gives a clue to human desires.

Throughout history, the human race has pursued the quest for comfortable shoes. With fully adjustable materials, the seemingly impossible goal of having shoes that both look good and feel good should finally be achieved. Shoes could keep your feet dry, and warm except in the Arctic, cool except in the tropics, and as comfortable as they can be with a person stepping on them.

Smart Furniture

Adaptive structures will be useful in furniture. Today, we have furniture that adapts to the human body, but it does so in an awkward and incomplete manner. It adapts because people grab cushions and move them around. Or a chair adapts because it is a hinged contraption that grudgingly bends and extends in a few places to suit a small range of preferred positions. Occasionally, one sees furniture that allegedly gives a massage, but in fact only vibrates.

These limitations are consequences of the expense, bulkiness, clumsiness, and unreliability of such things as moving parts, motors, sensors, and computers today. With molecular manufacturing, it will be easy to make furniture from smart materials that can adapt to an individual human body, and to a person’s changing position, to consistently give comfortable support. Smart cushions could also do a better job of responding to hints in the form of pats, tugs, and punches. As for massage—a piece of furniture, no matter how advanced, is not the same as a masseuse. Still, a typical massage setting on a smart chair would not mean today’s “vibrate medium vigorously,” but something closer to “five minutes of shiatsu.”

And So Forth . . .

This tour through of the potential of smart matter has shown how we could get walls that look and sound as we wish, clothing, shoes, and furniture of greater comfort, and clean solar power. As one might expect, this just scratches the surface.

If you care to think of further applications, here are some ground rules: Components made by molecular manufacturing can be many tens of times stronger than steel, but materials made by plugging many components together will be weaker. For these, strengths in the range of cotton candy to steel seem achievable. The components will be sensitive to heat, and at high temperatures they will break down or burn. Many materials will be able to survive the temperature of boiling water, but only specialized designs would be oven-safe. Color, texture, and (usually) sound should be controllable. Surfaces can be smooth and tightly sealed (this takes some cleverness). Motions can be fairly fast. Power has to come from somewhere; good sources include electricity, stored chemical energy, and light. If nanomachines or smart materials are dunked in liquids, chemical energy can come from dissolved molecules; if they are in the open, energy can come from light; if they are sitting in one place, they can be plugged into a socket; if they are moving around in the dark, they can run on batteries for a while, then run down and quit. Within these limits, much can be accomplished.

“Smart” is a relative term. Unless you want to assume that people learn a lot more about intelligence and programming, it is best to assume that these materials will follow simple rules, like those followed by parts of drawings on computer screens. In these drawings, a picture of a rectangle can be commanded to sprout handles at its corners; pulling a handle stretches or shrinks the rectangle without distorting its right-angle corners. An object made of smart matter could do likewise in the real world: a box could be stretched to a different size, then made rigid again; a door in a smart-material wall could have its position unlocked, its frame moved a pace to the left, and then be returned to normal use.

There seems little reason to make bits of smart matter independent, self-replicating, or toxic. With care, smart matter should be safer than what it replaces because it will be better controlled. Spray paint gets all over things and contains noxious solvents; the paperpaint described above doesn’t. This will be a characteristic difference, if we exercise our usual vigilance to encourage the production of things that are safe and environmentally sound.

Falling Costs

It may be fun to discuss wondrous new products, but they won’t make much difference in the world if they are too expensive. Besides, many people today don’t have decent food, clothes, and a roof over their heads, to say nothing of fancy “nanostuff.”

Costs matter. There is more to life than material goods, but without material goods life is miserable and narrow. If goods are expensive, people strive for them; if goods are abundant, people can turn their attention elsewhere. Some of us like to think that we are above a concern for material goods, but this seems more common in the wealthy countries. Lowering manufacturing costs is a mundane concern, but so are feeding people, housing them, and building sewage systems to keep them from dying of cholera and hepatitis. For all these reasons, finding ways to bring down production costs is a worthy goal.

For the poor, for the environment, and for the freeing of human potential, costs matter deeply. Let’s take a closer look at the costs of molecular manufacturing.

Can falling costs be realistic?

Inflation produces the illusion that costs rise, when the real story is that the value of money is falling. In the short term, real costs usually don’t change very quickly, and this can produce the illusion that costs are stable facts of nature, like the law of gravity or the laws of thermodynamics.

In the real world, though, most costs have been falling by a crucial measure: the amount of human labor needed to make things. People can afford more and more, because their labor, supplemented by machines, can produce more and more. This change is dramatic measured on a scale of centuries, and equally dramatic across the gulf between Third World and developed countries. The rise from Third World to First World standards of living has raised income (dropped the cost of labor time) by more than a factor of ten. What can molecular manufacturing do?

Larger cost reductions have happened, most dramatically in computers. The cost of a computer of a given ability has fallen by roughly a factor of 10 every seven years since the 1940s. In total, this is a factor of a million. If automotive technologies had done likewise, a luxury car would now cost less than one cent. (Personal computer systems still cost hundreds of dollars both because they are far more powerful than the giant machines of the 1940s and because the cost of buying any useful computer system includes much more than just the cost of a bare computer chip.)

Costs: A First Estimate

Some costs apply to a kind of product, regardless of how many copies of it are made: these include design costs, technology licensing costs, regulatory approval costs, and the like. Other costs apply to each unit of a product: these include the costs of labor, energy, raw materials, production equipment, production sites, insurance, and waste disposal. The per-kind costs can become very low if production runs are large. If these costs stay high, it will be because people prefer new products for their new benefits, despite the cost—hardly cause for complaint.

The more basic and easier to analyze costs are per-unit costs. A picture to keep in mind here is of Desert Rose Industries, where molecular machinery does most of the work, and where products are made from parts that are ultimately made from simple chemical substances. Let’s consider some cost components.

Energy: Manufacturing at the molecular scale need not use a lot of energy. Plants build billions of tons of highly patterned material every year using available solar energy. Molecular manufacturing can be efficient, in the sense that the energy needed to build a block of product should be comparable to the energy released in burning an equivalent mass of wood or coal. If this energy were supplied as electricity at today’s costs, the energy cost of manufacturing would be something like a dollar per kilogram. We’ll return to the cost of energy later.

Raw Materials: Molecular manufacturing won’t need exotic materials as inputs. Plain bulk chemicals will suffice, and this means materials no more exotic than the fuels and feedstocks that are, for now, derived from petroleum and biomass—gasoline, methanol, ammonia, and hydrogen. These typically cost tens of cents per kilogram. If bizarre compounds are used, they can be made internally. Rare elements could be avoided, but might be useful in trace amounts. The total quantity of raw materials consumed will be smaller than in conventional manufacturing processes because less will be wasted.

Capital Equipment and Maintenance: As we saw in the Desert Rose scenario, molecular manufacturing can be used to build all of the equipment needed for molecular manufacturing. It seems that this equipment—everything from large vats to submicroscopic special-purpose assemblers—can be reasonably durable, lasting for months or years before being recycled and replaced. If the equipment were to cost dollars per kilogram, and produce many thousands of kilograms of product in its life, the cost of the equipment would add little to the cost of the product.

Waste Disposal: Today’s manufacturing waste is dumped into the air, water, and landfills. There need be no such waste with molecular manufacturing. Excess materials of the kind now spewed into the environment could instead be completely recycled internally, or could emerge from the manufacturing process in pure form, ready for use in some other process. In an advanced process, the only wastes would be leftover atoms resulting from a bad mix of raw materials. Most of these leftover atoms would be ordinary minerals and simple gases like oxygen, the main “waste” from the molecular machinery of plants. Molecular manufacturing produces no new elements—if arsenic comes out, arsenic must have gone in, and the process isn’t to blame for its existence. Any intrinsically toxic materials of this sort can at least be put in the safest form we can devise for disposal. One option would be to chemically bond it into a stable mineral and put it back where it came from.

Labor: Once a plant is operating, it should require little human labor (what people do with their time will change, unless factories are kept running as bizarre hobbies). Desert Rose Industries was run by two people, yet was described as producing large quantities of varied goods. The basic molecular-scale operations of manufacturing have to be automated, since they are too small for people to work on. The other operations are fairly simple and can be aided by equipment for handling materials and information.

Space: Even a manufacturing plant based on nanotechnology takes up room. It would, however, be more compact than familiar manufacturing plants, and could be built in some out-of-the-way place with inexpensive land. These costs should be small by today’s standards.

Insurance: This cost will depend on the state of the law, but some comparisons can be made. Improved sensors and alarms could be made integral parts of products; these should lower fire and theft premiums. Product liability costs should be reduced by safer, more reliable products (we’ll discuss the question of product safety further in Chapter 12). Employee injury rates will be reduced by having less labor input. Still, the legal system in the United States has shown a disturbing tendency to block every new risk, however small, even when this forces people to keep suffering old risks, which are sometimes huge. (The supply of lifesaving vaccines has been threatened in just this way.) When this happens, we kill anonymous people in the name of safety. If this behavior raises insurance premiums in a perverse way, it could discourage a shift to safer manufacturing technologies. Since such costs can grow or shrink independent of the real world of engineering and human welfare, they are beyond our ability to estimate.

Sales, Distribution, Training . . .: These costs will depend on the product: Is it as common as potatoes, and as simple to use? Or is it rare and complex, so that determining what you need, where to get it, and how to use it are the main problems? These service costs are real but can be distinguished from costs of the thing itself.

To summarize, molecular manufacturing should eventually lead to lower costs. The initial expense of developing the technology and specific products will be substantial, but the cost of production can be low. Energy costs (at present prices) and materials costs (ditto) would be significant, but not enormous. They were quoted on a per-kilogram basis, but nanotechnological products, being made of superior materials, will often weigh only a fraction of what familiar products do. (Ballast, were it needed, will be dirt-cheap.) Equipment costs, land costs, waste-disposal costs, and labor costs can be low by the very nature of the technology.

Costs of design, regulation, and insurance will depend strongly on human tastes and are beyond predicting. Basic products, like clothing and housing, can become inexpensive unless we do something to keep them costly. As the cost of improved safety falls, there will be less reason to accept unsafe products. Molecular manufacturing uses processes as controlled and efficient as the molecular processes in plants. Its products could be as inexpensive as potatoes. This may sound to good to be true (and there are downsides, as we’ll discuss), but why shouldn’t it be true? Shouldn’t we expect large changes to come with the replacement of modern technology?

A Cycle of Falling Costs

The above estimate made a conservative assumption about future costs: that energy and materials will cost then what they do now, before molecular manufacturing has become available. They won’t, because lower costs lead to lower costs.

Let’s say that making one kilogram of product by molecular manufacturing requires one dollar for a kilogram of raw materials and four dollars for a generous forty kilowatt-hours of energy. These are typical present-day prices for materials and electrical energy. Assume, for the moment, that other costs are small. One of the resulting five-dollar-per-kilogram products can be solar cell paint suitable for applying to paved roads. A layer of paint a few millionths of a meter thick would cost about five cents per square meter to produce, and would generate enough energy to make another square meter of paint in less than a week, even allowing for nighttime and moderate cloud cover. The so-called energy payback time would thus be short.

Let’s assume that this smart paint costs as much to spread and hook up as it does to make, and that we demand that it pay for itself in a single month, so we charge ten cents per square meter per month. At that rate, the cost of solar energy from resurfaced roads would be roughly $0.004 per kilowatt hour—less than a twentieth the energy cost assumed in the initial production-cost estimate. By itself, this makes the cost of production fall to a fraction of what it was before. Most of that remaining fraction consists of the cost of materials.

But the products of nanotechnology will mostly be made of carbon (if present expectations are any guide), and carbon dioxide is too abundant in the atmosphere these days. With energy so cheap, the atmosphere can be used as source of carbon (and of hydrogen, nitrogen, and oxygen). The price of carbon would be a few cents per kilogram—roughly a twentieth the original price assumed for raw materials.

But now, both energy and raw materials are a twentieth the original price, and so the products become cheaper, including the energy-producing products and the raw-material—producing (atmosphere-cleaning) products….

The above scenario is simple, but it seems realistic in its basic outlines: lower costs can lead to lower costs. How far this process can go is hard to estimate precisely, but it could go far indeed.

Power Too Cheap To Meter?

This argument will remind some readers of an old claim—that nuclear energy would lead to “power too cheap to meter.” This assertion, attributed to the early nuclear era, has passed into folklore as a warning to be skeptical of technologists promising free goodies. Does the warning apply here?

Anyone claiming that something is free doesn’t really understand economics. Using something always has a cost equal to the most valuable alternative use for the thing. Choosing one alternative sacrifices another, and that sacrifice is the cost. As economist Phillip K. Salin says, “There’s no such thing as a free opportunity,” since opportunities always cost (at least) time and attention. Nanotechnology will not mean free goodies.

But, one might argue, nuclear power hasn’t even been inexpensive. If technologists could be so wrong back then, why believe a similar argument today? We are happy to report that the arguments aren’t similar: any argument for “nuclear power too cheap to meter” had to be absurd even given the knowledge at the time, and our argument isn’t.

Nuclear reactors boil water to make steam to turn turbines to turn generators to drive electrical power through power lines to transformers to local power lines to houses, factories, and so forth. The wildest optimist could never have claimed that nuclear power was a free source of anything more than heat, and a realist would have added in the cost of the reactor equipment, fuel, waste disposal, hazards, and the rest. Even our wild optimist would have had to include the cost of building the boiler, the turbines, the generators, the power lines, and the transformers, and the cost of maintenance on all these. These costs were known to be a major part of the cost of power, so free heat wouldn’t have meant free power. Thus, the claim was absurd the day it was made—not merely in hindsight.

In the early 1960s, Alvin Weinberg, head of the Oak Ridge National Laboratory, was a strong advocate of nuclear power, and argued that it would provide “cheap energy.” He was optimistic, but did his sums. First, he assumed that nuclear-power plants could be built a little more cheaply than coal-fired power plants of the same size. Then he assumed that the cost of fuel, waste disposal, operations, and maintenance for nuclear plants would be not much more than the cost of operations and maintenance alone for coal plants. Then he assumed that they might last for more than thirty years. Finally, he assumed that they would be publicly operated, tax free at low interest (which merely moves costs elsewhere) and that after thirty years the cost of the equipment would be written off (which is an accounting fiction). With all of that, he derived a power cost that “might be” as low as one half the cost of the cheapest coal-fired plant he mentions. He was clearly an optimist, but he didn’t come close to arguing for power too cheap to meter.

Low But Not Zero Costs

People have cried “Wolf!” before about new technologies leading to overwhelming abundance. It was said of nuclear power, and of steam power before it, and perhaps of water wheels, the horse, the plough, and the chipped rock. Molecular manufacturing is different because it is a new way to make almost anything, including more of the equipment needed to do the manufacturing. There has never been anything quite like this before.

The basic argument for low cost production is this: Molecular manufacturing will be able to make almost anything with little labor, land, or maintenance, with high productivity, and with modest requirements for materials and energy. Its products will themselves be extremely productive, as energy producers, as materials collectors, and as manufacturing equipment. There has never been a technology with this combination of characteristics, so historical analogies must be used with care. Perhaps the best analogy is this: Molecular manufacturing will do for matter processing what the computer has done for information processing.

There will always be limiting costs, because resources—whether energy, matter, or design skill—always have some alternative use. Costs will not fall to zero, but it seems that they could fall very low indeed.

Gaming the Future: The Book!