There is some very promising research that promises the development of a new kind of transistor.

A team of Duke University chemists has modified a method for growing long, straight, numerous and well-aligned carbon cylinders only a few atoms thick that paves the way for manufacturing reliable electronic nanocircuits.
The team had already described a method last April for growing the crystals, but the modification is targeted at making a process specifically for producing semiconducting versions of the single-walled carbon nanotubes, sometimes called “buckytubes” because their ends, when closed, take the form of soccer ball-shaped carbon-60 molecules known as buckminsterfullerines, or “buckyballs”.
The effort is being led by Jie Liu, Duke’s Jerry G. and Patricia Crawford Hubbard professor of chemistry.
“I think it’s the holy grail for the field,” Liu said. “Every piece is now there, including the control of location, orientation and electronic properties all together. We are positioned to make large numbers of electronic devices such as high-current field-effect transistors and sensors.”
A report on their achievement, co-authored by Liu and a team of collaborators from his Duke laboratory and Peking University in China, has just been published in the research journal Nano Letters.

What does this portend? Well quite a few things actually. Carbon Nano Tubes (CNTs) are five times as conductive as copper, electron mobility is about 70 times that of silicon and it should be able to withstand much higher temperatures than silicon without losing its semiconducting properties. Not only that, the material is abundant. So once the manufacturing process is worked out it will mean high power, low loss, extremely high speed transistors.
How soon you ask? First off not all the bugs have been worked out in the laboratory models.

That earlier JACS report described how the researchers coaxed nanotubes to form in long, parallel paths that will not cross each other to impede potential electronic performance. Their method grows the nanotubes on a template made of a continuous and unbroken kind of single quartz crystal used in electronic applications. Copper is also used as a growth promoter.
But that method left one unresolved issue blocking the use of such nanotubes as electronic components. Only some of the resulting nanotubes acted electronically as semiconductors. Others were the electronic equivalent of metals. To work in transistors, the nanotubes must all be semiconducting, Liu said.
The researchers now say they have achieved virtually all-semiconductor growth conditions by making one modification.
In their earlier work they had used the alcohol ethanol in the feeder gas to provide carbon atoms as building blocks for the growing nanotubes. In the new work, they describe how they tried various ratios of two alcohols–ethanol and methanol–combined with two other gases they also used previously–argon and hydrogen.
“We found that by using the right combination of the two alcohols with the argon and hydrogen we could grow exclusively semiconducting nanotubes,” Liu said. “It was like operating a tuning knob.” The inert argon gas was used to provide a steady feed of the ethanol and methanol, with hydrogen to keep the copper catalyst from oxidizing.
After making the nanotubes by chemical vapor deposition in a small furnace set to a temperature of 900°C, the researchers assembled some of them into field-effect transistors to test their electronic properties.
“We have estimated from these measurements that the samples consisted of 95 to 98 percent semiconducting nanotubes,” the researchers reported.

Now that is probably good enough for first generation transistors in some applications if those kind of numbers can be achieved in production. However what you want for general use is 99.9% or 99.99% semiconducting CNTs. The more nines the better. So how soon? I’d say pilot production in five years, and full scale production (10s of millions of devices) in about eight years. Fortunately it builds on the base of silicon semiconductor production so the equipment needed is likely to be very similar to what is already in use.
The best power conversion equipment we have using silicon has efficiencies topping out at around 95% with the more typical units running at 85% to 90% efficiency. With these new devices we could reach 99% or better. They could also mean 20 times faster computers that use 1/10th as much power as current devices. Considering that we already have chips on the market that deliver 25,000 MIPS for 360 milliwatts that would be something. It would be roughly equivalent to 1,000 Cray 1s in your pocket that could be powered from an AA cell for a month. Cell phones could run for weeks on a charge. Laptops that could run for many days. Faster please.
Cross Posted at Power and Control