Nantero’s alternative to flash memory has reached its sell-by date
Back in October 2001, a Woburn, Mass., start-up called Nantero said it was going to supplant flash memory chips. Business journalists took note: flash memory was then a US $7.8 billion market, and it was growing fast. Even more alluring, Nantero’s technology, based on carbon nanotubes, seemed to be the opening salvo in the nanotechnology revolution those journalists had been promising their readers for years.
Nantero called its prospective product the NRAM, the N standing for “Nanotube-based/Nonvolatile.” The company’s publicity people painted a picture of a computer freed from its last moving part—the hard drive—and thus capable of booting up instantly and surviving hard knocks.
What made the picture compelling was the way the technology seemed to lend itself to mass production. By relying on the properties not of individual tubes but of a mélange of them, Nantero would sidestep the material’s biggest bugbear, purity. By laying down the tiny tubes at random, as a kind of fabric, so that patches could reconfigure in response to electronic stimuli, the company would enable data to be encoded mechanically, and therefore permanently. Unlike costly static random access memory, the NRAM wouldn’t lose its contents when the power was switched off. Best of all, the company’s fabrication processes would be compatible with standard CMOS lithography, so that chip makers wouldn’t face the kind of retooling expenses that dog (and often doom) most new technologies.
These claims, buttressed by functioning prototype chips, got the interest of venture capital firms and garnered the company lots of breathless prose in the technology press. Yes, including us: this reporter, writing in this magazine, had only good things to say about it (see “10 Tech Companies for the Next 10 Years,” IEEE Spectrum, November 2004).
But that was more than three years ago. And it so happens that two years is the time Nantero has always cited in response to questions about when its product would be ready for the market: in 2002 the chips were supposed to be out by 2004; and in 2004, by 2006. Now Greg Schmergel, cofounder and chief executive officer of Nantero, cites the same figure, telling Spectrum that “with the right resources, commercial products could be out in a one- to two-year time frame.” He says he’s talking with major chip makers in the United States, Europe, and Asia and hopes to announce a major deal in a few more months.
What the Experts Say
“Now we know where the dumb money goes.”
“Nantero is impressive and ingenious in its use of characteristics of carbon nanotubes. But the delay in introducing commercial products—there are none yet—may indicate difficulty in achieving manufacturability.”
Then again, that’s what he was saying two years ago, and two years before that.
Skeptics doubt that such speedy development can be achieved. “Every change you make in the electronics industry, as simple as it may look—changing wiring from aluminum to copper, or adding a hafnium insulator—takes 10 years’ effort by the entire industry,” says Phaedon Avouris, group leader of the Nanoscale Science and Technology Group at IBM Research.
Avouris says the only way forward is to master the science of carbon nanotubes, particularly the problem of achieving the purity that Nantero tries to avoid dealing with by taking its average-of-many-tubes approach. Such averaging throws out almost all of the advantages peculiar to the material, Avouris says, and can be of use only in niche markets. “IBM has never made small trinkets,” he sniffs.
Might not little Nantero conceivably do very well off just such a niche market? Maybe, concedes Avouris, but where is the payoff? “They make assertions that they have products, but go look at their Web site,” he says. “I see they are offering solutions for others, providing technical support for $190 an hour.”
G. Dan Hutcheson, chief executive of VLSI Research, a top semiconductor analysis company in Santa Clara, Calif., is even more pessimistic: “I’ve been an unbeliever from the beginning,” he says. He doubts, specifically, that the NRAM technology can scale well enough to keep up with the ever-shrinking circuit components on standard chips. “The technology itself is very interesting, and it works in the lab, but look how long it took flash to succeed. A company like Nantero can’t take such long bets.”
Right now, Nantero is relying to a large extent on military R&D funds, says an engineering professor who asked for anonymity. The military wants NRAM chips for their ability to shrug off electromagnetic pulses, such as might be directed against a satellite by an enemy or even an errant solar flare. However, that market is far too small and specialized to serve as a launching pad for commercial electronics, the professor says. He also repeated Hutcheson’s contention that Nantero’s nanotube patches can’t scale down sufficiently to keep up with advances in CMOS technology.
“If you look at the picture of the nanotube fabric in their pictures,” he says, “you see it has 200-nanometer holes. Their fabric ribbons can’t be smaller than the holes, obviously.” For comparison, the smallest features of the CMOS transistors now being manufactured in state-of-the-art facilities measure roughly 45 nm. This professor says that three years ago the company had said it had hopes of finding new chemical processes to align the nanotubes better, and so reduce the pore size, but that it hadn’t worked out.
Schmergel rejects the criticism, saying that Nantero has demonstrated a prototype with critical dimensions as small as 22 nm. When it is pointed out that those features were carved with an electron beam, a laborious procedure that could never work in mass fabrication, Schmergel retorts that “even to make the 22-nm prototypes, we needed to solve the porosity problem.” Purely chemical fabrication methods capable of scaling down the NRAM will be possible, he adds.
Meanwhile, the years fly by and rival technologies continue to advance, not least among them plain-vanilla flash memory, which is getting better, cheaper, and smaller. That instant-on computer that Nantero sketched out more than six years ago? You can buy one right now for just $400; it’s called the iPhone. Cheap flash is the motive force behind One Laptop per Child’s XO (often erroneously called the “$100 laptop”), which is being churned out by the tens of millions for the poor countries of the world. It is also what has, in the past year, enabled both Samsung and Dell to introduce powerful instant-on laptop machines.
Nanotube-Based Nonvolatile Memory Chip
Goal: To store piles of data in a small space, permanently.
Why It’s a Loser: Its time has run out.
Where: Woburn, Mass.
Staff: Info not available
Budget: US $31 million in venture capital
Sure, flash won’t keep getting better forever, but what about the other storage technologies—using magnetism and other tricks—that have far more R&D muscle behind them than the NRAM? How can Nantero keep up?
“It is competing with large companies,” notes Dexter Johnson, a nanotech analyst at Cientifica, a consultancy based in London. “Samsung, for instance, has created a $4 billion market for themselves with flash memory. Do you think they are going to idly sit by while some start-up says they are going to make that business obsolete? Not likely; they have their own approach, which they are developing in conjunction with University of Cambridge.”
Nantero’s patented methods may well find application in completely different products. Just this past September, for instance, HP’s inkjet printing people signed an agreement to use Nantero’s nanotube techniques to print up inexpensive RFID tags. A few months earlier, Alpha Szenszor, a start-up also based in Woburn, acquired rights to Nantero’s technology for use in its medical diagnostic sensors. Such licensing agreements could turn out to be lucrative indeed; a miner who hasn’t struck gold can still make a living selling pickaxes to other miners.
But as a mass-market replacement for flash, the NRAM chip looks increasingly like a loser.
An error on the comparison of NRAM to static random access memory has been corrected. They are unlike with respect to volatility.