Susumu Koike is holding court over lunch at the Kyoto, Japan, offices of Matsushita Semiconductor, the company he heads. Matsushita Semiconductor is a subsidiary of $62-billion-a-year Matsushita Electric Industrial, the world’s second-largest consumer electronics manufacturer (behind Sony), whose product lines include Panasonic, Technics, and Quasar. A pugnacious 58-year-old engineer, Koike describes one of Matsushita’s latest achievements this way: “Reconfigurable FeRAM is our Mount Everest.”
Koike’s metaphor refers to the company’s broad goals for chips that use ferroelectric memory, a new technology that could help make a vast array of consumer electronics-from TVs to handheld computers to cell phones-more versatile and cheaper and, in some cases, faster and smaller. It should ultimately mean more multifunctional digital products, for less money.
To understand Koike’s enthusiasm, consider the cell phone. Yesterday, it was just a phone. Today, it has a built-in digital camera. Tomorrow, it might have a Global Positioning Systembased navigator, an MP3 player, an ID card-and who knows what else. The problem is one of squeezing extra features into the same gadget without raising costs or increasing power consumption. The most popular strategy for semiconductor makers is to cram as much circuitry as possible onto the same piece of silicon, an approach known as system-on-chip. To switch from one function to another, the chips need to be programmable or, in industry jargon, “reconfigurable.” Software effectively reprograms the chip’s circuits to do different jobs.
That approach requires lots of memory, however, to store the software, and that’s where Koike believes his company has climbed ahead. Matsushita didn’t invent ferroelectric memory, but it is the first to mass-produce it on a system chip. Ferroelectric memory, like today’s standard flash memory, retains data even when the power shuts down, but it’s far easier to fabricate. It also stores data much faster than flash, with much lower power consumption-a key consideration for battery-driven portable devices.
Matsushita’s new chips debuted commercially in December 2003, inside smart cards used as season tickets by Japan Rail commuters in Osaka. The chips are much smaller than conventional equivalents, which means they are cheaper to manufacture, and they run on only a fraction of the power.
These chips merely demonstrated the technology’s commercial viability; rail cards don’t need to be reconfigured. But from this modest beginning, the company intends to move forward aggressively. One early application where ferroelectric memory could have a large impact, Koike says, is in chips for cell phones that could be used around the world. Today, cellular-service providers in each country have different protocols, which means they require different types of phones. “It would be quicker to download the appropriate protocol into reconfigurable memory” and just use one phone, says Koike. But that’s just one idea; he predicts that the global market for system-on-chip with embedded ferroelectric memory will grow to $5 billion by 2010.
That’s a plausible projection, says Carlos Araujo, a professor of electrical engineering at the University of Colorado at Colorado Springs and chairman of Symetrix, the company that pioneered the type of memory the Japanese firm uses. “This is no longer like, Gee, we’ve got this new technology, and we’re trying to find something to do with it,’” he says. “This is more like, We have products out, and we’re realigning all of Matsushita around this technology, and all our products from cell phones to digital TVs will have it.’”
Of course, Matsushita’s rivals are working on their own next-generation memory technologies; this year, for example, Motorola is rolling out its first commercial magnetic random-access memory, or MRAM, chip, which also retains data when the power shuts down (see “A Chip Worth Remembering,” TR March 2004).
But whereas Motorola’s chip is too expensive for all but high-end applications, Matsushita’s is cheap enough for use in commuter-rail cards. Indeed, with production volumes running at millions of devices per month, Matsushita has proved that ferroelectric memory can be mass-produced. It has scaled the first peak. Whether other rivals also advance on the summit remains to be seen.
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
Data analytics reveal real business value
Sophisticated analytics tools mine insights from data, optimizing operational processes across the enterprise.
The Biggest Questions: What is death?
New neuroscience is challenging our understanding of the dying process—bringing opportunities for the living.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.