|
Chip engineers are the quintessential geeks. They thrive on problems that push the limits of physics, working in the shadows while their software counterparts soak up the glory (and the money). For these unappreciated figures, their anonymity in the tech world is a point of pride—only insiders truly appreciate the effort that goes into doubling chip performance every few years. Nearly 20 years ago, as a chip designer, I was part of this unassuming tribe of geeks myself.
But as I write in my latest
Technology Quarterly,
the semiconductor industry has now stepped out of the shadows and into the spotlight.
The rise of artificial intelligence
(AI) has boosted demand for specialist AI chips to such an extent that seven of the ten most valuable firms in the world are in the chipmaking business. Governments are dangling billions in sweeteners to lure semiconductor manufacturing to their shores, and policymakers are well-versed in industry jargon.
Nothing better captures the central role chips play in the economy than the investor frenzy around Nvidia, the world’s most valuable chipmaker. Blackwell, its latest superchip, crams over 208bn transistors (which includes two processors and eight stacks of high-bandwidth memory) into a single package. Those specs are cause for excitement. But as worries about manufacturing delays emerged, Nvidia’s market capitalisation dropped by almost $170bn in early August, more than the value of some major companies. This level of attention is new for the industry.
What is not new, though, is the pressure to keep churning out more powerful processors. For decades chipmakers could rely on the familiar trick of
shrinking the transistor
to squeeze ever more into the same space.
That era is over.
Chipmakers now need a whole
new bag of tricks,
from vertical transistor stacks and lightning-fast optical connections to clever packaging and even elements beyond silicon. Some believe that it might be time to
ditch digital computing
altogether, and instead try to replicate the efficient computing strategies of our brains.
Earlier this summer I visited Antwerp to glimpse the future of chipmaking. Europe’s diamond hub may seem like an odd place to gather such insights. But the city hosts the annual gathering of Interuniversity Microelectronics Centre (better known as IMEC), a Belgian chip-research organisation. IMEC’s neutrality allows researchers, chip rivals and policymakers to converge on neutral ground and hash out an industry roadmap. Despite all the newfound attention from investors and governments, most of the discussions still focused on “power walls”, 3D packaging and lithography. For the geek in me, it was a relief to see that the culture
had not changed all that much.
Elsewhere in The Economist:
Thank you to readers who responded to
last week’s newsletter
about the magic of breast milk. Colin, a former dairy farmer in New Zealand, reflects that farmers “inherently understood” the value of colostrum, the first breast milk a human or animal creates. It was essential in ensuring that new-born calves stayed healthy, he says. Ginette, meanwhile, cautions against viewing breast milk as a profit-making commodity rather than an integral part of the bond between a mother and baby. If you have any thoughts or questions about this week’s newsletter, or the chipmaking industry in general, reach us at
sciencenewsletter@economist.com.
|
