Chip Capacity Is Being Rebuilt Globally, Not in One Place

Texas Instruments does not usually make news for adding a production line. This one is worth a second look. The Dallas company said on 24 October 2024 that its Aizu, Japan, fab had started producing gallium nitride power semiconductors, and that the move, alongside existing GaN output in Dallas, quadruples its internal GaN manufacturing capacity. The announcement is on TI's own newsroom.

That is a capacity story. It is also a supply-chain story, because it shows where chipmakers think the risk sits.

TI says it qualified a 200mm GaN process for mass production in Aizu. Mohammad Yunus, the company's senior vice president of technology and manufacturing, is quoted in the release saying the company wants internal manufacturing to exceed 95% by 2030 while still sourcing from multiple TI locations. GaN is an alternative to silicon for power chips: adapters, motors, server power supplies. TI also said it piloted GaN process development on 300mm wafers earlier that year and that its GaN processes are fully transferable to 300mm, which is the company's stated path to scaling. None of these are consumer numbers. They matter because power chips sit under data centres, cars and air conditioners, and because TI is deliberately pulling that work in-house.

Meanwhile, one road in North Carolina

On 1 October 2024, Tom's Hardware reported that Hurricane Helene had hit Spruce Pine, North Carolina, home to high-purity quartz mines that supply the semiconductor industry. The report said the storm brought over two feet of rain to the town, cutting off road access and leaving residents without power, and that it was difficult to assess damage to the two mining companies, Sibelco North America and The Quartz Corp. The Quartz Corp told The Verge it was too early to say whether its high-purity quartz production facility had been affected, according to the same piece. Sibelco had confirmed a suspension of operations, Tom's Hardware wrote, but the outlet could not reach the company because phone service was down.

Quartz crucibles hold molten silicon while it is grown into ingots. The quartz has to be pure, because impurities migrate into the melt. Spruce Pine is where most of the pure material comes from.

But the widely shared claim that all chip manufacturing depends on Spruce Pine is too strong. Brian Potter, writing on Construction Physics on 8 November 2024, walked through the process and the alternatives: the piece says Spruce Pine is estimated to produce 180,000 to 200,000 tons of high-purity quartz per year, that the two miners there are Covia, formerly Unimin, a subsidiary of Belgian group Sibelco, and the Norwegian Quartz Corp, and that cutting off the supply would not choke off semiconductors completely but would lower yields and raise costs. High-purity quartz costs on the order of $10,000 a ton, Potter writes, a crucible costs several thousand dollars, lasts around 400 hours of operation, and one paper he cites estimated crucibles at 30% of the manufacturing cost of monocrystalline silicon ingots.

That is the useful framing. A single town in the Blue Ridge mountains is not the whole industry. It is one input with poor substitutes, and substitutes are being developed. Potter notes that a new crucible material would be a large deal, not only because it removes the bottleneck but because quartz constrains efficiency in ingot production generally.

India writes a bigger cheque

If the West's chip policy has been about bringing fabs home, India's latest move is about scale and depth at once. TechCrunch reported on 15 July 2026 that New Delhi unveiled a Mobile Phone Manufacturing Scheme worth ₹625 billion, about $6.5 billion, running five years, with incentives of 2.25% to 5% of eligible sales and an extra 1.5% for sourcing key components and sub-assemblies in India. The report also says the government committed a further ₹1.28 trillion, around $13.3 billion, to domestic semiconductor manufacturing, expanding a $10 billion chip incentive programme launched in 2021, with more support for equipment, materials, design and research.

The context matters more than the headline figure. China accounted for 63% of global smartphone production in 2025, against India's 18%, according to Counterpoint Research, as cited by TechCrunch. Apple began assembling iPhones in India in 2017 and about 25% of them are now made there, with suppliers including Foxconn and Tata Group. The smartphone scheme runs through March 2031; the government expects mobile phone production of about ₹39 trillion, roughly $405 billion, over that period and about 60,000 direct jobs. There is also a 3% extra incentive on eligible sales for product design and research aimed at Indian brands, IT minister Ashwini Vaishnaw said at the briefing.

"Apple stands to benefit directly," Navkendar Singh, associate vice president at IDC, told TechCrunch, adding that India's strengthening manufacturing and export credentials could give Apple more confidence to diversify away from China.

Tarun Pathak, research director at Counterpoint Research, told the same outlet that brands are looking to save every cent on component sourcing as memory prices reach record highs, and that local production could pay off over the longer term as a weaker rupee raises import costs. Both comments point at the same thing: assembly is not the prize.

Malaysia's angle

Malaysia has a different pitch. Amir Hamzah Azizan, the country's second finance minister, told CNBC's Karen Tso at the IMF spring meetings in Washington that the semiconductor space is in an upcycle and Malaysia will benefit. In the interview, published 18 April 2024, he said Malaysia does not see China, its largest trading partner, as a rival: "I think the reality of it all is, there is enough growth that will go around." He also said Malaysia would not go head on at the high end where the United States is bringing the parts, and that the focus is on extending the value chain and on economies of scale.

Malaysia holds 13% of the global market for chip packaging, assembly and testing, according to the Malaysian Investment Development Authority, cited in the CNBC piece, which said the figure came from a 18 February report. That is backend work, and the government says it wants to move up. The minister's argument is that diversification by end users is real and that supply chain connectivity is the thing to sell. It is a modest claim compared with India's, and it may be the more honest one.

The automation question

Capacity is not only about buildings. IEEE Spectrum published a history in December 2024 by Jesse Aronstein on Project SWIFT, an IBM effort from 1970 led by Bill Harding to fully automate wafer fabrication and produce integrated circuits in less than a day. The article says SWIFT averaged 5 hours to complete each layer of its process, while the fastest modern fabs take 19 hours per layer and the industry average is 36 hours. Today's chips have more layers and larger wafers, and the comparison is not like for like, but the gap is large enough to be interesting.

Aronstein was there. He writes that he first met Harding in 1962, when IBM was preparing to build the System/360, and that Harding bellowed "What the hell good is that?" at him during a demonstration of handling unpackaged semiconductor dice. Harding had been wounded three times in combat in the Second World War, in Patton's Third Army, and was, in Aronstein's words, rough around the edges for an IBM manager. The project's advances, the article says, are now commonplace in automated fabs, but its turnaround time has never been equalled.

So the industry is adding capacity in Japan, India, Malaysia and the United States, and it is still slower per layer than a line built by a brash middle manager in East Fishkill more than fifty years ago. The capacity race is real. The productivity race may be the older problem.