The AI accelerator bottleneck nobody priced in: one Japanese plant's glass cloth

Every conversation about AI accelerator supply starts with GPUs and foundries. The more interesting constraint sits further upstream, in a material most people have never heard of, made by a company with almost no competition.

According to Tom's Hardware, Nittobo controls roughly 90% of the global supply of specialist glass-fiber cloth known as T-glass. The material sits in the organic core of IC substrates, the interconnect layer between a chip and its printed circuit board. It keeps large, hot chip packages dimensionally stable as packaging gets denser. Tom's Hardware puts current prices up between 20 and 30%, with lead times for downstream materials such as copper-clad laminates stretching from a normal 8 to 10 weeks to beyond 20.

That is a supply chain problem priced into the wrong end of the market.

Why T-glass is hard to replace

Nittobo is tripling capacity at its Fukushima plant in Japan, but Tom's Hardware reports the new supply will not arrive on the market until mid-2027. The process needs specialised electric melting furnaces running between 1,600 and 1,700 degrees Celsius. They melt silica-rich glass, spin it into yarn and weave it into an ultrathin cloth. That is not a line anyone spins up in a quarter.

Bilal Hachemi, an analyst at Yole Group who tracks the IC substrate supply chain, told Tom's Hardware Premium that replacing T-glass is not straightforward because it "has specific dielectric and CTE values that work better for the AI chips, especially for the organic core." He also pointed at the industry's structural fragility: the IC substrate business has historically run on thin margins, so even modest demand surges trigger shortages. "Any increase in demand for build-up materials, ABF material, or T-glass can cause potential shortage, because it's against the basics of this industry," he said.

Bill Ho, an analyst at Yuanta, described the current state of the market in blunter terms. "With T-glass supply even more constrained now, suppliers are no longer providing lead times," he said.

E-glass, the cheaper alternative used in microcontrollers and older mobile processors, is not a drop-in fix. T-glass is preferred for 2.5D and 3D packaging, which is exactly where AI accelerators live.

The demand curve is bending the wrong way

The crunch is being driven by package sizes, not just unit volumes. According to data from Nvidia cited by Tom's Hardware, interposer sizes have grown from 814mm squared for the Hopper architecture to 1,700mm squared for Blackwell, a 109% increase, with the forthcoming Rubin and Feynman generations scaling further still. Bigger interposers and more complex substrates mean more T-glass per accelerator.

Bank of America estimates that Nittobo's electronic materials segment will nearly double sales from 40.9 billion yen, about $266 million, in 2025 to 87.7 billion yen by March 2028, with operating margins approaching 48%. "Demand for T-glass cloth seems likely to grow more than originally expected," Takashi Enomoto, a research analyst at Bank of America, said. He added that attention had focused on thick T-glass for GPU and CPU packages, but that ultra-thin T-glass demand is now likely to rise as leading-edge devices shift away from E-glass.

Margins near 48% explain why Nittobo is investing, and also why nobody has undercut it at scale.

Nvidia goes upstream, and the scramble starts

The most telling detail in Tom's Hardware's reporting is behavioural. Hachemi said Nvidia reaching out directly to an upstream material supplier such as Nittobo is unprecedented in that part of the supply chain. "For the first time, we are seeing Nvidia, the end customer, reaching out to the upstream material suppliers to secure the capacity and make sure they will get it," he said.

The implication is uncomfortable for everyone else building accelerators. Once the largest buyer locks down its allocation, rival chip buyers compete for whatever remains, and the material has no fast substitute.

Nittobo is not standing still. Beyond Fukushima, it is doubling raw yarn capacity at its Taiwan plant and importing yarn back to Japan for cloth manufacturing. It has also struck a collaboration deal with Nanya Plastics to outsource some weaving. By 2027, roughly 20% of Nittobo's glass cloth is expected to be woven by Nanya, according to the company's disclosure.

Partnering with one of your biggest competitors to ease a bottleneck is a fairly clear signal of how tight the market has become.

None of this makes AI accelerators unavailable. It does mean that forecasts built on foundry capacity and HBM supply are missing a constraint that takes years, not quarters, to relieve. The capacity is coming. It is just arriving in mid-2027, against demand that is already compounding.