TSMC — the Taiwan Semiconductor Manufacturing Company and the world’s largest chipmaker — has declined to rule out future price increases, with CFO Wendell Huang acknowledging at a shareholders’ meeting that inflation “did cause our costs to increase” and that the company “reflects its value.” Chairman and CEO CC Wei separately told shareholders he would “like” to raise prices, noting that competitors already have. TSMC has committed $165 billion to manufacturing expansion in Arizona, and is building facilities in Germany and Japan. But Huang pushed back against the narrative that this expansion is government-driven, insisting customers — not states — are pulling the investment. More significantly, Huang confirmed that the most advanced chip production will remain in Taiwan for the foreseeable future, and that replicating the full manufacturing ecosystem in the United States would take “five or 10 years, or even longer.”
The received wisdom
The progressive-technocratic consensus on semiconductor policy is broadly as follows: the concentration of advanced chip production in Taiwan represents an unacceptable strategic vulnerability for the West, the CHIPS Act and its allied equivalents in Europe and Japan represent a necessary, if late, correction to that vulnerability, and the inevitable disruption costs — including higher chip prices — are a reasonable price for strategic autonomy. On this view, TSMC’s price signal is not alarming but expected: domestic production is more expensive, and the additional cost is essentially an insurance premium against the scenario in which a conflict over Taiwan cuts off supply entirely. Proponents of this view point to the post-pandemic chip shortage as evidence that the market alone cannot manage strategic supply chain risk.
There is genuine merit here. Taiwan produces the majority of the world’s most advanced semiconductors, and Chinese President Xi Jinping has warned that mishandling the Taiwan question could put US-China relations in an “extremely dangerous situation”. The geopolitical logic for diversification is real and well-established.
A different read
The problem is not with the goal — diversification — but with the claim that the current industrial policy programme is achieving it on any useful timeline, and at manageable cost.
Huang’s timeline is the first problem. “Five or ten years, or even longer” to replicate the full ecosystem is not a reassuring answer for policymakers who have been arguing that the CHIPS Act represents an urgent security fix. It is closer to an honest acknowledgment that the manufacturing ecosystem TSMC operates in Taiwan — the network of suppliers, trained engineers, specialist chemical producers, and institutional knowledge accumulated over four decades — cannot be airlifted to Arizona by legislative fiat. The factories can be built. The ecosystem cannot be decreed.
This mirrors the pattern of Western industrial policy more broadly. The political impulse is to identify a strategic dependency, announce a massive subsidy programme, and declare the problem solved. The economic reality is that comparative advantage is accumulated over time and is not easily reversed. When Germany tried to build a domestic chip industry in the 1980s and 1990s through Siemens’ Infineon venture, the results were decades of subsidised struggle against Asian competitors who had already run down the learning curve. The CHIPS Act may ultimately prove more successful — the scale is larger, the technology more critical, and the strategic context more pressing — but the timeline Huang describes suggests policymakers and markets should adjust their expectations accordingly.
The price signal is the second problem. Any TSMC price increase flows directly into the cost of AI infrastructure — the data centres, training runs, and inference hardware that underpin the AI investment cycle. TSMC’s own shares have surged over the past year on AI chip demand, but tech shares in Asia have tanked recently following a US sell-off amid valuation concerns. If TSMC raises prices meaningfully, that adds another layer of cost pressure to AI infrastructure that is already being scrutinised for return on investment. The CFO’s insistence that “the AI boom is not a bubble” and that hyperscalers “are financially very strong” is reassuring but circular — it is essentially arguing that because the customers are rich, the demand is real.
The third problem is the assumption embedded in the diversification narrative: that TSMC in Arizona is strategically equivalent to TSMC in Hsinchu. It is not. The most advanced nodes will remain in Taiwan. What the Arizona fabs produce are chips that are one or two generations behind the frontier. In a conflict scenario, the vulnerability the CHIPS Act was ostensibly designed to address — loss of access to the most advanced chips — remains essentially unchanged.
None of this is an argument against the CHIPS Act or against allied diversification efforts. They are prudent hedges in a world of genuine geopolitical risk. It is an argument against the rhetorical inflation that has accompanied them — the suggestion that industrial policy has solved the semiconductor dependency problem when it has, at most, begun a slow and expensive partial mitigation.
What to watch
Watch whether TSMC formalises a price increase and at what magnitude — even a five percent increase would ripple significantly through the AI infrastructure supply chain. Watch for the Arizona fabs’ first commercial production runs and their actual yield rates, which will be the first real test of whether TSMC’s ecosystem can be transplanted. Watch also for any Chinese response to the simultaneous US designation of Chinese military-linked firms: a tit-for-tat that restricts rare earth or chemical exports could meaningfully disrupt TSMC’s Taiwan operations regardless of where its new fabs are located.
— J