Japan’s attempt at a silicon renaissance is absolutely fabulous

• Japan has attracted big silicon names to set up shop in the country
• And some of them are already getting into production with the type of cutting-edge chips everyone is keen to get their hands on
• But will the market end up with a glut?

After 20 years in the silicon doldrums, Japan’s digital turnaround strategy – which aims to attract foreign chip-makers and associated specialists to set up shop in the north-east Asia country – appears to be paying off. Companies such as IBM and Taiwan Semiconductor Manufacturing Company (TSMC), either alone or as part of a joint venture with Japanese companies, are now preparing to start significant advanced chip production from their newly built facilities in Japan. 

But there are challenges, not least the fact that numerous other countries have pretty much the same idea and are keen to attract chip production to their jurisdictions, including the US itself. The competition to bag a fab or three is intense.

Nevertheless, Japan remains keen to regain ground in semiconductor production, a field in which it used to excel when it held up to 50% of the global chip market before losing out to rivals across Asia and in the US.

Well before the Biden administration tightened US high-tech export restrictions to slow technology being leaked to China – and long before Donald Trump v2.0 set Asian nerves a-jangling by raising the spectre of an all-out trade war with tariffs levied on products manufactured outside the US – the Japanese government had been rolling out a welcome mat for foreign silicon specialists.

Encouraged by the US to “friend-shore” semiconductor operations in South-east Asia in order to mitigate potential supply problems, several semiconductor players had already moved some of their operations as a safeguarding measure in case of a falling out with one or the other superpower. Motivated by Japanese government inducements on the one hand and the lure of a comfortable, geographically bounded industrial ecosystem where they could rub shoulders and cross-fertilise personnel with like-minded competitors and partners on the other, some major silicon players have been busy building in Japan.

Glittering prizes

The biggest win in this regard for Japan may be the world’s largest semiconductor production specialist, TSMC. With sabres rattling (rising tensions is the politically correct term) and its extensive current production facilities on the Island of Taiwan being uniquely vulnerable (just a few miles from China), TSMC said it had begun mass production at its first Japanese fab in Kumamoto prefecture, on Japan’s most southerly island of Kyushu.

The plant has been designed to supply a wide range of chips, including the most advanced ever made in the country, according to Nikkei Asia, and to do so at scale.

A second TSMC fab is being planned, with initial work starting on the site later this year and due to start production in 2027. Taken together, the two facilities will cost a whopping $18.8bn and are expected to produce more than 100,000 300-millimetre wafers per month!

The TSMC build is an important part of the Japanese government’s big (and expensive) push to bolster Japan’s silicon renaissance. It plans to put up as much as 1.2tn yen to subsidise TSMC’s Japanese entry, a significant proportion of the 10tn yen it had earmarked for semiconductor and AI development by 2030 – see Japan chips in with $65bn AI-fuelled support package.

In addition to churning out oodles of advanced chips, it’s expected that TSMC will act as an anchor tenant for Japan’s semiconductor push, attracting other companies to the area to provide services, such as chip packaging and testing.

The other big win for Japan is Rapidus Corp, a heavily government-backed manufacturer of advanced logic semiconductors. It has the backing of a clutch of Japanese giants, including Toyota, plus US computing giant, IBM.

This venture, with plants on the northern island of Hokkaido, is gearing up to start producing much sought-after leading-edge 2 nanometer (nm) chips starting this year. It recently made a major step forward in this effort when it announced the delivery and installation of cutting-edge extreme ultraviolet (EUV) lithography equipment from Netherlands-based ASML – the chip-etching kit that the US administrations have been so keen to keep out of the hands of the Chinese. 

Rapidus is currently working on an advanced semiconductor development and manufacturing fab where the new kit is to be integrated in Chitose, Hokkaido.

In addition, Rapidus says it will install additional complementary advanced semiconductor manufacturing equipment, including automated material handling systems in its IIM-1 foundry to optimise 2nm generation ‘gate-all-around’ (GAA) semiconductor manufacturing. In short, this setup is what’s required to produce the chips necessary for generative AI model training – EUV lithography being one of the key technologies for realising 2nm semiconductors. 

There’s much competition to bag those relocating companies

Of course, other countries, including the US itself, have entered the competition to attract chip producers to their shores. One approach is to offer financial assistance: The US with its Chips and Finance Act, along with Japan’s big cash pledge, are the obvious examples, but diplomatic pressure is also being applied to countries and their companies. 

It all begs the question, are we going to end up with an oversupplied global advanced chip capability and subsequent crash in prices and general panic all round, followed by an extended silicon recession? Is this another feast-and-famine scenario the silicon business has historically been famous for?

Despite Japan’s early wins in the battle for silicon prowess, the relocation game is not just one-way traffic. Many Japanese companies are considering expanding their operations in the US for the same mix of reasons that their counterparts are setting up operations in Japan: to ensure that at least some facilities are planted on the right side of potential trade barriers, thus providing a hedge if US trade policy in the Trump era turns really nasty.

According to a recent survey by the Japan External Trade Organization (JETRO), 38.6% of Japanese companies considering overseas expansion have chosen the US, up from 28.1% a year ago. It’s thought that Trump’s tariff plans may have spiked this interest.

In fact, trade imbalances, and the apparent need to impose tariffs to correct them, had already seen many companies switching production away from China to Thailand, India, Vietnam, Mexico and Taiwan for political reasons. This bout of ‘friend-shoring’ has certainly helped diminish China’s deficit with the US, but has boosted the US deficit of the ‘friends’ and some of the friends are now in the Trump spotlight, having greatly increased their imbalances since 2018. For them, the friend-shoring idea no longer seems so secure. 

Japan, which has only racked up a small US deficit increase over that time, could see its US deficit increase if all goes according to plan with multiple big semiconductor players producing volumes. Trump has recently said he intends to levy blanket tariffs with no exceptions on US imports.

Risks posed to AI house of cards

The markets are also uncomfortably aware that the current AI boom has ballooned the valuations of many of the semiconductor specialists responsible for the crucial AI infrastructure, and they are nervously alert to signs of a market crash if and when sentiment turns sour on unwelcome news.

For instance, the performance of the open-source R1 chatbot application, released recently by Chinese AI company DeepSeek, immediately caused a spasm of self doubt amongst the global generative AI community, since the bot appeared to be capable of inferencing results using a fraction of the computer power apparently required by US-based chatbot rivals.

If usable ‘consumer’ GenAI could be offered for peanuts, there would surely be less demand for the advanced processor power required to train the models and infer answers to questions, so tech investors immediately took fright, causing a ripple of stock market losses.

Worst hit was advanced chip doyen Nvidia, which saw its share price crash by 17% losing nearly $600bn in market capitalisation, the biggest single-day loss in US history, reported Nikkei Asia. 

For respected tech sector analyst Richard Windsor, the big question, as posed in his Radio Free Mobile blog, was: If Deepseek had indeed engineered a major AI breakthrough, why did the Chinese government approve the release of an open-source version that could be used by rival (foreign) companies in the ongoing struggle for technology domination between China and the US?

Windsor suspects that the AI wizardry used to build the Deepseek programs probably “lies, not in the model, but in the methods that were used to create and run it so cheaply”. No detailed account of the methods deployed has so far been offered, but whatever the true motivations of the various parties in the Deepseek episode, the advanced tech share price movement illustrates the fragility of the sentiment driving both investment in digital infrastructure and the support of sky-high multiples for the technologies deemed the likely winners. It only takes a relatively minor technology breakthrough or vague political pronouncement on tariffs, say, to register multiple points on the AI richter scale.

– Ian Scales, Contributing Editor, TelecomTV