[Big read] One Taiwan, two speeds: The winners and losers of the AI boom

(Edited and refined by Grace Chong, with the assistance of AI translation.)

Wang Chun-hsiang (pseudonym), a semiconductor engineer with more than 15 years of experience, spends much of his free time studying Taiwan’s stock market. A seasoned investor who pays for professional market analysis software, he should be celebrating as the market repeatedly scales new highs. Instead, he has grown increasingly uneasy.

Even as Taiwan’s benchmark index continues to climb — with some analysts predicting it could eventually reach 50,000 points — Wang has noticed major institutional investors, the market’s so-called “smart money”, quietly trimming their holdings in Taiwan Semiconductor Manufacturing Company (TSMC), its most heavily weighted stock.

“It just doesn’t add up,” he said.

The day before speaking to Lianhe Zaobao (LHZB), outstanding margin loans on Taiwan’s main stock exchange exceeded NT$600 billion (US$19 billion) for the first time, setting another record. To Wang, it is yet another sign that the market is overheating.

“If the market crashes, will ordinary investors be able to withstand it?” he asked.

The boom in the capital markets has been driven largely by excitement over artificial intelligence (AI). As of 1 July, the benchmark weighted index had climbed 60% since the beginning of the year. Since the launch of ChatGPT at the end of 2022, it has risen by more than 200%.

Over the same periods, TSMC — often described as a company capable of carrying Taiwan’s stock market on its own — has gained 58% and 400% respectively.

To Wang, today’s AI boom marks the third major growth cycle for Taiwan’s semiconductor industry in little more than a decade.

The first came with the smartphone revolution, when the launch of Apple’s iPhone fuelled a global surge in demand for chips and accelerated the expansion of Taiwan’s semiconductor sector.

The second arrived during the Covid-19 pandemic, when the stay-at-home economy drove soaring demand for laptops, tablets and other consumer electronics.

The third is unfolding today. The rapid adoption of AI has sharply increased demand for advanced manufacturing processes, cutting-edge packaging technologies and AI chips.

Each successive wave has further cemented Taiwan’s position as a global technology powerhouse. Along the way, TSMC’s share price has climbed from below NT$100 when Wang first entered the industry to more than NT$2,000 today.

AI propels Taiwan into the trillion-dollar club

The three industry booms that Wang has witnessed mirror Taiwan’s broader economic transformation over the past two decades.

Driven by the AI supercycle, DBS senior economist Ma Tieying forecasts Taiwan’s nominal gross domestic product to exceed US$1 trillion this year, placing it alongside mainland China, Japan, India, South Korea and Indonesia as one of Asia’s trillion-dollar economies.

She also forecasts real economic growth of close to 10% this year. Earlier, Taiwan’s Directorate-General of Budget, Accounting and Statistics and the central bank raised their full-year growth forecasts to 9.64% and 9.45% respectively.

If one person has come to symbolise the AI boom, it is Nvidia founder Jensen Huang.

Every visit he makes to Taiwan attracts intense media attention. The technology banquets he hosts — dubbed the “trillion-dollar dinner” by Taiwanese media — have become an important bellwether for the AI industry. They often bring together the leaders of Taiwan’s largest companies and the businesses that have benefited most from the AI boom.

While TSMC remains the best-known name, the dinner also highlights the breadth of Nvidia’s Taiwanese supply chain.

The companies represented produce everything from circuit boards, power supplies and cooling systems to mechanical components and fully assembled AI servers.

Why Taiwan’s supply chain is so hard to replicate

Chris Wei, an industry consultant at Taiwan’s Market Intelligence & Consulting Institute (MIC), said ChatGPT marked the beginning of an unprecedented surge in demand for computing power, as generative AI requires vast numbers of graphics processing units (GPUs).

Producing those GPUs depends on cutting-edge semiconductor manufacturing and advanced packaging technologies, making TSMC one of the industry’s indispensable suppliers.

But AI systems rely on much more than chips alone.

Nvidia must also verify that the chips can operate reliably and stably once they are integrated into fully assembled systems. That requires close coordination not only with upstream suppliers such as TSMC, but also with downstream manufacturers such as suppliers of power and cooling systems.

Ansforce CEO Jefferey Chiu, who is known for explaining AI and semiconductor technologies in accessible terms, said Taiwan’s strength extends well beyond TSMC itself.

“Taiwan isn’t winning this race because of TSMC alone,” he told LHZB. “Its real advantage is that it controls more than 90% of the world’s advanced AI chip manufacturing capacity while also producing around 90% of the world’s AI servers.”

From chip design and advanced manufacturing to packaging, server assembly and final testing, the companies responsible for each stage of production are often located just a few hours’ drive from one another.

According to Chiu, this dense industrial cluster gives Taiwan a level of flexibility and efficiency that is “unmatched anywhere else in the world”. It is also one of the main reasons why global technology giants including Nvidia, AMD, Microsoft and Google continue to expand their investments on the island.

Factories can be built elsewhere, he said. What is much harder to replicate is the ecosystem that Taiwan has spent more than four decades developing — a tightly interconnected network of thousands of specialised small and medium-sized enterprises covering every stage of the semiconductor supply chain.

Booming markets, mounting risks

Chiu believes the AI industry is still in the early stages of what he describes as a period of “rational validation”.

Cloud service providers continue to increase capital spending, while the imbalance between the supply and demand for AI computing power persists. At the same time, AI adoption across the wider economy remains relatively limited, suggesting that the industry is still in the early phase of a long-term growth cycle.

This also means that while the AI boom is far from over, market valuations have, for now, run ahead of underlying fundamentals. To some extent, this reflects the abundant global liquidity of recent years, which has continued to channel capital into equity markets and push stock indices to successive record highs.

As AI becomes the principal engine of economic growth, Taiwan’s economy is increasingly moving at two distinct speeds.

The semiconductor industry and AI supply chain are booming as global investment pours into AI infrastructure. Meanwhile, traditional manufacturers and domestically oriented businesses have seen far fewer benefits and continue to grow much more slowly.

One economy, two realities

One Taiwan, two speeds. Beneath the economy’s continued rise, the gap in how different industries and social groups experience that growth is quietly widening.

At a forum on AI trends, Richard Liao, the fourth-generation successor of century-old glassmaker Hwa-Hsia Glass, described how the company has begun introducing AI into its factories to help ease labour shortages in traditional manufacturing.

With a touch of humour, he joked: “If you don’t have AI, all you’ll have is BI” — a play on words, as “BI” sounds like the Taiwanese word for sorrow (悲哀).

Behind the joke, however, lay a much harsher reality.

As investment, talent and technology continue to flow into the AI sector, many traditional manufacturers have found themselves struggling to share in the gains. Liao believes Taiwan is increasingly becoming “a two-speed economy”.

What caught Hwa-Hsia Glass even more off guard was the shift in the global geopolitical landscape.

After US President Donald Trump unveiled sweeping reciprocal tariffs in April last year, the company’s shipments plunged by 44%. Annual revenue fell by almost NT$600 million, eight production lines were shut down, and the company briefly faced a cash flow crunch.

Although orders have since recovered, Liao told LHZB that weak global consumer demand continues to weigh on the business.

The divide he describes is also reflected in Taiwan’s broader economic data.

The Economist recently observed that, excluding semiconductors and AI servers, Taiwan’s exports have fallen by about 40% since 2022. Yet while the electronics industry employs only around one in ten workers, average wages in the sector are roughly 70% higher than the overall average.

DBS’s Ma said Taiwan’s AI-led growth has also brought new challenges, including a more pronounced K-shaped growth and widening wealth inequality.

“Manufacturing is both capital- and technology-intensive,” she said. “The AI-driven manufacturing boom has not translated into significant job creation.”

Instead, most new jobs continue to come from the services sector.

According to figures released by Taiwan’s Directorate-General of Budget, Accounting and Statistics in mid-June, education, accommodation and food services, and support services were the island’s three lowest-paying industries.

When AI wealth leaves others behind

Hsiao-wei (pseudonym) recently completed her postgraduate studies and has just sat the examination required to obtain her teaching licence. While waiting for the results, she is preparing for the joint recruitment exam for public kindergarten teachers.

The attraction is obvious.

Teachers in public kindergartens can earn up to twice as much as those in private institutions. For many early childhood educators, securing a public-sector position is considered the pinnacle of the profession. Once appointed, they effectively enjoy a secure job for life unless they choose to resign.

Reflecting on Taiwan’s AI boom, Hsiao-wei could not help but laugh. “It really feels like two different worlds,” she said. “These days, even kindergarten teachers end up talking about stocks when we chat privately.”

Like many outside the technology sector, she watches friends in the technology sector enjoy rapidly rising salaries with mixed emotions. “It’s hard not to feel a little envious,” she admitted.

At the same time, she worries that many people have become overly optimistic, borrowing heavily to invest in a market that seems only to go up.

“What if they lose all the money they’ve worked so hard to save?” she asked.

Salary is not her only concern. Hsiao-wei believes early childhood education has long been undervalued in Taiwan. “People think kindergarten teachers just look after children,” she said. “But we’re professionally trained. We study child development and curriculum design, and we work closely with parents. It’s far more demanding than most people realise.”

If a better opportunity arose overseas — for example, in Singapore — she said she would seriously consider relocating.

Ma believes Taiwan’s relatively small economy and highly concentrated industrial structure mean that rapid growth in a handful of technology sectors is enough to lift overall economic performance significantly. As a result, the island’s K-shaped growth has become more pronounced than in larger economies such as the US, mainland China and South Korea.

In the near term, she expects that divergence to persist.

Over time, however, the gap could narrow if investment in AI begins to stabilise and AI applications spread more widely across the broader economy.

Until the benefits of AI become more widely shared, Taiwan’s two economies — the booming AI sector and the rest of the economy still waiting to catch up — are likely to continue coexisting.

New rivals, new risks 

Taiwan’s leadership in AI rests on a supply chain ecosystem that has been built over more than four decades. Yet industry experts warn that Taiwan’s AI industry continues to face two major challenges.

One, intensifying competition from mainland China, and two, a temporary production gap as Taiwanese manufacturers establish operations in the US, which could allow US-based companies to seize orders in the meantime.

Speaking at Nvidia’s GTC (GPU Technology Conference) Taipei conference in June, Jensen Huang said the AI industry is moving beyond generative AI towards agentic AI, with physical AI representing the next major phase of development.

In practical terms, that means AI will increasingly move beyond cloud-based models into end-user devices such as robots, autonomous vehicles and personal computers.

For Taiwan, that opens up another wave of opportunities across the AI supply chain — but it also raises the competitive stakes.

Ansforce’s Chiu said that while US export controls continue to restrict mainland China’s access to the most advanced semiconductor manufacturing technologies, its AI ecosystem is developing at remarkable speed.

Backed by a vast domestic market, abundant real-world application scenarios and the rapid progress of open-source models such as DeepSeek and Alibaba’s Qwen, mainland China has made striking advances in commercialising AI and improving AI models.

Chiu believes the biggest challenge that mainland China poses to Taiwan lies in mature-node manufacturing, general-purpose chips, and the development of the physical AI ecosystem. 

Mainland China continues to expand capacity for mature-node chips, increasing the risk of oversupply and price competition in communications, automotive and consumer electronics.

At the same time, its access to extensive real-world data from physical AI applications, including autonomous vehicles, smart factories and humanoid robots. This could enable it to compete directly with Taiwan’s hardware manufacturers in setting industry standards and developing application ecosystems.

Even so, Chiu believes Taiwan retains one important advantage.

As technological rivalry between the US and China intensifies, the reorganisation of global supply chains has created what he describes as a “trust dividend”.

The technological decoupling between the US and China, together with the US CHIPS and Science Act, has encouraged major American cloud service providers — including Microsoft, Amazon and Meta — to entrust production of their in-house chips and advanced AI infrastructure to trusted manufacturing partners in order to reduce geopolitical risks, further reinforcing the central role of Taiwan’s supply chain.

The geopolitical cost of staying ahead

Yet the same geopolitical tensions that have benefited Taiwan also pose one of its greatest risks.

“Because cross-strait tensions remain so sensitive, international customers increasingly want Taiwanese manufacturers to build backup production capacity outside Taiwan as part of their long-term strategies,” Chiu said. “That inevitably stretches their resources.”

MIC’s Wei said the Trump administration’s efforts to bring manufacturing back to the US have compelled Taiwanese companies to establish production facilities there as well.

But before those new factories become fully operational, American manufacturers could seize the opportunity to win business that might otherwise have gone to Taiwan.

AI server assembly offers one example.

Major US cloud service providers are developing their own AI chips to reduce their dependence on Nvidia. As shipment volumes continue to grow, they may choose to shift some assembly orders to domestic manufacturers instead.

For Taiwan, the AI boom continues to provide a competitive edge. However, its ability to maintain that advantage amid mainland China’s rapid progress, the restructuring of global supply chains, and mounting geopolitical risks will determine whether the “AI island” can remain at the forefront.

This article was first published in Lianhe Zaobao as “AI​引发​​​​​双速经济​两​台湾 新科技狂飙旧产业慢行”.

Related: [Big read] In Taiwan, birth rate soars in Hsinchu’s tech zone | [Big read] Too ‘cool’ to resist: China’s soft power captivates Taiwan’s youth