Asia dominates the global semiconductor supply. Five companies, spanning Taiwan, South Korea, and Japan, sit at the critical juncture of the AI buildout, controlling everything from fabrication to the equipment that makes chips possible. 

‍

Quick facts

• TSMC delivered $90 billion in revenue in 2024, with a 59% gross margin and shares up 55% in 2025.
• Advantest shares doubled (+102%) in 2025 as AI-driven chip testing demand surged.
• SK Hynix is Nvidia's primary HBM supplier, positioning it at the centre of the AI accelerator boom.

‍

1. Taiwan Semiconductor Manufacturing Co. (TSM)

TSMC is the world's largest contract chip manufacturer, producing advanced semiconductors for Apple, Nvidia, AMD, and Qualcomm. As a pure-play foundry, it leads in 5-nanometer (5nm) and 3- nanometer (3nm) chip production, with smaller nodes in development.

The company posted $90 billion in revenue for 2024 with a 59% gross margin and 36% return on equity. 

Shares delivered a total return of 55% in 2025, with analysts forecasting a further ~30% revenue increase in 2026, underpinned by its $100 billion US expansion programme.

The key risk for the company is its geopolitical exposure, with Taiwan Strait tensions remaining the sector's most-watched tail risk.

What to watch

• US expansion progress: Any delays, cost blowouts, or political friction concerning TSMC's $100 billion Arizona investment could weigh on sentiment.
• Customer order visibility: Watch for any guidance updates from Apple, Nvidia, or AMD on chip orders, as TSMC's revenue is highly concentrated among a handful of clients.
• Geopolitical developments: Any escalation of Taiwan Strait tensions could trigger sharp moves regardless of fundamentals.
• Next-node ramp: Progress on 2nm production and yield rates will be a key signal for TSMC's ability to maintain its technology lead.

‍

2. Samsung Electronics (KR:005930)

Samsung is one of the few companies globally that both designs and fabricates chips at scale. It competes across DRAM, NAND flash, and logic chip segments, and remains a core supplier to global tech giants.

Samsung's wide scope is a strength, but also a complexity. Its memory division faces margin pressure from inventory cycles, while its foundry business continues to lag TSMC in leading-edge yields. 

The AI-driven memory boom may provide a tailwind, though execution in HBM production has been slower than local rival SK Hynix.

What to watch

• HBM qualification progress: Samsung has been working to qualify its HBM3E chips with Nvidia. Any confirmation of a major supply win could be a meaningful catalyst.
• Memory pricing trends: DRAM and NAND spot prices could be an indicator of Samsung's margin trajectory.
• Foundry yield improvements: Samsung's logic foundry business has struggled with yields at advanced nodes; any credible progress here could re-rate the division.
• Management guidance: Following a period of earnings volatility, clarity on capex plans and divisional targets at upcoming results will be closely watched.

‍

3. Advantest (ATEYY)

Tokyo-based Advantest makes testing equipment used to verify chips meet performance and quality standards. 

It supplies to Samsung, Intel, Nvidia, Qualcomm, and Texas Instruments, allowing it to benefit from chip industry growth broadly, regardless of which foundry wins market share.

Advantest shares doubled in 2025 (+102%), and it raised its sales forecast by 21.8% and earnings forecast by 70.6% for the year ending March 2026.

What to watch

• Order backlog updates: Any contraction in Advantest's backlog could be an early warning sign after the strong 2025 run.
• AI chip testing demand: As chips grow more complex, testing time per chip increases. Monitor whether AI accelerator volumes from TSMC and Samsung start to drive outsized testing demand.
• FY2026 guidance: The next forecast update will be critical in confirming whether 2025's upgrade cycle has further to run.

‍

4. Tokyo Electron (T:8035)

Tokyo Electron is among the world's largest suppliers of semiconductor production equipment, specialising in deposition, etching, and cleaning tools. 

Every major chipmaker, including TSMC, Samsung, and SK Hynix, depends on TEL's systems to scale production.

As chipmakers invest billions to expand capacity, TEL's order book grows. The risk lies in potential US export restrictions on advanced equipment sales to China, which remains one of the primary revenue segments for the company.

What to watch

• US export control policy: China accounts for a significant portion of TEL's revenue. Any tightening of equipment export rules is the most immediate risk to watch.
• Chipmaker capex announcements: TSMC, Samsung, and SK Hynix's capital expenditure plans for 2026 directly translate into equipment orders. Any cuts could flow through to TEL's order book.
• New tool adoption cycles: Monitor whether TEL's next-generation deposition and etch tools are being adopted at leading-edge fabs.

‍

5. SK Hynix (KR:000660)

SK Hynix is the world's second-largest memory chip maker and has emerged as arguably the clearest AI-era beneficiary in the memory space. 

It is Nvidia's primary supplier of High Bandwidth Memory (HBM) chips, the specialised memory used in AI accelerators like the H100 and B200.

HBM demand has driven a dramatic re-rating of SK Hynix's revenue profile and market standing. With AI infrastructure spending showing little sign of slowing heading into 2026, the company's HBM franchise could remain a key differentiator. 

However, capacity constraints and the risk of Samsung and Micron closing the HBM gap are the primary concerns to watch.

What to watch

• Nvidia supply relationship: Any shift in Nvidia's supplier mix toward Samsung or Micron could be a key risk event.
• HBM4 development: The race to next-generation HBM is already underway. Watch for updates on SK Hynix's HBM4 readiness and whether it can maintain its lead.
• Conventional memory pricing: SK Hynix still derives meaningful revenue from standard DRAM and NAND. Spot price trends could be a gauge of the broader memory cycle.

‍

Bottom line

TSMC, SK Hynix, Samsung, Advantest, and Tokyo Electron collectively control the chokepoints of the AI buildout. 

The expected increase in AI infrastructure may support demand, but investors should weigh the risks carefully. 

Geopolitical exposure, US export restrictions, and the pace of HBM competition could all move the needle.

‍

Ready to trade beyond the majors?

‍Open an account · Log in

While all eyes are on the US AI narrative dominated by Nvidia, Microsoft, and Google, Asia has quietly been moving on AI and is home to some of the world’s most aggressive AI bets.

‍

Quick facts

• SoftBank has committed $41 billion to OpenAI, securing approximately an 11% ownership stake.
• Alibaba plans to invest more than $50 billion in AI infrastructure over the coming years.
• Baidu's Core AI-powered business revenue grew 48% year over year in Q4, with ~70% of search results now AI-generated.

‍

1. SoftBank Group (TYO: 9984)

SoftBank is the most AI-committed company in Asia by capital deployed and ambition. CEO Masayoshi Son has declared the company in "total offence mode," having completed a $41 billion investment into OpenAI for approximately an 11% ownership stake. 

Son has also launched a $100 billion initiative aimed at building a vertically integrated AI semiconductor champion (Project Izanagi), repositioning SoftBank as an "AI-era industrial holding company." 

SoftBank's fortunes are now deeply tied to the success of OpenAI and Son's ability to execute his semiconductor plan that puts it in direct competition with established players.

What to monitor

• OpenAI's trajectory: Any shift in OpenAI's competitive position, valuation, or path to profitability has direct implications for SoftBank's balance sheet.
• Project Izanagi progress: Watch for partner announcements, funding milestones, and whether Son can attract the engineering and manufacturing talent needed.
• Arm Holdings performance: SoftBank also has a listed stake in Arm. Arm's data centre and AI chip licensing momentum is worth tracking.
• Debt levels and Vision Fund exposure: SoftBank carries significant leverage. Rising interest rates or a correction in AI valuations could pressure the group's net asset value.

‍

2. Alibaba Group (BABA)

Alibaba has committed more than US$50 billion to AI infrastructure, making it one of the largest AI capex programmes in the world. 

Its Qwen family of large language models underpins a rebuilt AI-focused cloud platform, and the company has partnered with Nvidia on physical AI projects. 

Alibaba Cloud is also the leading cloud provider in China. The key commercial question is whether Alibaba's can convert this cloud leadership into durable revenue growth.

However, it will have to navigate ongoing regulatory scrutiny in China and competition from local rivals like Huawei and ByteDance.

What to monitor

• Cloud AI revenue growth: The clearest signal of whether the $50 billion investment is translating into commercial traction.
• Qwen model adoption: Enterprise and developer uptake of the Qwen model family could be an indicator of Alibaba's AI platform stickiness.
• Regulatory environment: Beijing's approach to large tech platforms and any renewed regulatory action could disrupt execution and sentiment.
• US-China tech tensions: Nvidia partnership activity and access to advanced AI chips could be affected by further export controls.

‍

3. Baidu (BIDU)

Baidu has made the most visible AI transformation of any company on this list. It has released a 2.4 trillion parameter omni-modal model (ERNIE 5.0) with approximately 70% of its search results now delivered as AI-generated rich media. 

Beyond search, its Apollo Go robotaxi service is now partnering with Uber to expand into Dubai and the UK.

Its Core AI-powered business generated RMB 11.3 billion in Q4 revenue, up 48% YoY. The question now is whether that momentum is sustainable and whether the robotaxi business can scale economically.

What to monitor

• ERNIE monetisation: Watch for updates on enterprise API revenue and advertising yield improvements driven by AI-generated search.
• Apollo Go expansion: Rider volume growth and cost per ride will indicate whether unit economics are improving.
• Search market share: Competition from ByteDance and emerging AI-native search alternatives in China is a potential structural risk.

‍

4. Tencent Holdings (HK: 0700)

Tencent's AI play is to allocate its GPU capacity to itself. This allows it to convert AI directly into efficiency gains across its ecosystem. 

With WeChat's 1.4 billion users providing an unmatched data engine, Tencent is embedding AI across gaming, payments, cloud, and search in a way that is difficult to replicate. 

This approach also offers greater resilience against AI chip export restrictions, since the compute stays internal.

The AI upside here is arguably underappreciated because it is embedded rather than a separate segment, which could also mean the market may find it harder to isolate and value that contribution.

What to monitor

• Advertising revenue trends: The most measurable near-term AI benefit is from ad targeting improvements translating into sustained advertising revenue growth.
• WeChat ecosystem AI integration: Watch for new AI-native features within WeChat, including search, mini-programs, and payments, as signals of platform deepening.
• Regulatory and geopolitical risk: Tencent operates under ongoing scrutiny from Chinese regulators and faces restrictions in some Western markets.

‍

5. Kakao (KRX: 035720)

Kakao is South Korea's dominant AI and internet platform, operating KakaoTalk, which is used by approximately 95% of South Koreans.

It is one of the most aggressively AI-focused non-Chinese tech companies in Asia, investing heavily in LLM development and AI-native services. 

The domestic dominance of KakaoTalk provides a captive distribution platform for AI products in a way few companies outside China can match. The key question is whether Kakao can monetise that distribution advantage before global competitors close the gap.

What to monitor

• KakaoAI product rollouts: New AI-native features within KakaoTalk and Kakao's broader service suite are the most direct signal of commercial AI progress.
• Cloud division growth: Kakao's cloud business is the infrastructure layer for its AI ambitions. Revenue growth and enterprise customer additions are key metrics.
• LLM competitive positioning: Monitor how Kakao's models benchmark against global and regional peers, and whether Korean enterprise customers are adopting them at scale.
• Corporate governance: Kakao has faced governance-related scrutiny in recent years; any developments here could affect sentiment independently of AI progress.

‍

Bottom line

Asia's AI landscape is far more complicated than a simple "follow the AI spend" narrative suggests. 

China's top companies are innovating rapidly but operate under regulatory and geopolitical constraints. Japan's SoftBank is making the biggest single bet, but at a level of concentration risk that demands scrutiny. And South Korea's Kakao offers a differentiated, lower-geopolitical-risk angle.

The AI push in Asia is real. But the range of outcomes across these five names is wide, making it pivotal to understand each company's specific exposure and risk profile, not just its AI narrative.

After three consecutive years in which mega-cap AI-linked names carried the Nasdaq, the mix of winners may be starting to change.

2026 is the "show me the money" year. Any hint of doubt about whether tech companies were correct to spend nearly US$700 billion on AI last year could have a major impact on market sentiment.

‍

Quick facts

• Global AI capex is projected to exceed US$600 billion in 2026.
• The total addressable market (TAM) for AI data centre systems is estimated to exceed US$1.2 trillion by 2030.
• Nvidia, Microsoft and TSMC are all trading below analyst fair value estimates, despite surging revenues.
• Broadcom's AI chip division is targeting US$100 billion in AI revenue by 2027.

‍

What is powering the AI trade?

Multiple macro forces are likely to underpin the AI investment theme through 2026. The direction of US interest rates, the scale of AI infrastructure spending and the geopolitical backdrop are all likely to matter.

Rates and valuations

The Federal Reserve delivered 75 basis points (bps) of rate cuts in 2025, and markets expect another 50 bps in 2026. Lower rates can reduce the discount applied to future tech earnings and typically support growth stocks, including AI-linked names.

Infrastructure spending and earnings expectations

On the spending side, Nvidia CEO Jensen Huang has said data centre operators could spend up to US$4 trillion annually by 2030, and AI capital spending is projected to reach US$571 billion in 2026 alone.

However, markets appear to have already priced in much of this optimism. Analysts are projecting 14% to 16% annual earnings per share (EPS) growth in 2026. That would require S&P 500 stocks outside the Magnificent 7 to roughly double the pace of earnings growth recorded in 2025.

Geopolitics and export controls

Geopolitics could also shape the outlook. US-China export controls on AI chips, along with reduced access to key international buyers, could weigh on data centre growth projections.

‍

Trade the US earnings season

‍

Top AI-linked stocks

Nvidia (NVDA) 

Nvidia remains the clearest expression of the AI trade. It holds a wide economic moat thanks to its market leadership in GPUs, hardware, software, and networking tools. 

Goldman Sachs and Morgan Stanley both carry price targets near $250 on NVDA, with Goldman's call based on a 2027 revenue forecast of over $380 billion. Bank of America sits in the $275 camp, effectively pricing in more AI upside on 2027 earnings.

At 21.6 times forward earnings, Nvidia is now trading below the broader S&P 500's multiple. Key risks include the overhang from US–China export restrictions and any softening in data centre capex guidance from major cloud providers.

Microsoft (MSFT)

Microsoft is down around 25% from its all-time high. During the second quarter of fiscal year 2026, Azure's revenue increased 39% year over year, and the company holds a US$625 billion backlog of contracted usage still to come. 

The gap between the stock's recent performance and its underlying revenue growth has drawn attention from analysts, though elevated valuations across the broader tech sector remain a risk to watch.

Broadcom (AVGO)

While Nvidia makes broad-purpose GPUs, Broadcom is winning business by going bespoke, designing custom AI chips tailored specifically to the needs of individual hyperscalers like Google and Meta.

During Q1 of FY2026, Broadcom's AI semiconductor division grew at a 106% pace to US$8.4 billion, and by the end of 2027 it expects its AI chip revenue to reach more than US$100 billion.

Broadcom trades at a significant premium to the broader market, which could amplify any downside if growth expectations are not met.

TSMC (TSM) 

Almost every major AI chip is manufactured by TSMC. The company holds approximately 70% market share in chip foundry, making it the single most critical piece of infrastructure in the entire AI supply chain. 

TSMC sales are projected to increase by 30% in 2026, with gross margins expected to remain above 60% as new fabrication capacity comes online.

The primary risk is geopolitical: any escalation in Taiwan Strait tensions could weigh heavily on the stock regardless of its underlying fundamentals.

Vertiv (VRT)

Less prominent than the semiconductor giants, Vertiv provides the power management, cooling, and data centre infrastructure that keeps AI hardware running. 

Nvidia, Broadcom, and Vertiv sit at different points in the AI build-out, including compute, custom silicon, networking and physical infrastructure.

Vertiv's revenue is tied to overall AI capex rather than any single chip maker, which gives it a different risk profile to the names above.

Corning (GLW)

Corning's stock rose 84% in 2025 thanks to surging demand from data centres for its fibre optic cables. Its optical communications segment has grown 69% YoY. 

At a Price-to-Earnings (P/E) ratio of roughly 37x, Corning trades at a discount to Nvidia and Broadcom while still carrying direct exposure to AI infrastructure spending. However, its valuation depends heavily on continued capex from the major hyperscalers.

‍

US market drivers for March 2026

‍

AI trades beyond the headline stocks

Energy and utilities

Training large-scale AI models is extraordinarily energy-intensive. A typical 1 gigawatt AI data centre facility requires upwards of US$60 billion in capital expenditure, with roughly half going directly to hardware. Utilities exposed to data centre power demand could also be affected by the AI build-out.

International spillover

South Korea's Kospi surged 76% in 2025 due to AI-linked chipmakers like SK Hynix. Japan's Topix, Germany's DAX, and the UK's FTSE 100 also saw gains of more than 20%. Memory supplier Kioxia was the world's best-performing stock, surging 540%.

Data centre infrastructure

Companies like Emcor, which provides critical electrical, HVAC, and power infrastructure to data centres, reported its contracted backlog surged 31.2% year over year to a record US$13.25 billion. These companies can offer different exposure to the AI capex cycle, but they carry their own execution, backlog, margin and valuation risks.

‍

What could derail the AI trade?

Valuation compression

Broadcom trades at about 50x earnings and AMD at 56x. Any disappointment in forward guidance could trigger a sharp contraction in multiples.

The return on investment test

Companies are investing today on the assumption that highly profitable business applications of AI will emerge over time. If the timing or scale of those returns disappoints, the AI trade could face pullbacks.

Index concentration

The 10 largest stocks in the S&P 500 account for about 40% of the index's total value. A rotation out of mega-cap tech could disproportionately affect broad indices.

Efficiency disruption

China's DeepSeek recently published research suggesting large language models may be developed more efficiently than previously assumed. If AI can be built with less compute, demand for GPUs and data centre hardware could fall short of current forecasts.

‍

Bottom line for traders

The AI trade is maturing but far from over. 2026 is shaping up to be a more nuanced chapter, spreading across the full AI value chain. 

The US earnings season will be closely watched for evidence that the hundreds of billions being poured into AI infrastructure are beginning to generate the anticipated returns.

‍

All data points referenced in this article were verified against primary sources on 18 March 2026.