Taiwan and South Korea’s AI Stocks

An artificial intelligence service may appear on a phone or computer screen, but behind it sits a network of processors, memory and other equipment. Building that infrastructure creates business for companies well beyond the software industry.

For investors looking at Asia, Taiwan and South Korea offer different ways to participate. The challenge is understanding what each company actually earns from AI, how much it must spend to support that business and what expectations are already reflected in its share price.

A growing industry can create an attractive opportunity. The investment still needs to make sense at the price paid.

Why this week’s announcements need careful reading

Samsung Electronics released its third-quarter earnings guidance on 8 October 2026. As of this article’s publication, TSMC has scheduled its third-quarter earnings conference for 15 October.

These announcements are at different stages. Samsung’s guidance provides preliminary estimates for the quarter. TSMC’s scheduled conference is a future event, so investors should not describe its third-quarter earnings as already reported.

That distinction matters because an estimate, a completed quarter’s results and management’s outlook for the next quarter answer different questions.

A company can report strong recent trading while becoming more cautious about future orders. It can also raise its sales outlook while facing higher costs. Reading the period, publication date and status of each figure helps prevent an encouraging headline from becoming a misleading investment argument.

What does each company actually sell?

TSMC operates a dedicated semiconductor foundry business. It manufactures chips designed by its customers, serving markets that include high-performance computing, smartphones, automotive and other applications.

An investment in TSMC therefore requires an assessment of manufacturing technology, production capacity and customer demand. Its overall revenue should not be treated as revenue exclusively from AI.

South Korea’s SK hynix supplies memory products, including DRAM and NAND flash. Its portfolio includes high-bandwidth memory, commonly called HBM, which is used alongside advanced processors in demanding computing applications.

Samsung Electronics has a broader mix of businesses. Alongside memory and foundry operations, it has smartphones, displays, televisions, appliances and other activities. Strong performance in one division does not establish that every part of the company is improving.

These examples also show why investors should avoid reducing an entire market to one product. South Korea has foundry operations as well as memory businesses, while Taiwan’s technology industry extends beyond TSMC.

The useful comparison begins with the company and the source of its profits.

Why high-bandwidth memory receives so much attention

A powerful processor needs data to arrive quickly enough for it to work efficiently. If data cannot move fast enough, some of that processing capability can sit unused.

HBM helps address this problem. It stacks layers of DRAM and provides a wide connection for transferring data. In simple terms, memory capacity describes how much data can be held, while bandwidth describes how much can move in a given time.

This is different from NAND flash, which is used for storage. An AI system can need both, but the products serve different purposes and should not be treated as interchangeable.

For investors, a useful question is whether a supplier can turn technical performance into repeat orders and profitable deliveries.

Sending samples to customers, passing customer testing and shipping products in large volumes are different stages. A promising announcement at an early stage should not automatically be counted as a substantial contribution to current earnings.

Where is the growth coming from?

Revenue can rise because a company sells more products, receives higher prices or sells a larger proportion of more expensive products. Several of these changes can happen together.

The distinction is particularly relevant to memory suppliers. Higher selling prices can improve earnings, but investors should examine what would happen if additional supply or weaker demand put those prices under pressure.

A more valuable product mix may improve profitability too. The benefit depends partly on production costs and how successfully the company manufactures the new products.

For a foundry, questions include which technologies customers are ordering, how much production capacity is being used and whether new facilities can operate efficiently.

Investors should also check how the company defines its categories. A broad label such as high-performance computing should not be assumed to mean AI alone. Where a precise AI revenue figure is not disclosed, that uncertainty should remain visible rather than being filled with an estimate presented as fact.

Growth also requires money

Factories, manufacturing equipment and supporting infrastructure require capital expenditure: money spent on assets intended to be used over several years.

That spending can support future revenue, but it also uses cash before the full commercial benefit arrives. Investors need to consider how the company will fund expansion and what demand assumptions support it.

Accounting profit and cash generation can differ. The cost of manufacturing equipment is generally spread across its useful life through depreciation, while payments for the equipment follow their own timetable.

One useful starting point is operating cash flow less capital expenditure, a common simplified measure of free cash flow. Companies may define this measure differently, so comparisons need a consistent basis.

A weak figure during expansion does not automatically mean the investment is poor. It does make the expected return from that expansion more important. The assessment should include existing cash, debt, future spending commitments and the possibility that new capacity takes longer than expected to become profitable.

Does the valuation depend on unusually strong earnings?

A low price-to-earnings ratio can look reassuring. It deserves closer examination when the earnings used in the calculation come from an exceptionally favourable period.

If earnings subsequently fall, the same share price produces a higher ratio. A company that appeared inexpensive on recent profits may look less attractive on a more cautious assessment.

Investors can compare several scenarios: continued growth, slower growth and a decline in earnings. The aim is to understand how much of the investment case depends on favourable conditions continuing.

Forecast-based ratios require care as well. A ratio using next year’s expected earnings should not be compared casually with one using last year’s reported earnings.

The same discipline applies when comparing companies. Differences in product mix, debt, cash requirements and the durability of earnings can justify different valuations. One headline ratio cannot settle the comparison.

Several holdings may depend on the same spending decisions

Owning a Taiwanese foundry and South Korean memory suppliers provides exposure to different businesses. It may still leave an investor sensitive to the same underlying demand for computing infrastructure.

For example, if a group of large customers delays planned data-centre spending, the effects could reach more than one supplier. The size and timing of that impact would depend on contracts, inventories and each company’s customer mix.

Investors should examine customer concentration where it is disclosed, while recognising that an immediate customer and the ultimate source of demand may be different businesses.

The same question applies to a technology fund or exchange-traded fund. Its name may suggest broad exposure, but the actual holdings and their weights determine how concentrated it is.

Counting companies is only the beginning. Understanding which spending decisions support their earnings gives a clearer picture of the portfolio’s risks.

Location and currency still matter

A company’s listing location does not describe its entire business. Investors should examine where it manufactures, where its customers operate and how it depends on suppliers and transport links.

Disruption at a factory, changes in trade restrictions or difficulties delivering equipment could affect costs and production. These are risks to examine through company disclosures, rather than events to assume will happen.

Currency adds another consideration. A return measured in New Taiwan dollars or Korean won may differ from the return received in the investor’s home currency.

Buying a fund or security quoted in another currency does not by itself establish that the underlying currency exposure has been removed. Any hedging arrangements, fees and conversion costs should be checked separately.

These details can affect the realised result even when the original view about the company proves broadly correct.

What should investors check before buying?

Before investing in an AI-related company in either market, ask whether:

• The business’s role in the supply chain is clear, including its activities beyond AI.
• Reported results are distinguished from estimates, guidance and product-development milestones.
• Revenue growth is supported by identifiable changes in volume, prices or product mix.
• Cash generation and funding can support the planned investment in capacity.
• The valuation remains reasonable under a less favourable earnings scenario.
• Other portfolio holdings depend on many of the same customers or spending plans.
• Manufacturing locations, currency exposure and investment costs are understood.

Taiwan and South Korea offer access to companies that perform different tasks within the computing industry. Understanding those differences helps investors judge where an opportunity may lie and what could weaken it.

The strongest investment argument should explain how demand becomes revenue, how revenue becomes cash and why the purchase price leaves room for an acceptable return. That explanation deserves as much attention as the technology itself.