The AI industry chain enters a rational re-evaluation period: The differentiation logic behind the Taiwan stock market's 28,000-point defense battle

The Market Is Rewriting Pricing Rules

This round of comprehensive adjustments in the AI sector is not unfounded. When Broadcom announced AI orders totaling up to $73 billion but experienced a sharp drop in stock price, Wall Street began to realize that the market’s game rules are changing.

Over the past two years, the logic behind the rise of the AI track was simple and straightforward: as long as a company was associated with AI concepts, it could achieve huge valuation premiums in the stock market based on “order growth.” However, with key signals emerging from the financial reports of industry leaders like Oracle and Broadcom, the market’s pricing logic is shifting—from simply chasing growth scale to examining the profitability quality of that growth, investment return cycles, and the certainty of order realization.

Broadcom’s post-earnings plunge did not see funds rushing in to buy the dip on that day, which is a clear indicator of this transition. This chip giant repeatedly emphasized that the company is transitioning from “selling high-margin chips” to “selling systems,” implying that future profit margins may not meet expectations.

The Truth Behind the Break of the 28,000 Point Barrier in Taiwan Stocks

Last Friday, after the four major US stock indices closed in the red, Taiwan stocks staged a thrilling correction today. The weighted index opened with a gap down and once fell more than 500 points during the session, reaching a low of 27,684 points, breaking the 28,000 point level.

This was not just a simple technical breakdown. Electronics heavyweight stocks were hit hardest, with TSMC ADR plunging 4.2%, dragging the cash stock down by 30 NT dollars to 1,450 NT dollars with a gap down. The stock king, Xinhua, saw fierce battles around the 6,600 NT dollar mark, briefly dropping to 6,590 NT dollars before strong support emerged, becoming a bullish indicator for high-priced stocks. Most of the “thousand-dollar stocks” (priced above 3,000 NT dollars) closed in the red, creating a rare scene.

However, it is worth noting that although the market declined, the fall was not as severe as expected. The main reason is that the market still recognizes the demand for AI, but there has been a fundamental change in valuation judgments for different companies in the supply chain.

The Risks of OpenAI Orders Are Being Exposed

Among Oracle’s $523 billion worth of orders, $300 billion come from OpenAI, but the market is beginning to doubt whether these orders can truly translate into substantial, high profits. The long return cycle and lower-than-expected profit margins make the deals between Broadcom and OpenAI uncertain.

Oracle’s new co-CEO attempted to reassure the market, stating that the company has over 700 AI clients, and even if OpenAI defaults, it can “reallocate infrastructure to other clients within hours.” However, this statement exposes a potential risk: OpenAI may “bite off more than it can chew” with these orders.

Public companies deeply tied to OpenAI—including Oracle, SoftBank, Microsoft, and NVIDIA—have seen their stock prices collectively plunge since late October. This is not a coincidence but a collective market pricing of supply chain concentration risk.

The Divide Between Winners and Losers Is Becoming Clear

Against the backdrop of overall adjustments, market funds are undergoing refined structural adjustments. Companies with strong performance, such as Megachip, saw their stock price surge over 8%, reaching a new high of 2,370 NT dollars. This company benefits from the inventory momentum of next-generation smartphones and high-end tablets, with combined revenue of NT$4.415 billion in the first 11 months, up nearly 40% year-over-year, reflecting genuine revenue growth and profit support.

The stock king, Xinhua, also demonstrated resilience. Benefiting from a smoother supply chain, its shipment performance exceeded expectations, leading to a second upward revision of this quarter’s outlook and optimism for 2025 as the peak year, with order visibility extending into the second quarter of next year. Companies with solid fundamentals like these often find support during market corrections.

Meanwhile, funds have not completely exited the market. Oil, electricity, and gas stocks performed the best, rising 3.09%; followed by networking and shipping stocks, which increased by 1.33% and 1.25%, respectively. Funds are simply flowing out of the crowded midstream AI supporting sectors and into assets with clear cash flow, valuations not excessively inflated, and less sensitive to interest rate environments.

This precisely confirms the essence of this transition—funds are not denying the AI industry but are seeking more certain value anchors amid industry chain differentiation.

Google’s Vertical Integration Advantage Is Expanding

From a mid- to long-term perspective, Google possesses what OpenAI lacks most: cash flow and a complete industry chain.

Google’s expected capital expenditure in 2026 will account for 56% of its operating cash flow, the most efficient among the giants. This vertical integration provides extreme cost advantages—Google’s TPUv7’s TCO (total cost of ownership) is about 44% lower than NVIDIA’s G8X servers.

What does this mean? In the long-term competition for AI infrastructure, Google has the strongest cost pricing power. While companies like OpenAI struggle with high computing costs, Google has already secured a dominant position.

The Three Major Year-End Tests Intensify Uncertainty

Taiwan stocks are currently facing three major year-end tests, which together increase market uncertainty.

US stock index fluctuations directly influence foreign capital allocation to Taiwan stocks; the upcoming implementation of IFRS 17 for the life insurance industry next year may also trigger selling pressure. This wave of life insurance stock sales is not a sign of pessimism about fundamentals but a passive adjustment driven by the system. As IFRS 17 and TW-ICS are fully integrated before the end of 2025, stocks classified as FVOCI will no longer realize gains or losses upon sale but will be recorded in capital reserves, cutting off the previous practice of beautifying EPS through stock disposals.

During this week’s “Super Central Bank Week,” the Bank of Japan’s expectation of a 1 basis point rate hike may lead to the exit of interest rate spread trades, further adding to market variables.

Long-term Judgment: This Is Not a Bubble Burst

From a mid- to long-term perspective, the recent correction in the AI sector is not a bubble burst but an inevitable step toward market maturity.

In the future, sector differentiation within AI will become the norm. Companies relying solely on “AI concepts,” with single customer structures and lacking profitability support, may face ongoing valuation compression; whereas companies with core technologies, stable profitability, diversified customer bases, and clear growth paths will stand out through market rationality.

When Taiwan stocks lose the 28,000 point barrier, it does not signify the end of the AI era but marks the beginning of a more rigorous market assessment of every participant. Those companies that can withstand the test will become the protagonists of the next wave.