Is 2025 AI stocks still worth bottom-fishing? From the industry chain perspective, the real opportunities in AI concept stocks

Since the advent of ChatGPT, the industry investment boom around artificial intelligence has continued for over two years. However, unlike other tech waves, the performance of AI concept stocks has shown clear differentiation—some companies’ stock prices have multiplied several times, while others have fallen into difficulties amid hype. The question is: Which AI concept stocks are still worth关注ing? How should investment logic be adjusted?

Why Have AI Concept Stocks Become an Investment Focus?

The definition of artificial intelligence (AI) is no longer unfamiliar—enabling machines to acquire human-level cognitive abilities such as learning, reasoning, problem-solving, and language understanding. From voice assistants to autonomous driving, from medical diagnostics to financial forecasting, AI applications have penetrated all aspects of life.

What are AI concept stocks essentially? They are not the creators of AI technology but beneficiaries along the AI industry chain—chip designers, server manufacturers, cloud infrastructure providers, and even suppliers of cooling systems and power solutions. In other words, investing in AI concept stocks is investing in the “shovel business” behind this technological revolution.

According to the latest IDC data, global enterprise investment in AI-related technologies and solutions will reach $307 billion by 2025, surpassing $632 billion by 2028, with a compound annual growth rate (CAGR) of 29%. Among these, spending on accelerated servers will account for over 75% by 2028, becoming the core hardware supporting AI deployment. This data illustrates a fact—there is still enormous growth potential in the AI industry chain.

Differentiation in the AI Industry Chain: Who Are the True Beneficiaries?

To understand investment opportunities in AI concept stocks, it is first necessary to grasp the true structure of the industry chain.

Upstream (chips and basic hardware): This is currently the most capital-intensive area. From GPUs to customized ASIC chips, and high-efficiency cooling systems, these are the core components of AI servers. Industry giants like NVIDIA, Broadcom, and AMD rely on technological barriers to form relatively stable market positions.

Midstream (servers and data center infrastructure): Companies like Quanta (2382) and Wistron-KY (3661) are in this position, responsible for integrating chips into complete AI server solutions. As cloud service providers (such as Microsoft Azure, Google Cloud) accelerate the expansion of AI data centers, their order volumes continue to rise.

Downstream (AI applications and enterprise services): Tech giants like Microsoft and Google deliver AI capabilities directly to end-users through products like Copilot and Azure AI. These companies have stronger monetization capabilities, but competition is also fiercer.

According to Bridgewater Associates’ 13F report for Q2 2025, they increased their holdings in core nodes such as computing power, chips, and cloud computing, including holdings in NVIDIA, Alphabet, and Microsoft. This trend reflects the consensus among institutional investors: In the short term, upstream hardware and infrastructure companies remain the most direct beneficiaries.

Taiwanese AI Concept Stocks: Growth Momentum vs. Valuation Risks

Which Taiwanese companies are truly benefiting from the AI wave? Here are some market-recognized names:

Quanta (2382) is the top player in Taiwan’s AI server field. This global largest notebook OEM has successfully pivoted, with its subsidiary QCT focusing on servers and cloud solutions, serving clients like NVIDIA and international cloud providers. In 2024, Quanta’s revenue is expected to reach NT$1.3 trillion, with a continuous increase in AI server share. By 2025, Q2 revenue surpassed NT$300 billion, up over 20% year-on-year, reaching a new high for the same period. Foreign analysts’ target prices range from NT$350 to NT$370, leaving room for upside.

Wistron-KY (3661) specializes in ASIC customized chip design, with clients mainly being global cloud giants and leaders in high-performance computing. In 2024, full-year revenue reached NT$68.2 billion, up over 50%. In Q2 2025, quarterly revenue exceeded NT$20 billion, doubling compared to the same period last year, with gross margin and net profit margin continuing to rise. As generative AI applications expand, the market is optimistic about its long-term growth, with target prices between NT$2,200 and NT$2,400.

Delta Electronics (2308) has successfully entered the AI server supply chain from its leadership in power management, mainly providing high-efficiency power supplies, cooling, and cabinet solutions. In 2024, full-year revenue is about NT$420 billion, with data center and AI application performance steadily increasing. In Q2 2025, revenue is approximately NT$110 billion, up over 15% year-on-year, with high gross margins reflecting strong AI infrastructure demand.

MediaTek (2454), although known for mobile chips, has actively布局 in AI in recent years. Its Dimensity series now includes enhanced AI computing units and collaborates with NVIDIA on automotive and edge AI solutions. In 2024, revenue is NT$490 billion, with gross margins improving quarter by quarter. In Q2 2025, revenue is about NT$120 billion, up 20% year-on-year. Analysts are optimistic about its dual-drive potential in mobile AI and automotive AI, with target prices between NT$1,300 and NT$1,400.

Airoha (3324) might be the most overlooked AI concept stock in the market. This cooling solutions leader has successfully positioned itself in the global AI server supply chain with its advanced liquid cooling technology. In 2024, revenue reached NT$24.5 billion, up over 30%. As major cloud service providers accelerate adoption of liquid cooling solutions in 2025, shipments of water-cooled AI server modules from Airoha have surged. Foreign analysts’ target prices are mostly above NT$600, reflecting confidence in its profitability.

US AI Concept Stocks: Technological Barriers and Market Share

In the US stock market, the AI landscape is dominated by absolute technological leaders:

NVIDIA (NVDA) needs no introduction. Its GPUs and CUDA software platform have become industry standards for AI training and inference. In 2024, revenue reached $60.9 billion, up over 120% year-on-year. In Q2 2025, revenue hit a new high of about $28 billion, with net profit increasing over 200%. The GPU architectures (B200, GB200) are in high demand from cloud providers and large enterprises, with data center business continuously setting records. Analysts expect exponential growth in demand as AI shifts from training to inference, and its dominant position is widely recognized.

Broadcom (AVGO) plays a key role in AI chips and network connectivity. In fiscal 2024, revenue was $31.9 billion, with AI-related product revenue rapidly increasing to 25%. In Q2 2025, revenue grew 19% year-on-year, driven by cloud providers accelerating AI data center deployments and demand for Jericho3-AI chips and Tomahawk5 switches. Foreign target prices are mostly above $2,000.

AMD (Advanced Micro Devices) is a challenger in the AI accelerator market. Its Instinct MI300 series and advanced CDNA 3 architecture have successfully entered the predominantly NVIDIA-led market, providing cloud providers with a secondary supply source. In 2024, revenue was $22.9 billion, with data center business up 27%. In Q2 2025, revenue increased 18% year-on-year, with MI300X accelerators adopted by major cloud providers, leading to multiple-fold growth in AI-related revenue. The market is optimistic about its expanding share as an alternative, with target prices mostly above $200.

Microsoft (MSFT), through its exclusive partnership with OpenAI and Azure AI platform, has become a platform for enterprise AI transformation. In fiscal 2024, revenue was $211.2 billion, with Azure and other cloud services growing 28%. In Q1 2025, intelligent cloud revenue first exceeded $30 billion, driven by large-scale deployment of Copilot and exponential growth in Azure OpenAI usage. Most institutions see Microsoft as the most certain beneficiary of the “enterprise AI popularization” wave, with target prices around $550 to $600.

The Real Risks of AI Concept Stocks: Not All Companies Can Outperform Long-term

Investing in AI concept stocks requires acknowledging a historical lesson—not all companies along the industry chain can create long-term value.

Referring to Cisco Systems (CSCO) during the dot-com bubble, this network equipment leader hit a peak of $82 in 2000, representing the cutting-edge “shovel business” at the time. But after the bubble burst, its stock plummeted over 90% to $8.12. Even after 20 years of sustained good management, the stock has not returned to its former high. What does this tell us? While upstream hardware suppliers benefit in the short term, high growth is difficult to sustain permanently.

The same lesson applies at the internet application layer. Yahoo, once a leading internet company, was eventually overtaken by Google (GOOGL), with its stock showing a pattern of “soaring then continuously declining.” Even giants like Microsoft and Google experience significant pullbacks at market peaks, with years of difficulty in recovery.

What is the key takeaway? Companies in different positions along the industry chain show vastly different long-term performance. Upstream hardware vendors face risks from technological iteration and increased competition, while downstream application companies, though strong in monetization, must continuously innovate to maintain competitiveness.

How to Invest in AI Concept Stocks Scientifically?

Given the high volatility and long-term uncertainty of AI concept stocks, investment strategies should be adjusted accordingly:

Direct stock picking: Advantages include low transaction costs and ease of trading, but risks are concentrated in individual companies. Suitable for investors with in-depth knowledge of specific companies.

Theme-based funds: Managed by fund managers selecting a portfolio of stocks, offering relatively diversified risk but higher transaction costs. The First Financial Global AI Robotics and Automation Industry Fund is a typical example.

ETFs: Track indices passively, providing risk diversification with lower transaction costs and management fees. Products like Taishin Global AI ETF (00851) and Yuan Da Global AI ETF (00762) allow investors to deploy across multiple segments such as AI applications, infrastructure, cloud, and big data in one go.

Recommended approach: Use a dollar-cost averaging (DCA) strategy, gradually entering the market to average out purchase prices. The holdings changes of Bridgewater also show that although AI continues to develop rapidly, the positive news is not necessarily concentrated in the same companies—some stocks may have already priced in the AI optimism. Staying updated and adjusting accordingly can maximize investment performance.

Regarding trading platforms, for Taiwan stocks, open an account with a Taiwanese broker. For US stocks, use a Taiwanese broker’s omnibus account, an overseas broker, or a contract for difference (CFD) platform. Short-term traders may consider CFD platforms, as they allow long and short trading, no commission fees, and higher leverage.

Outlook for AI Concept Stock Investment from 2025 to 2030

From an industry development perspective, AI technology will inevitably change human life and production modes like the internet did—this is almost universally accepted. But from an investment standpoint, the future pattern will feature “long-term bullishness with short-term volatility.”

Short-term opportunities: Rapid advances in large language models, generative AI, and multimodal AI will continue to drive demand for computing power, data centers, cloud platforms, and dedicated chips. During this phase, chip and hardware suppliers like NVIDIA, AMD, and TSMC will remain the biggest beneficiaries.

Medium to long-term opportunities: As AI applications in healthcare, finance, manufacturing, autonomous vehicles, and retail gradually materialize, they will generate tangible revenue for enterprises, fueling the overall growth of AI concept stocks. Companies with concrete, sustainable applications that can continuously create value will be more favored.

Unavoidable risks:

• Industry uncertainty—AI technology develops rapidly, making it difficult even for the most knowledgeable investors to keep pace, leading to stock price swings driven by hype.
• Unproven companies—Some AI startups have short histories and weak fundamentals, with higher operational risks than established tech giants.
• Macroeconomic policy changes—Interest rate policies by the Federal Reserve and other central banks directly impact high-valuation tech stocks; issues like data privacy, algorithm bias, and copyright may lead to stricter regulations.
• Capital flow risks—New energy and other emerging themes may attract funds, causing short-term fluctuations in AI stocks.

Advice for general investors: Focus on long-term allocation rather than short-term gains. Prioritize chipmakers, server providers, or companies with tangible applications like cloud services and fintech. Diversify through AI-themed ETFs to reduce the risk of individual stock volatility. Only by doing so can investors participate in AI growth dividends while avoiding shocks from short-term market turbulence.