2025 AI Stock Investment Map: From Hardware Positioning to Application Implementation, What Are Institutions and Retail Investors Betting On?

Still debating whether to invest in AI stocks? Instead of blindly following the trend, it’s better to see how major institutions are acting.

According to the latest data, Bridgewater Associates adjusted its holdings in Q2 2025, significantly increasing positions in key companies along the AI supply chain such as NVDA, Alphabet, and Microsoft. The logic behind this is simple: AI is shifting from concept to reality, from hype to demand, with hardware infrastructure becoming the first wave of dividends.

Distribution of AI Industry Chain Dividends: Who Can Most Benefit?

Don’t be fooled by the three words “AI stocks.” True AI stock investment is actually investing in the “iron rooster” of the AI era.

According to IDC’s latest forecast, global corporate spending on AI solutions will reach $307 billion in 2025. Where will this money go? It mainly flows into three areas:

Layer 1: Chips and Accelerators. Training large models requires computing power, and inference also needs it. GPUs from NVDA, AMD’s MI series, and downstream ASIC chip design companies are direct beneficiaries of this round.

Layer 2: Infrastructure Providers. Servers’ cooling, power management, network switches—these seemingly insignificant components are the foundation of AI data center operations. That’s why orders for companies like Delta Electronics, Sunway, and Broadcom are booming.

Layer 3: Cloud Platforms and Applications. Azure, Google Cloud, various AI SaaS—these are the “money-making machines” that convert chips and computing power into real revenue.

By 2028, total AI expenditure is expected to surpass $632 billion, with over 75% spent on accelerating servers. In other words, infrastructure layer profits are the easiest and most sustainable.

U.S. AI Leaders: Who Are the Real Cash Cows?

NVDA (Nvidia): If you had to pick an AI stock representative, it’s this one. In 2024, revenue reached $60.9 billion, up over 120% year-over-year. Entering 2025, the company continues to break through—Q2 revenue of $28 billion, net profit up over 200%. Demand for Blackwell architecture GPUs remains hot, with cloud service providers lining up to buy.

Honestly, NVDA’s position is almost unshakable in the short term. From chip design, CUDA software ecosystem, to system integration, it has built a complete moat. Several institutions target its stock price above $600.

AMD: As a challenger, AMD’s offensive has become more aggressive. The MI300X accelerator is being adopted by major cloud providers, and MI350 is coming in the second half. Although Q2 2025 revenue only grew 18% YoY, AI-related business doubled. The key is that it provides customers with alternatives, which is significant for industry competition. Target prices mostly above $200.

Broadcom (AVGO): This company is often overlooked but plays a key role in AI data centers. Custom ASICs, network switches, optical communication chips—an all-in-one solution. AI product revenue accounted for 25% of FY2024. Q2 revenue grew 19% YoY, steady growth though not as explosive as NVDA. Target prices generally above $2,000.

Microsoft (MSFT): The biggest winner in enterprise AI monetization. Exclusive partnership with OpenAI, deep integration of Copilot into Office/Windows ecosystem, exponential growth in Azure AI usage. In Q1 2025, intelligent cloud revenue broke $30 billion for the first time. Most institutions see MSFT as the most certain beneficiary of “enterprise AI popularization,” with target prices between $550 and $600.

Taiwan Stock AI Opportunities: Sector Rotation and Sub-sectors

Taiwan’s AI story isn’t about chip design (that’s TSMC’s domain), but ecosystem support and application deployment.

Quanta Computer (2382): From a notebook OEM giant to a new player. Through its subsidiary Quanta Cloud Technology (QCT), Quanta successfully entered the AI server supply chain. In 2024, revenue reached NT$1.3 trillion, with AI servers’ share continuously rising. In Q2 2025, revenue exceeded NT$300 billion, up over 20% YoY, hitting a new high for the same period. Foreign investors’ target price is NT$350-370, with room for growth.

Unigroup-KY (3661): Taiwan’s purest AI concept stock. ASIC chip design, serving top global cloud and HPC giants. In 2024, revenue was NT$68.2 billion, up over 50%. In Q2 2025, quarterly revenue surpassed NT$20 billion, doubling from last year, with gross margin and net margin both improving. Foreign investors’ target price NT$2,200-2,400. Its growth rate is among the top in Taiwan stocks.

Delta Electronics (2308): A versatile player in power supplies, cooling, and cabinet solutions. In 2024, revenue hit NT$420 billion, with data center business rapidly increasing. In Q2 2025, revenue NT$110 billion, up over 15% YoY. Gross margin remains high, showing it’s a company truly profiting from the AI era.

MediaTek (2454): A layout player in edge AI and mobile AI. The Dimensity series now includes enhanced AI computing units, and it collaborates with NVIDIA on automotive AI solutions. In 2024, revenue was NT$490 billion; in Q2 2025, NT$120 billion, up about 20% YoY. Long-term growth supported by mobile AI and automotive AI, with target prices NT$1,300-1,400.

Sunway (3324): The hidden champion of liquid cooling. AI server chips’ power consumption exceeds 1 kW, and traditional air cooling has become a bottleneck, making liquid cooling a must. In 2024, revenue NT$24.5 billion, up over 30%. As mainstream cloud providers adopt liquid cooling on a large scale in 2025, this company’s prospects look even brighter. Foreign target price above NT$600.

How to Snag AI Investment Opportunities? Three Strategies Compared

Strategy 1: Buy individual stocks directly

• Pros: Self-selected, low transaction costs, no dilution of gains
• Cons: Research pressure, high risk of single stock, easy to miss opportunities
• Suitable for: Investors with time to research and able to tolerate volatility

Strategy 2: Buy thematic funds

• Pros: Fund managers select a diversified portfolio, balancing risk and return, hassle-free
• Cons: Higher management fees, performance depends on manager’s skill
• Suitable for: Those wanting to participate in AI but lacking stock-picking skills
• Example: First Financial Global AI Robotics and Automation Industry Fund

Strategy 3: Buy AI-themed ETFs

• Pros: Lowest transaction costs, passive index tracking, low fees, high liquidity
• Cons: Prone to premiums or discounts during trading
• Suitable for: Those seeking low-cost long-term participation
• Examples: Taishin Global AI ETF (00851), Yuan Da Global AI ETF (00762)

Morningstar data shows that as of the end of Q1 2025, global AI and big data funds have assets exceeding $30 billion, indicating that the market’s view of AI investment has shifted from “concept” to “asset allocation.”

The smartest approach is dollar-cost averaging—whether choosing stocks or funds, building positions gradually to average costs, participating in long-term trends while avoiding short-term volatility.

The Other Side of the Coin: Three Major Pitfalls in AI Stock Investment

Not all AI stocks are worth holding long-term. Lessons from the internet bubble are instructive.

Cisco Systems (CSCO) soared to $82 during the 2000 dot-com bubble, then fell 90%. Yahoo was once an internet giant but was eventually crushed by Google. What does this tell us? In early industry stages, hardware and infrastructure companies’ valuations tend to overshoot, and bubbles burst with sharp declines.

Three major risks facing AI stocks:

Risk 1: Uncertain technological path. What’s the next step for large models? Optimization of inference, multimodal evolution, or a completely different direction? No one can guarantee. If the tech direction shifts, some chips and solutions may become obsolete.

Risk 2: Policy and regulation. Governments worldwide see AI as a strategic industry and are increasing subsidies. But at the same time, regulations on data privacy, algorithm bias, and copyright are tightening. Once new rules are implemented, some AI companies’ business models could be impacted.

Risk 3: Overstretched valuations. Many AI stocks have already priced in growth for the next three to five years. If growth slows, stock prices could adjust rapidly.

Investment Logic for AI Stocks (2025-2030)

Short-term (2025-2026): Infrastructure dividends release, chip and server suppliers profit most. Capital continues to flow in, but volatility will increase—macroeconomic policies, news, and capital flows will cause short-term shocks.

Mid-term (2027-2028): AI applications gradually land in industries like healthcare, finance, manufacturing, and autonomous driving, generating real revenue. The entire chain from hardware to applications benefits, but growth rates will diverge.

Long-term (2029-2030): AI becomes a basic infrastructure, as common as electricity. Companies selling AI will see profits squeezed; the real winners are downstream enterprises using AI to improve efficiency.

Investment advice:

1. Chasing highs in the short term is a dead end. AI stocks react sensitively to news, with large fluctuations. Buying at peaks often results in being trapped.

2. Prioritize infrastructure. NVDA, AMD, TSMC, Broadcom—these hardware suppliers remain the safest bets in 2025-2026.

3. Moderately focus on application layer. Companies like Microsoft that can directly monetize AI, though growth may be slower than chips, tend to have lower risk.

4. Diversify your investments. Using AI-themed ETFs or funds can effectively reduce single-stock volatility, suitable for most retail investors.

5. Dollar-cost averaging over timing. Avoid putting all your money in at once; building positions gradually can significantly reduce missed opportunities and being caught in downturns.

Ultimately, the core of AI investing is: Know what you are investing in, when to get on board, and when to get off. The infrastructure wave’s dividends are real, but so are valuation bubbles. Stay alert to hype, cherish opportunities.