Artificial intelligence is emerging as the most influential long-term narrative in the US stock market. Since 2023, breakthroughs in generative AI technology have attracted significant capital inflows into related publicly traded companies. As of June 2026, the total market capitalization of companies directly linked to AI in the S&P 500 has surpassed $10 trillion.
However, the "AI stocks" label encompasses companies with vastly different business models, risk profiles, and growth drivers. Simply grouping all AI-related firms together can lead to serious misjudgments in investment decisions.
Why the Compute Infrastructure Layer Is the Biggest Beneficiary of the AI Boom
The compute infrastructure layer includes companies that design AI chips, semiconductor foundries, and manufacturers of high-bandwidth memory chips. This segment sits at the very top of the AI industry chain, with demand growth driven directly by the rigid compute spending of all downstream companies.
NVIDIA stands out as the flagship company in this space. Its GPUs dominate the AI training and inference market, and its data center business has posted over 300% revenue growth in the past six quarters. AMD is gradually gaining market share in inference scenarios with its MI series accelerators, while Broadcom and Marvell focus on custom AI chips (ASICs), delivering lower power consumption and higher efficiency solutions for cloud service providers.
TSMC, the leading semiconductor manufacturer, also plays a central role. Its advanced process nodes (such as 3nm and 2nm) have maintained utilization rates above 90% for extended periods, and orders from AI chip clients have become the main driver of revenue growth. Memory manufacturers like Micron and SK Hynix benefit from surging demand for HBM chips, which are critical components for high-performance AI accelerators.
The core advantage of this segment lies in the verifiability of its revenue. Each chip sold has a clear customer and price, and market demand growth can be tracked directly through order volumes, capacity utilization, and revenue data. The downside is its pronounced cyclicality—when cloud providers’ capital expenditures enter a downturn, chip orders can contract rapidly.
Why Foundation Model Companies Still Face High Uncertainty in Their Business Models
The foundation model layer refers to companies developing large language models, multimodal models, and offering APIs or product services directly to clients. Key players include OpenAI (closely partnered with Microsoft), Anthropic, Google, Meta, and xAI.
The main challenge for this segment is the unclear path to profitability. Training cutting-edge models costs tens of millions, or even hundreds of millions of dollars. While inference services generate substantial revenue, gross margins are pressured by compute costs and pricing competition. As of June 2026, only a handful of leading foundation model companies have achieved overall profitability.
The competitive landscape is also unstable. Open-source models are rapidly approaching the performance of closed-source ones, eroding pricing power for proprietary providers. Enterprise clients often connect to multiple model vendors simultaneously to reduce supplier dependency, making it difficult for any single provider to monopolize market share.
For US stock investors, direct investment opportunities in the foundation model layer are relatively limited. OpenAI and Anthropic are not publicly traded, and Google and Meta do not report AI business revenue separately, instead combining it with other segments. This makes it challenging for investors to accurately assess the financial performance of this layer independently.
How Application Software Companies Leverage AI for Revenue and Cost Optimization
The application software layer comprises companies that integrate AI capabilities into specific work scenarios. This segment covers verticals such as office software, customer service automation, code generation, marketing copywriting, medical diagnostics, and legal document processing.
Two types of companies stand out in this segment. The first are established software giants like Microsoft, Salesforce, Adobe, and Autodesk. They embed AI features into their existing product suites, boosting incremental revenue by raising subscription prices or attracting new users. Microsoft’s Copilot product line is a prime example, with enterprise subscriptions priced significantly higher than standard versions.
The second type are AI-native startups, such as Cursor in generative code assistance and Runway in AI video generation. Some of these firms have already gone public or been acquired, offering investors exposure distinct from traditional software giants.
The core logic of the application software layer is the combined effect of revenue growth and cost control. On the revenue side, AI features support higher pricing or attract users migrating for new functionality. On the cost side, AI automation reduces labor costs in customer service, content moderation, and coding, thereby improving profit margins. This dual effect gives the segment unique advantages in profit improvement potential.
However, competitive barriers pose a long-term risk. When every software company can access similar model APIs, product differentiation may shrink rapidly, making price wars inevitable.
What Role Do Cloud Service Providers Play in the AI Value Chain?
Amazon AWS, Microsoft Azure, and Google Cloud are three cloud service giants that play a pivotal role in the AI value chain. They sit between the infrastructure and application layers: they purchase chips to build data centers and offer Model-as-a-Service (MaaS) to enterprise clients.
Cloud providers’ AI revenue comes from three sources: compute leasing, model hosting services, and sales of proprietary AI products. All three benefit from rising enterprise demand for AI capabilities. In 2025, the AI-related revenue growth rates for the three major cloud providers exceeded 40% year-over-year, far outpacing their traditional cloud businesses.
Their core advantage lies in revenue diversity and customer stickiness. Even if AI compute leasing demand slows in a given quarter, traditional enterprise IT migration, data storage, and analytics still provide stable cash flow. This makes their risk profile smoother compared to pure chip or pure model companies.
Investors should monitor metrics such as cloud business capital expenditure guidance, the impact of AI services on overall profit margins, and the depth of client engagement with MaaS products. When capital expenditure growth slows, it typically signals upstream chip orders will face pressure; conversely, continued expansion suggests the entire value chain remains in a growth cycle.
What Are the Fundamental Differences in Valuation Logic Across AI Segments?
Valuation for the compute infrastructure layer mainly relies on price-to-earnings ratios and alignment with capital expenditure cycles. Chip companies’ revenue and profits are highly cyclical, so the market usually prices them based on forward P/E ratios. During periods of expanding AI compute demand, P/E multiples may reach historical highs; but when the market expects cloud providers to cut procurement, valuations quickly revert.
Foundation model layer valuations are more often based on revenue scale and user growth. Since most companies are not yet profitable, investors use price-to-sales ratios or enterprise value-to-revenue as benchmarks. P/S ratios above 20x are not uncommon, provided the market believes the company’s models can maintain technological leadership and eventually deliver strong profits. The risk here is that if competition drives prices down or users leave, revenue growth cannot support high multiples.
Application software layer valuations blend P/E and P/S ratios. For profitable traditional software companies, P/E remains the core metric, and AI-driven margin improvements can lower P/E, boosting share prices. For AI-native firms, P/S is more common, but the market focuses on customer retention, unit economics, and payback periods for customer acquisition costs.
Cloud provider valuations are influenced by both overall cloud computing growth and the incremental value from AI. The market often uses a sum-of-the-parts approach, valuing cloud and consumer internet businesses separately. AI’s contribution to cloud business growth is the main source of current valuation premiums.
How Does Capital Flow and Rotate Among AI Stock Segments?
Tracking capital flows in the US stock market from 2023 to 2026 reveals a clear pattern of sector rotation.
Phase one (early 2023 to mid-2024) saw capital highly concentrated in the compute infrastructure layer. The market’s core question was "Who can supply the compute needed to train large models?" During this period, chip companies like NVIDIA outperformed other segments by a wide margin.
Phase two (late 2024 to end of 2025) saw capital spreading to cloud providers and the application software layer. The market realized that compute is just the first step; platforms and tools that turn AI into sustainable revenue may create lasting value. Microsoft, Amazon, and Salesforce attracted significant inflows during this stage.
Phase three (2026 to present) features layered allocation. Large institutional investors now deploy capital across all three segments, but weightings depend on their views of AI commercialization progress. Investors bullish on persistent compute shortages overweight chip stocks; those optimistic about model monetization shift toward cloud providers; those focused on software efficiency gains pay more attention to the application layer.
This rotation pattern offers a key insight for retail investors: there’s no need to treat the AI narrative as a monolith. Instead, understand the driving cycle of each segment. When signs of slowing compute demand emerge, reducing upstream exposure and increasing downstream allocations may be a sensible rebalancing strategy.
Which Metrics Should Retail Investors Focus on When Screening AI Stocks?
Based on the analysis above, you can build a screening framework starting from segment characteristics:
For the compute infrastructure layer, focus on order backlog, capacity expansion plans, customer concentration, and inventory turnover days. Order backlog reflects revenue visibility for the next two to four quarters; capacity expansion plans indicate management’s view of long-term demand.
For cloud providers, monitor capital expenditure guidance, the proportion of AI service revenue, and trends in operating margins. Expanding capital expenditures usually signal upstream strength, but if margins fall due to over-investment, valuations may be affected.
For the application software layer, focus on AI feature pay conversion rates, customer retention, and differentiation versus competitors. Pay conversion rate is a key indicator of real user demand; feature differentiation determines whether a company can maintain pricing power in long-term competition.
Universal metrics for all segments include insider trading records, analyst rating changes, and institutional holding shifts. These can be found in earnings reports and SEC filings, serving as important supplementary indicators of changing capital sentiment.
Summary
AI stocks are not a single category, but a value chain composed of four core segments: compute infrastructure, foundation models, application software, and cloud services. Each segment differs fundamentally in revenue verifiability, competitive landscape, valuation methods, and capital rotation rhythm.
The compute infrastructure layer benefits from rigid demand but is highly cyclical; the foundation model layer represents the technological frontier but faces unclear paths to profitability; the application software layer combines revenue growth and margin improvement but risks feature commoditization; cloud providers offer diversified income and high customer stickiness but face ongoing capital expenditure pressures.
For investors trading US AI stocks via Gate, understanding this layered logic helps build allocation strategies that better match your risk preferences. In any segment, making decisions based on fundamentals rather than market sentiment is the basic principle for controlling risk.
Frequently Asked Questions (FAQ)
Q: Which US AI-related stocks are available on Gate?
A: Gate’s US stock trading section offers core AI value chain stocks including NVIDIA (NVDA), AMD, Microsoft (MSFT), Google (GOOGL), Amazon (AMZN), Salesforce (CRM), Adobe (ADBE), and more. You can view the complete list and real-time market data on the platform.
Q: Is it too late to invest in AI stocks now?
A: AI commercialization is still in its early stages. While some stocks have already seen significant gains, penetration rates vary widely across segments of the value chain. We recommend using the segment analysis framework above and factoring in your own risk tolerance and investment horizon for independent judgment.
Q: What are the main sources of correction risk for AI stocks?
A: Risks include overvaluation, slowing capital expenditure growth by cloud providers, intensified competition leading to margin erosion, slower-than-expected AI application rollout, and broad market adjustments triggered by macroeconomic shifts.
Q: How can I get the latest market data for AI stocks on Gate?
A: Log in to the Gate platform, go to the US stock trading section, and search for relevant stock tickers to view real-time prices, historical trends, and fundamental data. All market data is updated in real time from the exchanges.
Q: Is it better to invest in a single AI segment or diversify across multiple segments?
A: There’s no one-size-fits-all answer. If you have deep insight into a particular segment, concentrated allocation can make sense. But given the cyclical differences among segments, diversification helps smooth portfolio volatility. We suggest making allocation decisions based on your research capabilities and risk preferences.
