AI Stocks July 12, 2026
The Idea
Artificial intelligence has moved from “promising technology” to core business infrastructure. In 2026, AI is reshaping software, chips, cloud computing, cybersecurity, healthcare, advertising, industrial automation, and financial services. But for long-term investors, the key question is not whether AI is important. It is whether a company can turn AI demand into durable cash flows.
The most useful approach is to evaluate AI firms the way Warren Buffett and Charlie Munger evaluated any business: focus on economics, competitive advantage, management discipline, and price. AI may be new, but valuation math is not. A great technology can still be a poor investment if expectations are too high or margins are structurally weak.
How It Works
Evaluating AI firms for long-term value starts with separating AI users, AI infrastructure providers, and AI-native platforms. Each group requires different metrics.
- Revenue quality: Look for recurring, contract-based, or usage-driven revenue that expands over time. A company with high renewal rates and rising customer spend is usually more attractive than one relying on one-time implementation fees.
- Gross margin durability: AI can be computationally expensive. Software firms with 75% gross margins may see those margins pressured by inference costs. Infrastructure providers may have lower margins but benefit from scale. Investors should ask: does each additional customer become more profitable over time?
- Free cash flow conversion: Earnings can be noisy in high-growth technology. Free cash flow shows whether the business is funding itself or constantly depending on capital markets. Seth Klarman’s margin-of-safety framework is especially relevant here: cash generation reduces the need for optimistic assumptions.
- Customer retention and net revenue retention: For enterprise AI companies, net revenue retention above 100% suggests existing customers are expanding usage. But investors should also examine whether growth comes from mission-critical use cases or experimental budgets.
- Return on invested capital: AI firms may spend heavily on data centers, model training, sales teams, and talent. High revenue growth is less valuable if it requires endless reinvestment at mediocre returns. As Aswath Damodaran often emphasizes, value comes from growth only when that growth earns more than the cost of capital.
- Data advantage and switching costs: Durable AI companies often improve as customers use their products. Proprietary data, workflow integration, regulatory approvals, and ecosystem lock-in can create moats. Without these, AI tools risk becoming commodities.
- Valuation versus realistic growth: Price-to-sales ratios alone are incomplete, but they are useful warning signs. Investors should model what revenue, margins, and cash flow must look like five to seven years out to justify today’s price.
The practical logic is simple: favor AI businesses that combine real customer demand, improving unit economics, strong balance sheets, and defensible advantages. Avoid companies whose investment case depends entirely on being “an AI play.”
Why This Has Worked Historically
History shows that transformative technologies do not automatically create broad investment winners. Benjamin Graham’s core lesson in The Intelligent Investor was that investors should distinguish between a good business narrative and a good investment price. That lesson mattered during the dot-com bubble, and it matters again in the AI era.
A concrete example is Amazon after the 2000 technology crash. The internet was clearly revolutionary, but many internet stocks disappeared because they lacked viable economics. Amazon survived because it built scale, customer trust, logistics capabilities, and eventually cloud infrastructure. Investors who focused only on near-term losses missed the longer-term operating leverage; investors who bought any internet stock indiscriminately were often wiped out.
Another example is Nvidia’s evolution from gaming graphics chips to accelerated computing. The long-term value creation did not come merely from being associated with a hot trend. It came from a combination of product leadership, software ecosystem depth, developer adoption, and expanding end markets. The lesson is not to chase yesterday’s winner, but to identify the underlying qualities that allowed the winner to compound.
Peter Lynch’s advice to “know what you own” is useful here. In AI, investors should understand whether a company sells foundational hardware, rents compute capacity, embeds AI into enterprise workflows, or applies AI to a specific industry. Different models have different margins, capital needs, and competitive threats.
How to Apply It Today
In July 2026, investors can apply a disciplined AI framework without trying to predict every technological breakthrough. Start with a simple checklist.
- Classify the business: Is it infrastructure, platform software, application software, data provider, semiconductor equipment, cloud services, or consulting? Avoid lumping all AI exposure together.
- Read the revenue notes: Look for customer concentration, contract duration, remaining performance obligations, usage-based revenue trends, and renewal commentary.
- Track AI cost structure: For software firms, monitor gross margin trends as AI features scale. For infrastructure firms, monitor capital expenditures, depreciation, utilization, and pricing pressure.
- Compare growth to spending: If revenue growth requires rapidly rising stock-based compensation, sales expenses, or capital investment, future shareholder returns may be diluted.
- Stress-test valuation: Build a conservative scenario. What happens if growth slows, margins peak lower than expected, or competition reduces pricing power? Howard Marks’ risk framework is relevant: risk often comes from paying too much for a widely loved asset.
- Look for customer proof: Strong AI companies should be able to show measurable customer outcomes: lower costs, higher productivity, better fraud detection, faster drug discovery, improved cybersecurity, or higher conversion rates.
- Prefer optionality with discipline: The best AI firms may have multiple growth paths, but management should still allocate capital rationally. Munger often warned against confusing activity with progress.
A useful rule of thumb: the more exciting the story, the more demanding the evidence should be. If a company trades at a premium valuation, it should have premium growth, premium margins, premium retention, and a credible path to premium free cash flow.
Risks to Keep in Mind
The first risk is commoditization. AI models, tools, and interfaces can improve quickly, but competition can also reduce differentiation. If customers can switch vendors easily, early growth may not translate into long-term profitability.
The second risk is capital intensity. Training and serving AI models can require massive spending on chips, data centers, energy, and engineering talent. A company can grow revenue rapidly while still producing disappointing shareholder returns if reinvestment needs remain too high.
The third risk is regulatory and legal uncertainty. Data privacy, copyright, model transparency, cybersecurity, and sector-specific compliance can all affect adoption. AI firms operating in healthcare, finance, defense, and education may face especially high scrutiny.
The fourth risk is narrative valuation. Nassim Taleb’s work reminds investors to respect uncertainty and avoid overconfidence. AI forecasts often sound precise, but small changes in assumptions can dramatically alter intrinsic value estimates. A business can be real, useful, and growing, yet still be overpriced.
Finally, investors should watch for “AI washing.” Companies may use AI language to make ordinary software, consulting, or automation businesses appear more innovative than they are. Long-term value is not created by terminology. It is created by customers paying more, staying longer, and generating attractive returns on the capital required to serve them.
