Semiconductor Stocks September 06, 2026
The Idea
Value investors are often portrayed as hunters of cheap banks, industrials, insurers, and overlooked cash-flow machines. But the core discipline is not about buying “old economy” stocks. It is about paying less than a business is worth, with a margin of safety. That is why emerging AI startups deserve serious attention this fall.
The opportunity is not to chase every company with “AI” in its pitch deck. It is to study where durable value may be forming before it becomes obvious. Artificial intelligence is moving from experimentation to infrastructure, workflow automation, drug discovery, cybersecurity, software development, customer support, and industrial productivity. Many startups will fail. A small number may become essential businesses.
For value investors, the goal is simple: identify real economics beneath the excitement. The best AI opportunities may not be the loudest. They may be companies solving expensive problems, reducing labor intensity, improving decision quality, or creating software that becomes deeply embedded in customer operations.
How It Works
A value approach to emerging AI startups starts with business quality, not valuation shortcuts. Traditional metrics like price-to-earnings may be useless for young private companies or newly public firms that are still reinvesting aggressively. Instead, investors should focus on evidence of future cash-generation potential.
Think of it as underwriting a business model. The key questions are practical:
- Is the product mission-critical? AI tools used for compliance, fraud detection, code generation, medical analysis, or security may have stronger staying power than novelty applications.
- Does the company have pricing power? If customers save meaningful time or money, the vendor can often capture part of that value.
- Are switching costs rising? A startup that becomes integrated into workflows, data systems, and employee habits can build a moat over time.
- Is there proprietary data or distribution? In AI, access to unique data, trusted customer relationships, or embedded channels can matter as much as the model itself.
- Can unit economics improve? Investors should watch gross margins, customer acquisition costs, retention, usage trends, and compute expenses.
This is where Benjamin Graham’s margin of safety meets modern technology analysis. A company can be innovative and still be a poor investment if expectations already assume perfection. Conversely, a misunderstood AI business with sticky customers and improving economics may offer value even before it looks statistically cheap.
Why This Has Worked Historically
Some of the best long-term investments looked expensive or uncertain early on, yet rewarded investors who understood the economics before the consensus did. Peter Lynch wrote in One Up on Wall Street that individuals could gain an edge by observing powerful business trends before Wall Street fully appreciated them. That principle applies to AI adoption today, provided investors remain disciplined.
A useful historical example is Amazon after the dot-com crash. In 2001 and 2002, many investors dismissed it as another internet bubble casualty. The stock had collapsed, losses were significant, and sentiment was awful. But investors who focused on customer loyalty, scale advantages, fulfillment infrastructure, and long-term cash-flow potential saw something different: a company building a durable retail and technology platform. The market eventually recognized those economics.
Another example is Salesforce in the early cloud era. Many traditional investors were skeptical of subscription software because reported profits were limited and sales spending was high. But the recurring revenue model, high retention, and expanding customer relationships created enormous value. The lesson is not that every hot technology company becomes a winner. It is that new categories can look unattractive under old accounting lenses while quietly building durable assets.
Warren Buffett and Charlie Munger often emphasized staying within one’s circle of competence. That does not mean avoiding technology forever. It means expanding knowledge carefully. Buffett eventually invested in Apple after recognizing not just a hardware company, but a consumer ecosystem with loyalty, pricing power, and recurring services. The same mindset can help investors evaluate AI: avoid what cannot be understood, but do not ignore economic reality because the category is new.
How to Apply It Today
This fall, value investors should build an AI watchlist rather than a buy list. The objective is to monitor business progress, valuation resets, and evidence of durable advantage.
- Follow the customer, not the demo. Look for startups gaining adoption among enterprises with strict procurement standards. Real customers are more meaningful than viral attention.
- Track retention and expansion. The best software businesses often grow because existing customers spend more over time. Net revenue retention can be more revealing than headline growth.
- Separate infrastructure from applications. AI chip demand, cloud platforms, model providers, and vertical applications have different economics. Each layer will not capture equal value.
- Watch compute costs. A company with impressive revenue but poor gross margins may struggle if inference costs remain high or pricing pressure increases.
- Study management incentives. Seth Klarman has long stressed risk control and discipline. In startups, investor-friendly governance, rational capital allocation, and realistic communication matter.
- Wait for better entry points. Howard Marks reminds investors that price is what turns a good asset into a good or bad investment. Patience is an edge when hype is high.
Investors without access to private markets can still benefit. Study IPO filings, venture funding trends, customer case studies, partnership announcements, and earnings calls from larger companies that buy from or compete with AI startups. The goal is to understand where value is migrating before public markets fully price it.
Risks to Keep in Mind
The biggest risk is confusing technological promise with investment merit. Nassim Taleb’s work on uncertainty is relevant here: outcomes in emerging technologies can be highly skewed, with many failures and a few huge winners. Forecasting the winner too confidently can be dangerous.
AI startups also face rapid commoditization. If multiple companies can build similar features on top of the same foundation models, pricing power may evaporate. A flashy product today can become a built-in feature from a larger software platform tomorrow.
Valuation is another risk. Even excellent businesses can disappoint investors if purchased at prices that assume flawless execution. As Aswath Damodaran often argues, narratives must eventually connect to numbers. Growth, margins, reinvestment needs, and risk all matter.
Finally, regulation, data privacy, intellectual property disputes, and customer trust will shape the industry. Value investors should prefer companies that can survive scrutiny, not merely ride enthusiasm. Emerging AI startups are worth watching this fall because some may become tomorrow’s compounding machines. But the value investor’s advantage remains unchanged: patience, skepticism, and a relentless focus on business economics.
