← Back to list
Oakmark FundsQuarterly30 Jun 2026Source: oakmark.com

The certainty trap | Fixed income market commentary 2Q 2026

Oakmark is the mutual fund family launched in 1991 by Harris Associates, the Chicago deep-value firm founded in 1976 (about $105bn AUM). Bill Nygren runs the flagship Oakmark Fund and David Herro the Oakmark International Fund, buying businesses at large discounts to intrinsic value and holding them like owners — publishing quarterly fund commentaries, market commentaries and insight articles.

Bill Nygren、David Herro · 1991 · 美国芝加哥Deep value / contrarian long-term

In plain words

This report warns against the 'certainty trap'—the habit of seeing things as black or white, like assuming AI will either change everything or be a bubble. The reality is a spectrum. For everyday investors, the key is not to bet on one outcome. Instead, look for companies like Meta or Oracle that can profit from multiple AI scenarios because they have diverse businesses. Also, today's AI investors (like Meta) are cash-rich giants, not dot-com startups relying on stock markets, so they have a longer runway. The lesson: don't trust anyone who claims to know AI's final outcome—stay open-minded.

AI SummaryAI-generated · may contain errors · verify against the original

Oakmark’s Second Quarter 2026 Fixed Income Market Commentary focuses on the "Certainty Trap," pointing out that investors often fall into binary, black-and-white thinking—such as AI either transforming everything or turning into a bubble. The report emphasizes that real-world outcomes tend to fall a

~4 min full read · 5 sections
Deep Analysis

Theme and Background

This chapter centers on the concept of the "Certainty Trap," critiquing investors' tendency to fall into binary, black-and-white thinking amid significant uncertainty. Drawing on President Kennedy's refusal of the binary choice between "invasion or acceptance" during the Cuban Missile Crisis—opting instead for a third option of a naval blockade—the report argues that investors' most dangerous mistakes often begin with accepting others' predefined frameworks for a problem.

Core Thesis

The author contends that AI is the most typical "Certainty Trap" today: the market broadly simplifies it into two outcomes—"changing everything" or "turning into a bubble." However, real-world results fall along a spectrum, not just two possibilities. Harris | Oakmark's core investment principle is: Do not predict the future; instead, seek securities where the current price already compensates for uncertainty. The author explicitly opposes two extremes: completely ignoring AI (assuming it is irrelevant) or going all-in on a single outcome (assuming the result is already known).

Key Arguments and Data

1. Key Difference Between AI and the Internet Bubble: Financing Structure:

  • 25 years ago: Tech companies relied on equity financing; when speculative capital vanished, their business models collapsed.
  • Today: The largest AI investors (e.g., Meta, Oracle) collectively generate over $500 billion in annual operating cash flow, and their balance sheets can support hundreds of billions in additional investment without impairing credit quality.
  • Conclusion: AI has a longer "runway" to prove its economic value, but time does not guarantee success—it only shifts where uncertainty lies.

2. Sources of AI Uncertainty (not a single factor):

  • Energy generation, semiconductor supply, regulation, enterprise adoption, competition, inference economics, etc.
  • Some of the most important future value drivers have not even emerged yet.

3. Comparative Data:

Dimension Internet Bubble Era (circa 2000) Current AI Investment Cycle
Primary Financing Source Equity market (speculative capital) Internal cash flow (>$500 billion annually)
Company Viability Business model collapses when capital disappears Balance sheet supports additional hundreds of billions in investment
Time Horizon Determined by capital markets Determined by the company itself

Companies/Assets Covered

  • Meta (bullish): Daily active users of approximately 3.56 billion (Facebook, Instagram, WhatsApp, Messenger). AI value creation paths are diverse (advertising, engagement, content discovery, productivity) and do not depend on the success of a single LLM. Management (Zuckerberg) has a track record of adapting to technological shifts.
  • Oracle (bullish): Enterprise relationships, proprietary software, and decades of customer data position it to benefit from multiple AI competitive landscapes. The investment thesis does not bet on a single LLM (e.g., OpenAI) but is based on: an investment-grade balance sheet, one of the deepest enterprise software ecosystems globally, mission-critical customer relationships, high recurring cash flow, and the opportunity to help existing customers apply AI to their existing data, applications, and workflows.
  • Investment-grade, single-tenant data center bonds (bullish): Built for hyperscale computing, with long-term contracts tied to a single customer (one of the world's largest tech companies). Investment thesis: does not bet on the success of a single AI application, but on the resilience of the anchor tenant, moderate AI computing demand, and contractual protections. These bonds can earn nearly twice the credit spread of the anchor tenant and benefit from substantial principal amortization before maturity, reducing principal loss risk in adverse scenarios.

Investment Implications

  • Avoid Binary Thinking: Do not go all-in on AI due to its immense potential, nor completely avoid it due to uncertainty. Instead, seek companies that can create value across multiple AI outcomes.
  • Focus on Financing Structure: Current AI investment is dominated by cash-rich giants with far stronger survival capabilities than companies during the internet bubble era. This reduces the risk of systemic collapse, but attention must be paid to whether investment returns are ultimately reasonable.
  • Specific Directions:
  • Prioritize companies with diversified businesses that do not rely on a single AI path (e.g., Meta, Oracle).
  • Consider assets with structural protections (e.g., single-tenant data center bonds), whose risk-return profile is superior to directly holding the anchor tenant's credit bonds.
  • Beware of the "Certainty Trap": Any strategy claiming to have already predicted AI's final outcome is suspect.