Why the AI Trade Is No Longer a Single Story: Key Takeaways
The recent market downturn, particularly last Thursday, saw significant drops in the NASDAQ (6%) and S&P 500 (4%), with much of the blame attributed to the "AI trade." However, the concept of an undifferentiated "AI trade" is becoming increasingly problematic as the AI industry matures. While in its early stages, most AI companies tended to move in unison, this is no longer the case. It's time to differentiate between the various players in the AI ecosystem.
Deconstructing the AI Ecosystem
The AI business can be broken down into four distinct groups, each with unique characteristics and financial dynamics:
AI Suppliers: These companies provide the foundational ingredients for AI architecture.
- Examples: Nvidia (chips), power companies, electrical equipment companies, real estate companies (selling land for data centers).
- Role: They benefit first and most directly from increased AI investment, as their products are purchased upfront.
- Financial Impact: Revenues and earnings pick up almost immediately.
AI Builders: These are the companies that construct the AI infrastructure, primarily data centers.
- Examples: Hyperscalers like Alphabet, Meta, Microsoft, Amazon, Oracle, CoreWeave.
- Role: They acquire chips and other components from suppliers to build large-scale data centers.
- Financial Impact: Their investments are treated as capital expenditure (capex), which negatively impacts current earnings and cash flows but is amortized over time. They aim to lease out this infrastructure.
AI Product and Services Companies: These businesses utilize the AI infrastructure to develop and offer AI agents and products.
- Examples: OpenAI, Anthropic (LLMs), Palantir, Salesforce (integrating AI into existing products).
- Role: They lease compute power from AI builders to create and sell AI-driven solutions to end-users.
- Financial Impact: Lease agreements create significant compute costs, either as operating expenses or financial expenses. Revenues are generated from selling AI products. Many are pre-revenue or have small revenues with high R&D and employee costs in early phases.
AI Users: These are the individuals or enterprises that purchase and utilize AI products and services.
- Role: They adopt AI to improve efficiency, reduce costs, increase revenues, or simply to keep pace with competitors.
- Financial Impact: Spending on AI products is recorded as an expense, with the expectation of benefits outweighing the cost.
The Evolving Phases of AI Development
The AI industry progresses through distinct phases, with the financial performance and interdependencies of these groups changing over time:
Hope and Hype Phase (e.g., 2023-2024):
- Suppliers: See immediate benefits, with revenues and earnings rising.
- Builders: Invest heavily in capex, showing negative cash flows and no immediate returns.
- Product Companies: Nascent, often pre-revenue, with high R&D costs.
- Users: Primarily test users and early adopters.
- Overall: Promise over product; all groups benefit from the perception of a large AI market.
Investment Phase (e.g., 2024-2025):
- Suppliers: Businesses are established, but growth rates may start to moderate as they scale.
- Builders: Begin to lease out infrastructure, generating initial, albeit small, revenue streams. Returns on invested capital are still low.
- Product Companies: Start introducing products, experiencing high growth rates due to small initial scale.
- Users: Early adopters and experimenters in specific industries.
Business Building Phase (Current):
- Suppliers: Models are well-established; focus shifts to continued scaling.
- Builders: Leasing out significant portions of architecture, with clearer visibility on potential returns on capital.
- Product Companies: Jockeying for market position, locking in compute power through long-term contracts to differentiate.
- Users: Targeted with sales pitches emphasizing the transformative potential of AI.
Interconnections and Divergent Interests
While all groups ultimately depend on a growing and monetized end-market for AI products, their interests are not always aligned:
- Interdependence: If the end-market for AI products shrinks, product companies are hit first, leading them to reduce compute power agreements, which impacts builders, who then reduce orders from suppliers.
- Circular Financing: Suppliers like Nvidia have an interest in the health of builders (their customers). They might offer support or financing to ensure builders continue to purchase their chips.
- Competing Interests:
- Builders vs. Shareholders: Shareholders of hyperscalers might worry about excessive capex spending on data centers that may not be fully utilized.
- Suppliers vs. Builders: Suppliers benefit from builders investing heavily, even if those investments prove to be mistakes for the builders.
- Builders vs. Product Companies: Long-term compute power contracts are beneficial for builders (locking in revenue, reducing risk) but can be restrictive and create financial obligations for product companies (LLMs) before they are ready.
This divergence of interests highlights why a monolithic "AI trade" is no longer an accurate representation of the market.
The "Big Market Delusion" and Future Correction
The rapid growth and excitement around AI have led to a "big market delusion." This phenomenon occurs when:
- A large, promising market (like AI) attracts overconfident entrepreneurs.
- These entrepreneurs secure funding from overconfident venture capitalists.
- A cluster of overconfident individuals collectively overestimate potential revenues and company valuations.
The inevitable outcome is that the market is not large enough to sustain all the companies that enter it. This leads to:
- Overpricing: Companies are collectively overvalued in the early stages.
- Cleanup Phase/Correction: Eventually, the market will recognize this overreach, leading to a correction. This is a feature, not a bug, of disruptive change.
Implications of a Correction:
- Uneven Pain: Not all AI companies will suffer equally; some may emerge as superstars.
- Timing Uncertainty: Predicting the exact timing or catalyst for a correction is difficult. It could be a major event or a series of smaller disappointments.
- Behavioral Phenomenon: Since the delusion is driven by overconfidence, fundamental analysis alone may not predict the correction. Technical indicators might be more useful for those skilled in charting.
Investment Strategy During a Correction:
- For Existing Holders: Consider selling portions of holdings gradually to manage risk and avoid an "all or nothing" decision.
- For New Investors: A correction presents an opportunity to buy. Identify well-liked companies now (even if overpriced) and set limit orders at desired valuation points to capitalize on market panic.
Anthropic's IPO and Valuation Challenges
The impending IPO of Anthropic, scheduled for mid-November, faces challenges due to the shifting market sentiment.
- IPO Pricing vs. Valuation: IPOs are priced based on demand and supply, reflecting market mood and momentum, rather than intrinsic valuation. A darker market mood will pressure bankers to lower the offering price.
- Valuation Sensitivity: Valuing young growth companies like Anthropic is highly sensitive to assumptions. Recent political and regulatory developments (e.g., difficulties in building data centers in Texas, increased regulatory hostility towards AI) can significantly impact valuation.
- Example: Adjusting the "sales to capital" ratio (revenue generated per dollar of capital invested) from $3 to $2.5 (implying more investment needed for growth) and increasing the "drag" on margins (due to regulatory pressure) from 5% to 7.5% can drop an estimated equity value from $1.45 trillion to $1.25 trillion.
Conclusion
The "AI trade" as a singular concept is outdated. Investors must differentiate between AI suppliers, builders, product companies, and users, understanding their distinct financial characteristics and interdependencies. While a "big market delusion" suggests an eventual correction, this is a natural part of disruptive innovation. Astute investors will prepare for this by managing existing holdings and identifying future buying opportunities, always asking critical questions rather than succumbing to extreme optimism or pessimism.
Takeaways
- The AI market can no longer be treated as a monolithic "AI trade" because the ecosystem consists of distinct groups—suppliers, builders, product companies, and users—each with different financial dynamics.
- In the current business‑building phase, AI suppliers see stable scaling, builders generate revenue from leasing infrastructure, product firms compete for compute contracts, and users are targeted with transformation pitches.
- Divergent interests among these groups mean that a downturn hits product companies first, which then cascades to builders and suppliers, explaining why stock moves are now uneven.
- The "big market delusion" describes how overconfidence leads to inflated valuations and an inevitable correction, where only the strongest AI firms will survive and others will be pruned.
- Investors should manage risk by gradually trimming existing positions during a correction and use the pull‑back to buy well‑liked AI stocks at lower valuations, while monitoring technical signals for timing.
Frequently Asked Questions
What does the "big market delusion" refer to in the AI context?
The "big market delusion" is the phenomenon where overconfident entrepreneurs and venture capitalists overestimate the size and revenue potential of the AI market, leading to inflated valuations for many startups. This collective optimism creates a bubble that eventually corrects when the actual market cannot sustain all entrants.
How do the interests of AI builders clash with those of product companies?
AI builders seek long‑term compute contracts that lock in revenue and reduce risk, while product companies often need flexibility because they may not yet have mature products. These contracts can become financial burdens for product firms, creating tension between builders’ desire for stable cash flow and product firms’ need for agility.
Who is Aswath Damodaran on YouTube?
Aswath Damodaran is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.
Does this page include the full transcript of the video?
Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.
(6%) and S&P 500 (4%), with much of the blame attributed to the "AI trade." However, the concept of an undifferentiated "AI trade" is becoming increasingly problematic as the AI industry matures. While in its early stages, most AI companies tended to move in unison, this is no longer the case. It's time to differentiate between the various players in the AI ecosystem. ## Deconstructing the AI Ecosystem The AI business can be broken down into four distinct groups, each with unique characteristics and financial dynamics: 1. **AI Suppliers:** These companies provide the foundational ingredients for AI architecture. * **Examples:** Nvidi
(chips), power companies, electrical equipment companies, real estate companies (selling land for data centers). * Role: They benefit first and most directly from increased AI investment, as their products are purchased upfront. * Financial Impact: Revenues and earnings pick up almost immediately.
Helpful resources related to this video
If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.
Links may be affiliate links. We only include resources that are genuinely relevant to the topic.