Chamath Palihapitiya
· Chamath Palihapitiya
· April 22, 2026 at 18:56
· ⏱ 43 min read
| Read on Substack ↗
Summary
Chamath Palihapitiya argues that investment returns are deeply lagging indicators and that the real work lies in present-moment decisions. He frames AI as a transformational but overhyped technology, identifies physical infrastructure (energy storage, actuation, critical minerals) as the true fulcrum of value, and warns that terminal value is collapsing as AI makes software moats ephemeral. For markets, this means a secular shift away from growth-equity terminal-value narratives and toward physical-asset-backed, short-duration investments, with the US and China in a bipolar race that may ultimately yield peace through mutual unexploitability.
•Groq was licensed by NVIDIA for $20B, but Palihapitiya says that decision was made ten years ago; 2025 was a year of tactical decisions, not outsized new returns.
•AI industry messaging (doomerism, 'building God') is immature and has driven U.S. public favorability to net -20 (lower than Democrats and Iran).
•The AI stack bifurcates into digital AI (abundant text data, runs in data centers) and physical AI (sensory data, must run locally on robots with limited power).
•Fulcrum assets in the AI era are physical: critical minerals, chemical processing, energy storage, and actuation—because factories cannot be repurposed across battery chemistries.
•Only the US and China can build frontier AI due to deep talent, massive compute, near-infinite energy, and capital markets; a bipolar world is forming around two intelligence stacks (US closed-source, China open-weight).
•The Trump Doctrine aims for strategic optionality: economic resilience, geopolitical optionality, fundamental security, and internal population health; bilateralism replaces multilateral frameworks.
•Terminal value (60–80% of tech equity) is collapsing as AI unbundles software moats; Duolingo fell 67.5% from its 2025 peak and Adobe fell 24.6%, exemplifying the shift.
•Private equity bought 1,900 software companies for $440B (2015–2025), and software now makes up ~25% of private credit; default rates on those loans will be the canary for AI disruption.
Palihapitiya explicitly notes that 'silicon companies, the ones building the physical infrastructure AI runs on, went up. Broadcom, SK Hynix, and the semiconductor supply chain' — directly naming Broa
Palihapitiya explicitly notes that 'silicon companies, the ones building the physical infrastructure AI runs on, went up. Broadcom, SK Hynix, and the semiconductor supply chain' — directly naming Broadcom as a beneficiary of the AI infrastructure buildout.
Risk: Capex cyclicality and potential demand normalization if AI investment ebbs; high valuation multiples.
The letter uses Duolingo's 67.5% peak-to-year-end decline as the flagship example of software companies whose products 'can be replicated by a better prompt' — an explicit critique of terminal value v
The letter uses Duolingo's 67.5% peak-to-year-end decline as the flagship example of software companies whose products 'can be replicated by a better prompt' — an explicit critique of terminal value vulnerability in the AI era.
Risk: Duolingo could adapt or integrate AI product features; the market may have overreacted or the example may be backward-looking.
Adobe fell 24.6% from its 2025 peak and is cited alongside Duolingo as a software company suffering from multiple compression as AI makes its product more replaceable.
Adobe fell 24.6% from its 2025 peak and is cited alongside Duolingo as a software company suffering from multiple compression as AI makes its product more replaceable.
Risk: Adobe has a strong enterprise moat with Creative Cloud; AI integration (e.g., Firefly) could offset some disruption.
This newsletter, published April 22, 2026,
features Chamath Palihapitiya
discussing AVGO, DUOL, ADBE.
3 trade ideas extracted by AI with direction and confidence scoring.