Jeremy Raper: Cycles, AI, and Activism

Jeremy Raper opens the podcast with personal updates about his move to Australia and new family situation. The conversation pivots to a detailed analysis of the AI semiconductor cycle. Raper argues that despite claims these are structurally different post-cyclical businesses, memory companies like Micron and SanDisk will face a significant earnings cliff as peak pandemic-era profitability normalizes. He uses the example of SanDisk, which generated $900M EBITDA in 2024, spiked to $10B, with consensus expecting $35-40B, only to likely revert to $10-15B normalized levels. This presents a massive valuation risk for stocks trading at 10x current-year free cash flow.

Raper explains that cyclical commodities and commodity-adjacent businesses have long histories of collapsing when supernormal earnings normalize—citing refiners, coal, and thermal pricing as precedents. Long-term supply agreements provide less downside protection than proponents believe, as powerful oligopsonist customers (Meta, Google, Apple) will pressure prices down when spot markets soften. He emphasizes trading cyclicals requires discrete timing assumptions about peak earnings duration and a glide path to normalized profitability.

On artificial intelligence's broader impacts, Raper shares his practical experience using Claude for legal document analysis and special situations work. He estimates replacing $350,000 in legal fees with AI-generated documents and legal analysis that costs near-zero, dramatically reducing friction costs for activism and taking action on opportunities. This democratizes access to sophisticated legal and analytical capabilities previously available only to well-capitalized investors. He hard-codes confirmation bias safeguards into AI prompts, instructing it to push back aggressively.

Raper theorizes that AI's global accessibility will collapse valuation disparities between developed and emerging markets. Since an entrepreneur in Bangladesh can access identical AI tools as Silicon Valley investors, geographic arbitrage on expertise diminishes. He predicts this structural shift may reverse the decades-long tendency for developed markets to outperform, particularly for value stocks globally that were previously locked up due to high friction costs of extracting value.

Regarding his personal investment process, Raper works approximately 4 hours per day on investment activities despite the cultural pressure to be constantly engaged. He emphasizes that like an Olympic athlete, extensive practice ("reps") creates capacity for explosive short-term decision-making during rare windows. He recounts Bill Ackman's Washington Mutual holding company investment during the financial crisis as an example requiring a two-hour analysis for a $500M decision.

Raper discusses his failed Australian special situation involving a 30% blockholder preventing value extraction despite regulatory findings of bad faith. He learned that certain ownership thresholds make opposition nearly insurmountable and that sizing positions to justify campaign costs is economically irrational. He's now focusing on Australian small-cap opportunities where he identifies structural gaps in professional coverage due to the market's pension-driven large-cap focus and retail gambling orientation toward mining stocks.

Finally, Raper walks through European Lithium (EUR), an Australian-listed cash box holding 45.5% of CRML (Critical Minerals, NASDAQ-listed), which owns the Tanbreeze rare earth deposit in Greenland. A proposed merger ratio of 58 cents per EUR share when CRML was $12 has deteriorated to 26 cents as CRML traded to $5. Since EUR has 18 cents per share of net cash and the CRML stake implies 40 cents value, Raper expects the deal will require substantial reconsideration. He argues the CRML leadership needs this deal far more than EUR shareholders do, and independent fairness opinions limit how much the consideration can deteriorate, making the asymmetry actionable.