Interview: Nihardustinbliss — 2026-05-06

Key Themes

Semiconductor Industry Structure Nihar provided a clear mental model of the semiconductor stack: logic (TSMC near-monopoly, Intel yield-challenged at 14A), memory (3-player DRAM oligopoly: Hynix, Micron, Samsung), advanced packaging, and equipment (ASML, Applied Materials, Tokyo Electron, Lam Research, KLA). Consolidation is structural and one-directional — surviving players are conservative with balance sheets precisely because of the industry’s brutal cyclicality. No new entrants at the frontier: capex requirements are prohibitive.

Chinese Parallel Ecosystem as Entity List Artifact The entire Chinese domestic semiconductor ecosystem exists because of entity list restrictions. Every local Chinese semi-cap maps 1:1 to a Western counterpart it was built to replace. Shell company arbitrage — subsidiaries buy restricted tools, transfer to listed parents — is standard operating procedure, not an edge case. The cat-and-mouse dynamic with regulators is chronic and structural, not episodic.

Equipment Supplier Non-Compliance as Structural Feature The most durable insight: ASML, Tokyo Electron, and Applied Materials actively resist compliance because China represents 20–30% of their revenue. US restrictions unilaterally shift share to Japanese/Dutch competitors. The three-party coordination failure (US, Netherlands, Japan) is chronic — Dutch and Japanese governments have divergent interests and resist being pulled into US export control posture. Chinese companies simply redirect purchasing to the less-restrictive national supplier.

Memory as Most Promising Financialization Target Nihar flagged memory as the most commodity-like layer — spot pricing on certain types, long-term capacity reservation agreements (LTAs) emerging. With only ~15 bilateral relationships (3 manufacturers × 4–5 hyperscalers covering ~80% of demand), a formal exchange may be unnecessary. His reframe — ‘look for commodities not yet commodified rather than end chips’ — is the most actionable product signal.

Investor Information Edge (Nihar’s Approach) Nihar differentiates from channel-checkers (Taiwan real-time contacts) by mapping AI architecture evolution (dense→sparse, training→inference→agents) to 3–5 year chip supply implications. Long-short growth fund, ~$1B AUM on ~$8B platform, transitioning to Gildred Dagden with ~50% AI hardware allocation.

Notable Quotes

  • “The only reason the entire local China ecosystem exists is because of the list.”
  • “The people you need to comply — like ASML — don’t actually want to comply. Because they want to keep selling these tools and make a lot of money.”
  • “You can basically map every single local Chinese semi-cap to one of those four or five global ones.”
  • “Look for commodities not yet commodified rather than end chips.”

Surprises

  • Equipment suppliers don’t passively fail to comply — they actively lobby against restrictions that cut their China revenue. This is a structural enforcement failure, not a monitoring gap. Directly reinforces the Applied Materials $252M settlement context from prior synthesis.
  • The geopolitical trilemma (US/Netherlands/Japan coordination failure) means export controls have a built-in leak — Chinese companies route purchases through whichever national supplier faces the least restriction at a given moment. This is an enduring feature, not a bug to be fixed.
  • The LTA (long-term agreement) trend in memory is relatively new and suggests procurement workflows are actively evolving — potential entry point for tools.

Open Questions

  • What does the procurement workflow actually look like inside a major memory buyer? How far in advance are LTAs negotiated, and who owns that relationship?
  • Can PDF Solutions provide operational supply chain data, or is their focus narrower (yield/process intelligence)?
  • Is there an insurance angle specific to capacity reservation agreements — i.e., what happens when an LTA counterparty can’t deliver?
  • How does Nihar think about the compliance wedge thesis given his view that ASML/Tokyo Electron actively resist restrictions?