Interview: 390 Jonathan Bliss Dustin — 2026-05-08
Key Themes
Compliance wedge is dead (for now). The team’s original go-to-market thesis — using export compliance as the data extraction wedge — has hit a wall. The core problem: companies in the semiconductor supply chain are incentivized toward opacity, not transparency. If goods are flowing through Singapore to China, the seller prefers not to know. This is a fundamental structural problem, not a product gap.
New thesis: Glencore of memory chips. The conversation pivoted hard toward inventory management and commodity trading as the new direction. The post-2020 chip shortage cost the auto industry ~$200B in lost revenue — not from price spikes but from pure unavailability. Companies now want buffer inventory but don’t want to own or manage it. The Glencore oil trading analogy was the intellectual engine of the session: Glencore profits by being a physical option holder at scale, owning storage infrastructure, seeing market flows, and capturing volatility spreads. Could a similar model work for commoditized chips (memory, trailing edge)?
Arrow/Avnet as the foil. Arrow ($30B revenue, $500M net income, $9B market cap) is the existing distributor model — and its thin margins reveal how little defensible value distributors currently capture. The hypothesis is that the Glencore model requires something Arrow doesn’t have: either storage moat, data asymmetry, or financing infrastructure.
Supply chain anatomy clarified. Significant portion of the call was spent mapping the semiconductor value chain: TSMC/Samsung (foundries) → fabless designers (Nvidia, Elmos, ARM) → tier one suppliers → OEMs (Toyota, GM). Tesla’s supply chain position (does it bypass tier ones?) flagged as an open question.
Notable Quotes
- “I’m selling to a firm in Singapore — who knows what they’re doing with it.” (On why compliance is not a real pain point)
- “They don’t want to manage the inventory themselves.”
- “Glencore makes billions because they’re physical option holders at industrial scale.”
- “The automobile industry loses $200 billion of revenue — not because the chips went up by that much, but because they couldn’t get the chips.”
Surprises
- The compliance pivot failure was acknowledged openly and quickly — the team has internalized it and is already moving. That’s a healthy signal.
- The Glencore analogy emerged organically and generated real energy. It’s a strong mental model but the team immediately identified the key disanalogies (no storage moat for chips, obsolescence risk, custom vs. commodity).
- Jonathan flagged that memory chips may be more commoditized than logic chips — this is the crux of whether the Glencore model is viable and hasn’t been validated yet.
Open Questions
- How commoditized are memory chips vs. logic chips, really? Is there a DRAM-like spot market already?
- What are the actual storage and maintenance requirements for semiconductors? Do chips require “rebaking” or other active maintenance that creates real barriers?
- Why do Arrow and Avnet have such low margins at $30B+ revenue? What does that reveal about defensibility?
- Does Tesla buy from tier ones or go direct to chip designers?
- What would it actually take to be a physical option holder in chips — storage, capital, information edge?