Back to Flash News

Investor Geoffrey Woo: The upstream hardware supply chain remains the same sources, only more intermediaries and cross-border transactions

Venture capitalist Geoffrey Woo pointed out that once mineral sourcing is laid bare, the "diversified supply chain" touted by hardware venture capitalists will seem untenable: the upstream still consists of the same batch of mineral sources, only with more intermediaries and increased cross-border transactions.

He is not referring to the financial reports of a specific startup, but rather the concentration of processing of key minerals. According to data from the International Energy Agency, the combined share of the top three refining countries for energy minerals has risen from about 82% in 2020 to about 86% in 2024, with about 90% of new supply coming from a single leading supplier: nickel from Indonesia, cobalt, graphite, and rare earths from China. The mining sector is also tightening, with the share of the top three mining countries increasing from about 73% to about 77%. Lithium is one of the few minerals where new entrants like Argentina and Zimbabwe are present, but refining remains highly concentrated.

The midstream is narrower than the mines. About 90% of rare earth refining and separation occurs in China, with approximately 94% of permanent magnets produced there; under Goldman Sachs' metrics, about 85% to 90% of rare earths from mine to metal refining, and about 92% of rare earth magnets. Lithium can be mined in hard rock in Australia and brine in Chile, with almost all Australian spodumene concentrate shipped to China for processing. Cobalt from the Congo and nickel from Indonesia also reflect a "mined dispersion, concentrated processing" model. Japan has reduced its reliance on rare earth imports from China from over 90% during the 2010 dispute to about 60-70% by relying on Lynas' Mount Weld in Western Australia and the separation plant in Kuantan, Malaysia, indicating that the real structural change comes from the separation plants, not just signing another trade contract.

U.S. policy is also focused on acquiring processing rights rather than narratives. The Department of Defense invested about $400 million for a 15% stake in MP Materials to promote domestic rare earth magnets; it also holds about 10% of Trilogy Metals, targeting Alaskan mining areas. Recycling is being treated as a third source of minerals, with companies like Ascend Elements and Redwood Materials creating closed loops for batteries and electronic waste, as new mines often take nearly a decade from exploration to production.

Hardware venture capital labels combinations as "diversified" based on country names, border counts, and the number of suppliers, while sourcing is accounted for by molecules and smelting plants. Adding another layer of traders or customs declarations does not change who owns the refining furnace. Licensing, separation chemistry, waste streams, capital expenditures, off-take agreements, and geopolitics are the hard constraints on expansion.

In market mechanisms, what is being repriced are the processing capacities, magnet factories, and recycling facilities that can prove mineral sources and separation paths; hardware projects that only change the delivery location without altering upstream chemistry are being discounted. Capital is flowing from "multi-country supplier lists" to assets capable of separation, sintering, and closed-loop recycling. Beneficiaries include countries that control refining and magnets, as well as mining companies that have secured government equity or off-take agreements; those under pressure are hardware combinations that treat border counts as risk hedges, and mines that still sell concentrates to the same batch of smelting plants.

Source: Public information

ABAB AI Insight

Geoffrey Woo does not come from a mining background. With a computer science background from Stanford, his first company Glassmap was sold to Groupon, followed by Ketone-IQ, and later co-founding Anti Fund with a partner from a boxing background, managing over $180 million, with investments including OpenAI, Anduril, and Ramp. He repeatedly mocks the rare earth discussion in a software industry tone: when engineers ask why production cannot be expanded, the mining side cites licensing, separation, waste, off-take agreements, and nine-year cycles, while the software industry immediately interprets it as "just create a market." This statement reflects the same misalignment: hardware funds count suppliers in a software-like manner, while the mining side accounts by molecules.

Capital is changing its channels. The U.S. buying equity in MP Materials directly injects finances into magnet factories; Japan, in 2011, provided about $250 million in loans and equity to Lynas through Jogmec and Sojitz, locking in non-Chinese pathways. Mining capital like Wyloo is acquiring nickel assets, taking majority stakes in the Yangibana rare earth project, and holding shares in Neo Performance Materials, focusing on mining plus separation plus materials, rather than just finding another trader. Recycling is being termed the "fourth mine" because the time and community costs of greenfield mines do not align with the expansion timelines of chips and electric vehicles.

Specific comparisons are necessary. After China restricted rare earth exports in 2010, Japan did not rely on signing three more traders to get through, but instead made Mount Weld and the Kuantan plant into a closed loop. Australia is the top lithium miner, yet almost all concentrates still go to China, indicating that mining diversity can coexist with processing monopolies. Companies like Waymo are competing on fleets, while here the competition is over furnaces. The industry is in a phase of "verbal dispersion, actual re-concentration": the mining map is becoming colorful, while the refining map is turning redder.

Structural judgments belong to the transfer of pricing power. The mechanism is: the constraints on chips, motors, and batteries do not lie in the number of assembly plants, but in separation chemistry and waste processing; whoever controls the midstream can simultaneously raise prices or cut supplies for upstream ores and downstream hardware. Adding another layer of intermediaries only increases freight and compliance costs, without increasing alternative molecular sources. Therefore, once sourcing is opened up, the narrative of dispersion will collapse, not because funds are fraudulent, but because they are counting passport stamps instead of refining capacities.

ABAB News · Cognitive Laws

  1. More intermediaries do not equal more sources

  2. Dispersed mines cannot manage concentrated furnaces

  3. Once sourcing is opened, narratives will collapse before inventories