DeepSeek — AI supply-chain exposure
The model reads DeepSeek primarily as a producer in Models. Its largest modeled sensitivity is a shock at GPU (constraint β 10).
The structural read · model-generated
The model reads DeepSeek primarily as a producer in Models. Its largest modeled sensitivity is a shock at GPU (constraint β 10).
In depth · editorial + model · written 2026-07-13
DeepSeek is a Hangzhou model lab that builds open-weights reasoning models. Its V3 and R1 releases mattered less for raw capability than for cost: they reset global expectations of how cheaply a strong reasoning model could be trained and served. That sits it on the model layer as a producer, pretraining its own foundations and specialising in the step-by-step reasoning that has become the frontier's proving ground.
Its structural hook is cost-per-token itself. By showing a capable model can be produced far below assumed budgets, it exerts downward pressure on the whole layer's pricing, and open weights let others build on that base. Its exposure is almost purely at the model level, spanning pretraining, reasoning and the economics of inference, which is why the model centres it despite its light physical footprint.
Where it's exposed
Chain footprint by layer
How it participates
Every part DeepSeek touches
Geographic concentration
Frequently asked
What is DeepSeek's role in the AI supply chain?
The model reads DeepSeek primarily as a producer in Models. Its largest modeled sensitivity is a shock at GPU (constraint β 10).
Which parts of the AI value chain is DeepSeek exposed to?
DeepSeek is mapped to 3 parts of the AI value chain, most strongly Reasoning models, Pretraining, Cost per token. It sits primarily in the Models layer as a producer.
Does DeepSeek own an AI bottleneck?
Not in the current model — DeepSeek is exposed to constrained parts but sits downstream of them rather than producing them.
What is DeepSeek's biggest AI supply-chain risk?
Its largest modeled sensitivity is a shock at GPU (constraint β 10). 4 nodes depend on it; pressure 78/100
Who are DeepSeek's closest peers by AI-chain position?
By shared chain dependencies: xAI, Safe Superintelligence, High-Flyer Quant, Moonshot AI.
Go live on DeepSeek
- The interactive dependency graph and full company Nexus
- The analyst bull / bear thesis and valuation lens
- Live signals, today’s movers and the read-through
- Track it in your Portfolio Cockpit — positions, P&L, valuation, thesis
model v0.7.0 · research, not advice
Chain analytics are illustrative, order-of-magnitude estimates from our model of the AI value chain — not investment advice.