Yield Curve
Yield Curve (Layer 3)
Layer 3 is the intelligence layer. It sits on top of the futures market and turns thousands of individual locked-rate contracts into a single, continuous picture: a yield curve for AI compute — what the market believes inference will cost at every point in the future.
This is data nobody else has, because nobody else has the futures market underneath it. You own the curve because you built the market.
The intelligence layer
On Layer 2, every action a user takes produces a price point. Someone locks DeepSeek at $0.14/M until September. Someone else locks it at $0.16/M until December. Another at $0.19/M until March. Each contract is one coordinate.
Individually those are just trades. Plotted together — price per million tokens against time to expiry — they form a term structure: the same yield curve that defines every mature commodity and rates market, now for machine compute.
What the yield curve is
The yield curve maps the price of compute across a series of future expiry dates. On Auton, each model gets its own curve:
- X-axis — time to expiry (1 week, 1 month, 3 months, 6 months, and beyond).
- Y-axis — locked rate in $/M tokens implied by the futures at that expiry.
- One curve per model — DeepSeek, Llama, Qwen, GPT, Gemini, Claude — so you can compare term structures across providers.
How the curve forms
The curve is built directly from real market activity, not a model's list price:
- Each open and settled futures contract contributes a (expiry, rate) point for its model.
- Points are grouped by model and sorted along the expiry axis to form the term structure.
- The front of the curve is anchored to the live spot rate (the blended price the gateway pays today), so the near end always reflects reality.
- As more contracts open across more expiries, the curve gets denser and more accurate — it improves automatically with volume.
Reading the curve
The shape of the curve is itself the signal:
| Shape | What it means | Signal |
|---|---|---|
| Contango (sloping up) | Future compute is priced higher than today | Market expects scarcity, demand growth, or a model launch |
| Backwardation (sloping down) | Future compute is priced lower than today | Market expects new supply / efficiency to drive prices down |
| Local spike | A bump around a specific expiry | An anticipated event — e.g. a major model drop that period |
Architecture
The curve is a read-only data layer derived from Layer 2. Nothing new is traded here — it indexes the market that already exists and serves it back as financial-grade analytics:
Who uses it
- Builders — decide when to lock in futures. If the curve is in contango, lock now; if it's in backwardation, wait.
- Traders — spot mispricings between expiries or between the curve and live spot, and express them with leveraged positions.
- Analysts — track where the compute market is heading across models, the way rates desks read a yield curve.
Why it matters for $AUTO
This is the moment Auton stops looking like a crypto product and starts looking like a Bloomberg Terminal for machine resources. The yield curve is pure data infrastructure — and it is defensible precisely because it can only exist on top of a real futures market:
- The curve is a data moat: no market underneath means no curve, and Auton owns the market.
- More usage on Layer 2 → a richer curve → more builders, traders, and analysts → more usage. The intelligence layer compounds demand for the underlying market that drives $AUTO.
- It positions $AUTO as the settlement and data layer of the machine-resource economy, not just another token.
Status
The yield curve is the planned Layer 3. The data that feeds it — locked rates, expiries, and on-chain settlements — is already being produced by the live Layer 2 futures market today. This page describes the design; the terminal dashboard is the next build, and it grows more valuable with every contract opened in the meantime.