Discover The Residual Value of Compute Hardware

A: The maximum physical operating life is a configurable parameter — currently estimated at around 8 years based on historical end-of-life data disclosed by hyperscalers (e.g., AWS, GCP) for previous GPU generations like the P100 and V100. However, the economic life is handled dynamically by our Utilization decline curve and the Forward Curve's pricing. If an architecture becomes obsolete faster than its physical lifespan, the projected utilization and rental rates will crater within the model, driving the residual value to near-zero well before the maximum parameter is reached.

A: Utilization follows a continuous exponential decay of the form k·e^(−rt), where k is current utilization, r is the decay rate, and t is time from the estimation date — not a linear drop or a step function. Generational shocks such as a Blackwell release are captured through the market side of the model: the Forward Curve's rental pricing reprices immediately, while the decay curve represents the gradual competitive erosion as newer, faster architectures absorb workloads.

A: The current model uses a blended global average operating-cost ratio covering electricity, rack space, and labor — a standardized macroeconomic baseline for valuation rather than a site-level cost model. Region-specific cost modifiers may be introduced as market data granularity improves.

A: The model holds the last observed Forward Curve rate and applies a compounding tail decay anchored to the month-36 state, with a decay rate set well above the initial curve's to reflect accelerated obsolescence of aging architectures. Because value at that point is already heavily discounted, the tail assumption adds little sensitivity to the final present value.

A: Yes. The residual value recalculates dynamically as the underlying GPU Forward Curve inputs change, so it reflects the true, real-time mark-to-market value of the asset rather than a static depreciation schedule fixed at purchase.