What AI Is Made Of

Run 04 equipment mass balance

Run 04 applies independently constructed equipment recipes and proxies to the fixed Run 03 plant inventory. It reports partial equipment stocks in metric tonnes per 1 GW of commissioned IT capacity. Low, base, and high use joint input constructions with fixed equipment counts. Other feasible designs could fall outside their range. Transfer from the source products to the hypothetical campus depends on the stated proxy assumptions.

The cumulative modeled base contains about 16.92 kt of copper-grade stock, 2.48 kt of aluminum stock and named alloys, 44.28 kt of disjoint ferrous stock, and 24.0 t of elemental lithium nested within its identified host compounds. The largest incremental historical differences for copper, carbon steel, and aggregate ferrous are in B2 facility equipment. The generated findings show the engineering ranges, leading equipment contributors, replacement effects, and comparison by boundary.

The comparison CSV includes each endpoint, its display-rounding tolerance, and missing or unreported estimates. The archived material-specific envelopes and the modeled joint input cases have different uncertainty semantics.

The coverage register identifies modeled scopes, residual mass budgets, and gaps. High-impact missing inputs include installed conductor route lengths and sections, equipment-specific transformer drawings and material masses, and selected battery and cooling equipment bills of materials. Civil works and unallocated offsite infrastructure remain outside the equipment boundary.

Run python scripts/reproduce_run04.py from the repository root to regenerate the outputs offline. Use the scenario methodology and historical comparison method (repository-only reference) to interpret the results. The manifest (repository-only reference) records input and output hashes. The preservation audit (repository-only reference) compares frozen Run 03 bytes with their original Git blob IDs.

Source details

Run 04 research. Equipment material estimates with documented assumptions and scope. Web prose uses current site terminology; downloadable records preserve original wording. Hashes identify the downloadable records.

Source: RUN_04_FINDINGS.md
SHA-256: 005cf65e5a0bbaaaddaaba25cc244433f7b7ff98e713600ebb82d7104b6aea15
Download research file

Optional: tell Sid you’re reading
The document is open; sharing your name/email is optional.