The deceleration of Moore's Law and Dennard scaling has imposed an escalating cost ceiling on centralised computation. Frontier model training compute is growing exponentially at 5x per year since 2020 (doubling every 5. 2 months) while the financial cost of training is increasing at 3. 5x per year (doubling every 7 months), far outstripping the ~1. 37x annual improvement in hardware price-performance. This paper presents empirical and architectural evidence that this compute ceiling acts as a key rate constraint on labour displacement (GD), driving a transition to a distributed, three-layer Hybrid Swarm Architecture. We show that this transition suggests a potential inversion of the dominant infrastructure stranding thesis: the massive 700B-725B aggregate hyperscaler capital buildout projected for 2026 may be better interpreted not as a stranded asset bubble, but rather as the productive centralised hub of a coexisting human-AI economy (AGI-C). We present verified May 2026 data on hyperscaler capex, Nvidia Blackwell standard deployments (81. 6B record Q1 FY2027 revenue), illustrative indicators of enterprise workload repatriation (including a Cloudian-commissioned survey of 203 IT decision-makers showing a 93 per cent repatriation rate), and near-future 1. 58-bit ternary on-device co-design. This formalises the Economic Forcing Function (EFF) as the physical and economic rate-limiter of labour market transition speeds, suggesting how the depreciation and capital pressures identified by bubble theorists act as structural mechanisms driving a metamorphic evolution that preserves and sustains infrastructure value. This Zenodo record includes the main preprint PDF and a companion formula/theory guide documenting the mathematical structure, parameter definitions, sensitivity framing, and theoretical boundaries of the Economic Forcing Function framework.
Vito Henjoto (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: