Key points are not available for this paper at this time.
Hyperscale data center construction has become one of the largest concentrated capital flows in the global built-environment sector, with global hyperscale capital expenditure exceeding USD 230 billion in 2024 and the leading hyperscalers committing more than USD 325 billion for 2025. Generative artificial intelligence workloads have driven a structural shift in mechanical and electrical design, increased delivery-pace demands, and concentrated supply-chain pressure on a small number of specialist general contractors. This paper develops and tests a microfoundational model of dynamic capabilities for hyperscale data center project delivery using fuzzy-set Qualitative Comparative Analysis (fsQCA) on a sample of N = 18 primarily North American hyperscale-active general contractors. Capabilities are formalized as fuzzy sets X ⊆ Ω with membership functions μX: Ω → 0, 1 over the case space Ω; necessity and sufficiency consistency are computed using the canonical formulas ConsN(X→Y) = Σi min(μX(i), μY(i))/Σi μY(i) and ConsS(C→Y) = Σi min(μC(i), μY(i))/Σi μC(i), and necessity is interpreted as the fuzzy-set containment PDPhigh ⊆ X. Capability scores are coded from secondary public sources—ENR rankings, SEC filings, trade press, and company disclosures—rather than from primary survey or interview data. Three findings emerge: process microfoundations and seizing capability are necessary conditions for both project delivery performance and competitive advantage (ConsN ≥ 0.97); the dominant causal recipe for high project delivery performance is the conjunction SEN ∩ SEI ∩ TRA ∩ ED with consistency 1.000 and coverage 0.771; the dominant recipe for high sensing capability is the conjunction IND ∩ PRO ∩ STR ∩ INT with consistency 1.000 and coverage 0.799. The findings are positioned against prior dynamic-capabilities studies and microfoundations theory, confirming convergence on conjunctive bundle structure while extending literature into a previously unstudied AEC segment with sharper containment relations.
Shafaghat et al. (Fri,) studied this question.