Vicor widens AI server power patent fight, seeking new US import ban on Foxconn, Quanta and Delta

Vicor's second import ban attempt this year is not a routine IP dispute—it is a structural chokepoint play on AI compute infrastructure. The patent portfolio sits precisely at the transition from board-level multiphase DC-DC to modular brick architectures, the exact migration path every hyperscaler is executing now. Locking Foxconn, Quanta (ODMs) and Delta (Tier-1 power) simultaneously means the power delivery chain for NVIDIA GB200/GB300 clusters in North America faces a direct rupture. The compliance shock is not a line-item cost: ODMs must re-qualify power modules within 60-90 days, and Delta's production scheduling will cascade into Q3-Q4 AI server shipment slippage. More critically, when import bans become a normalized tool, supply-chain compliance redundancy shifts from optional to mandatory—hyperscalers will need two to three independent power paths in parallel. Competitively, Infineon and MPS will absorb the displaced volume, but AI power lead times could stretch past 20 weeks near term. The 12-24 month trajectory: AI power delivery is converging toward on-package power delivery, which is precisely where Vicor's patent coverage is thinnest. The long-tail effect of these bans may paradoxically accelerate the industry past Vicor's architecture, compressing the adoption curve for next-gen power topologies.