Chip and Hardware Competition
Key Questions
What new NVIDIA and power semiconductor technologies are advancing AI hardware?
NVIDIA's Vera Rubin is in full production, supported by Navitas' 97.5% efficient 800V-to-6V PDB, Advanced Energy's 98.2% efficient converters, and EPC's 7th-gen GaN FET. Lotus Microsystems' 96% efficient vStrata vertical delivery and Navitas' GaN/SiC solutions for NVIDIA are also key.
How are alternative chip architectures challenging GPU dominance?
Foxconn/Intel are partnering on AI rack systems with Xeon 6, Supermicro's Arm partnership claims 2x performance per rack, and Marvell reports 27% revenue growth with custom silicon XPUs on track for $10B. Qualcomm's ByteDance ASIC targets energy-efficient NPUs by 2026.
What supply chain pressures are affecting AI chip and hardware scaling?
Power semiconductor shortages, memory supply squeezes, and emerging component bottlenecks are intensifying, prompting AMD's acquisition of MEXT for memory tiering. Enphase is pivoting to solid-state transformers for an 11 GW US market opportunity.
How is efficiency becoming central to IT hardware planning?
HPE confirms efficiency as core to planning, with shifts toward tokens per dollar metrics and 48V ecosystems from Allegro. Solid-state transformers have raised $280M and seen acquisitions, validating power conversion tech.
What partnerships are accelerating AI rack and interconnect solutions?
Foxconn/Intel and Supermicro Arm deals target rack-level performance, while Marvell's 102.4 Tbps switch and 1.6T interconnect catalysts support scaling. Google-SpaceX arrangements underscore ongoing GPU scarcity pressures.
NVIDIA Vera Rubin production. Navitas 800V-to-6V PDB with 97.5% efficiency, stock surge. Advanced Energy 98.2% efficient 800V DC converters. EPC 7th-gen GaN FET for 48V-to-6V converters. Lotus Microsystems introduces 96% efficient vStrata vertical power delivery. Foxconn/Intel partnership on AI rack systems with Xeon 6+ challenges GPU dominance. Supermicro Arm partnership claims 2x performance per rack. Marvell 102.4 Tbps switch. Power semiconductor shortages. Shift to tokens per dollar. Marvell Q1 earnings show 27% revenue growth, custom silicon XPU on track for $10B, 1.6T interconnect catalysts. Qualcomm ByteDance ASIC for energy-efficient NPUs. Google-SpaceX deal underscores GPU scarcity. Memory supply squeeze emerging as new component bottleneck; AMD acquires MEXT for memory tiering. Enphase Energy pivots to data center solid-state transformers (SST), targeting 11 GW US addressable market by 2031. Navitas highlights GaN and SiC for NVIDIA. Allegro expands 48V ecosystem. Solid state transformers raise $280M and see acquisition. HPE confirms efficiency is core IT planning. Qualcomm Dragonfly Arm processor targets efficient AI inference, challenging GPU dominance. Corning emerges as key AI infrastructure bottleneck via massive fiber deals. NVIDIA introduces fully liquid-cooled system eliminating fans and nearly zeroing water use.