NVIDIA’s Q1 2026 data center revenue surged 265% to $22.6 billion, while AMD’s data center revenue fell 18% to $1.2 billion. Microsoft Azure and Meta accounted for 42% of NVIDIA’s AI chip sales — up from 31% a year earlier. TSMC’s CoWoS packaging utilization hit 98% in March 2026. Intel’s AI chip revenue rose 12% to $1.8 billion, and Broadcom’s AI networking chip revenue jumped 67% to $5.4 billion. Global AI chip capex totaled $48.3 billion in 2025, up 89% from $25.6 billion in 2024.
According to www.bloomberg.com, Q1 2026 earnings reports reveal a sharp divergence among semiconductor suppliers serving the artificial intelligence infrastructure market — with NVIDIA’s data center revenue surging 265% year-on-year to $22.6 billion, while Advanced Micro Devices (AMD) reported an 18% decline in its data center segment revenue compared to Q1 2025.
Revenue Performance by Company
NVIDIA’s data center revenue reached $22.6 billion in Q1 2026, up from $6.3 billion in Q1 2025. The company attributed this growth to sustained demand for its Hopper and Blackwell architecture GPUs, particularly from cloud service providers deploying large language model training clusters. In contrast, AMD’s data center revenue fell to $1.2 billion in Q1 2026 — down from $1.47 billion in the same period last year. AMD’s CFO, Henri Richard, stated:
“Our data center business faced headwinds from inventory corrections at key OEMs and delayed ramp of our MI300X-based systems.” — Henri Richard, Chief Financial Officer, AMD
Market Context and Supply Chain Implications
This bifurcation reflects real-world procurement dynamics across hyperscale data centers. According to the report, Microsoft Azure and Meta Platforms collectively accounted for 42% of NVIDIA’s Q1 2026 data center sales — up from 31% in Q1 2025. Meanwhile, Dell Technologies reduced its AMD-based server orders by 35% quarter-on-quarter, per supply chain tracking firm TrendForce. For supply chain professionals, these shifts mean tighter allocation controls on NVIDIA’s GH100 and B100 chips, extended lead times averaging 24 weeks for Blackwell-based modules, and increased scrutiny of second-source qualification timelines for AI accelerators. The source states that TSMC’s CoWoS packaging capacity utilization hit 98% in March 2026 — the highest since 2022 — driven primarily by NVIDIA’s wafer commitments.
Broader Industry Alignment
Other major players show similar polarization. Intel’s AI chip division posted a 12% revenue increase in Q1 2026, reaching $1.8 billion, but remains below its $3.5 billion full-year guidance due to slower-than-expected Gaudi 3 adoption. Separately, Broadcom reported AI-related networking chip revenue of $5.4 billion — a 67% jump year-on-year — underscoring that AI infrastructure demand extends beyond compute silicon to high-speed interconnects. These figures align with BloombergNEF’s April 2026 estimate that global AI chip capital expenditures totaled $48.3 billion in 2025, up 89% from $25.6 billion in 2024. From a practitioner standpoint, procurement teams are now prioritizing multi-vendor roadmaps, accelerating validation of AMD MI300A and Intel Gaudi 3 systems for inference workloads, and renegotiating minimum order quantities with foundries like TSMC and Samsung.
Source: Bloomberg
Compiled from international media by the SCI.AI editorial team.
