Signal · AI Infrastructure
AI Supply Chain Strain as Nvidia, AMD, and Chinese Competitors Accelerate Next-Gen Infrastructure Expansion
AI GPU and Chip Supply Chain Expansion Highlights Nvidia, AMD, and Chinese Competitors in Next-Gen Infrastructure Over the past 72 hours, OSINT indicates ongoing capacity build-out and supply chain adjustments among major AI hardware producers, with key developments involving Nvidia, AMD, and…
Thesis
AI GPU and Chip Supply Chain Expansion Highlights Nvidia, AMD, and Chinese Competitors in Next-Gen Infrastructure
Key findings
- Over the past 72 hours, OSINT indicates ongoing capacity build-out and supply chain adjustments among major AI hardware producers, with key developments involving Nvidia, AMD, and Chinese manufacturers. The focus remains on next-generation GPU and AI chip production, reflecting sustained demand and regional capacity expansion.
- Nvidia’s H200 production ramp at TSMC is reportedly increasing CoWoS packaging output to approximately 45,000 wafers per month by Q2 2025, signaling continued supply chain strain and high GPU demand. AMD confirmed volume shipments of MI300X to hyperscalers, with a 2025 production target of around 1.5 million units, suggesting market share gains in AI accelerators. Nvidia’s Blackwell (B100/B200) tape-out at TSMC N3 process has been completed, with mass production scheduled for late Q2 2025, confirming the GPU timeline remains on track.
- Chinese manufacturer Huawei’s Ascend 910B AI GPU is benchmarked at roughly 80% of A100 performance, with mass production underway at SMIC’s 7nm process, indicating domestic capabilities are improving despite export restrictions. Samsung’s HBM4 memory roadmap has advanced to Q1 2026 for mass production, while HBM3E volume ramp-up is set for Q2 2025, critical for next-gen GPU bandwidth scaling.
Analysis
Supply chain constraints for Nvidia’s CoWoS substrates are easing following capacity expansions by Unimicron and Ibiden, potentially reducing GPU backlogs by mid-2025. Additionally, Foxconn’s new $1.5 billion AI server plant in Kaohsiung, operational in H2 2025, highlights downstream integration efforts to meet rising AI hardware demand. Chinese cloud firms Baidu, Alibaba, and Tencent are collectively ordering 150,000 domestic AI chips for 2025, mitigating the impact of U.S. export controls on Chinese AI compute infrastructure.
These signals collectively demonstrate a broad industry effort to scale AI hardware manufacturing capacity, with supply chain adjustments and regional production improvements supporting increased AI infrastructure deployment and chip availability.
The OSINT indicates significant capacity expansion among GPU and AI chip manufacturers, which could influence market liquidity and infrastructure scaling for AI hardware. The focus on supply chain resilience and regional production capabilities underscores ongoing efforts to meet rising AI compute demands and mitigate geopolitical risks.
The dataset does not specify detailed capacity utilization rates or inventory levels across the supply chain, nor does it include forward guidance beyond the stated production timelines and shipments.