The global AI chip market is witnessing a significant bifurcation between the US and China, driven by geopolitical tensions, supply chain disruptions, and divergent technological strategies.
The latest TrendForce research provides insights into the strategic maneuvers of US and Chinese CSPs (Cloud Service Providers) in the AI chip sector, highlighting emerging trends, market shares, and potential trajectories.
US AI Chip Market Developments
Shift Toward In-House ASICs
US-based CSPs are accelerating their in-house ASIC (Application-Specific Integrated Circuit) development to reduce dependency on dominant chip suppliers like NVIDIA and AMD. This trend is driven by the need for cost control, enhanced performance, and supply chain flexibility.
Google: Leading the charge with its TPU v6 Trillium, Google has moved from a single-supplier strategy (Broadcom) to dual-sourcing, adding MediaTek as a strategic partner. This approach enhances design flexibility and supply chain resilience while advancing AI chip capabilities.
AWS: AWS is focusing on its Trainium series, co-developed with Marvell, to handle generative AI workloads and LLM training. Trainium v3 is in collaboration with Alchip, and is expected to see significant shipment growth in 2025.
Meta: Meta has developed its MTIA accelerator for inference workloads and is now working with Broadcom on MTIA v2, with a strong focus on energy efficiency and low-latency architecture.
Microsoft: Despite its heavy reliance on NVIDIA GPUs, Microsoft is pushing its Maia series of ASICs, partnering with Marvell to mitigate risks and enhance chip design capabilities for the Azure platform.
China’s Strategic Shift in AI Chip Market
Reducing Reliance on US Chips
China’s AI chip market is undergoing a transformation in response to US export controls introduced in April 2025, aiming to reduce imported chips’ share from 63 percent in 2024 to 42 percent in 2025.
Huawei: Huawei is expanding its Ascend series to cater to local AI infrastructure needs, including LLM training and AI-powered telecom networks. Supported by national policies, Huawei is positioned to challenge NVIDIA’s market share in China.
Cambricon: Cambricon is ramping up its Siyuan (MLU) chip series, targeting both training and inference for major Chinese CSPs. Following feasibility tests in 2024, widespread deployment is anticipated in 2025.
Alibaba, Baidu, Tencent: Major Chinese CSPs are developing proprietary ASICs to reduce dependence on US chips. Alibaba has launched the Hanguang 800 for inference, Baidu is advancing its Kunlun series for training and inference, and Tencent is integrating Enflame’s ASIC solutions to bolster its AI capabilities.
Emerging Trends and Implications
Bifurcated Ecosystem: The AI chip market is rapidly bifurcating into two ecosystems—one within China driven by domestic CSPs and chipmakers, and another dominated by US technology giants.
Government Policies and Subsidies: Government support is pivotal in shaping both markets. The US is incentivizing in-house ASIC development, while China is strategically investing in domestic semiconductor manufacturing to mitigate the impact of US export controls.
Supply Chain Resilience: Dual-sourcing strategies by US CSPs (e.g., Google, AWS) and China’s aggressive pursuit of domestic alternatives (e.g., Huawei, Cambricon) underscore the critical importance of supply chain diversification.
The AI chip market industry is evolving rapidly, driven by strategic shifts in both the US and China. While US CSPs are intensifying in-house ASIC development to maintain technological leadership, Chinese players are leveraging national policies and local market demand to establish AI chip independence. As geopolitical tensions persist, the global AI chip market is likely to remain divided, with each region striving to secure its own supply chain and technological ecosystem.
Baburajan Kizhakedath
