AI NEWSLatest AI NewsArticleGoogle Plans to Stock 15 Million Self-Developed AI Chips by 2028, Directly Challenging Nvidia’s Dominance in Computing PowerPublished in Latest AI NewsTime :Aug 3, 2026Read :3minuteAccording to a market analysis report cited by Taiwanese media Jutai, Google plans to deploy 12 million to 15 million of its self-developed TPU v9 AI acceleration chips by 2028. Considering that the analysis suggests NVIDIA’s AI GPU shipments may reach around 12.4 million units by 2028, Google is expected to surpass NVIDIA in server AI accelerator shipments.TPU is an ASIC chip specifically designed by Google for AI training and inference. Its performance has been continuously iterated through internal use and has never been publicly sold. Shifting from renting NVIDIA GPUs to developing in-house chips is a key move for Google to reduce computing costs and reduce reliance on a single supply chain.4 Compute Die Architecture, Sharp Increase in Advanced Process Capacity DemandIn terms of chip architecture design and process technology, the report suggests that TPU v9 will adopt a 4 Compute Die structure. This design will further challenge the capacity limits of advanced processes and packaging, meaning that Google will have to seek manufacturing partners other than TSMC.Intel and Samsung foundry services are expected to gain new business opportunities - when a single supplier cannot meet Google’s capacity needs, a multi-supplier strategy becomes inevitable. This also reflects a deeper transformation in the industry: leading tech giants are accelerating from simply purchasing general-purpose GPUs to developing customized AI chips. As Google prepares to build an independent computing infrastructure with millions of self-developed chips, it may be facing the strongest challenger to NVIDIA’s dominance in AI chips.