Reshaping the Market Landscape: The Strong Rise of Google’s TPU
Google is collaborating with Broadcom to develop the TPU v7p, set to launch in 2026. The new generation Ironwood TPU can connect up to 9,216 chips in a single cluster, addressing data bottlenecks for complex models. This technological breakthrough directly impacts NVIDIA’s market position, causing NVIDIA’s stock price to drop over 7% at one point, marking its largest single-day decline in seven months, with a total market value evaporating by $1 trillion from its historical peak.
Even more concerning is that Meta Platforms is in talks to invest billions of dollars to purchase Google’s TPU chips for its own data center construction. Meta plans to rent TPU computing power through Google Cloud starting in 2026 and deploy such chips in its own data centers by 2027. This large-scale procurement plan signifies that Google’s TPU is beginning to disrupt NVIDIA’s core customer base.
Technical Strength Comparison: Performance and Ecosystem Battle
Following the significant drop in stock price, NVIDIA rarely reassured the market, claiming that its GPUs are a generation ahead of Google’s AI chips. NVIDIA stated that its technology remains a generation ahead of the industry and is the only platform capable of running all AI models and applicable in all computing scenarios. However, Google’s TPU technology indeed demonstrates strong capabilities. Google’s TPU is gaining market attention as a replacement for NVIDIA’s Blackwell chips, with training and inference performance improved fourfold compared to the previous generation, and Google TPU’s shipment volume is expected to maintain a leading position. Analysts point out that Gemini 3, trained on TPU and outperforming GPT, may replicate the earlier impact of DeepSeek, shaking NVIDIA’s GPU demand.
Industry Chain Impact: Comprehensive Transformation from Chips to Applications
This technological change is triggering a restructuring of the entire industry chain. Amazon AWS has announced a $50 billion investment to build infrastructure specifically for artificial intelligence and high-performance computing for the U.S. government, with the project expected to commence in 2026, adding nearly 1.3 gigawatts of computing capacity. Meanwhile, the global foundry market is also being driven by strong growth in AI demand, with global foundry revenue expected to reach $199.4 billion in 2025, a year-on-year increase of over 25%. The market size is projected to grow another 17% in 2026, surpassing $230 billion, with a compound annual growth rate of 14.3% from 2025 to 2030.
Opportunities and Challenges in the Chinese Market
In this AI chip competition, the Chinese market is showing a unique development trend. Alibaba’s Qianwen has surpassed 10 million downloads, with cloud business growth of 34%. The AI application “Qianwen” from Alibaba achieved over 10 million downloads in its first week of public testing. At the same time, Alibaba Cloud’s revenue grew by 34% year-on-year, with AI products achieving triple-digit growth for nine consecutive quarters. More notably, Singapore’s National Artificial Intelligence Strategy (AISG) is undergoing significant strategic adjustments, having abandoned the Meta model in its latest Southeast Asian language model project in favor of Alibaba’s Tongyi Qianwen open-source architecture, marking a key expansion of the influence of Chinese open-source AI models on the global stage.
Investment and Market Outlook
Alibaba revealed during its earnings call that it will “actively” invest in enhancing its artificial intelligence capabilities, not ruling out additional investments beyond the previously committed 380 billion RMB over three years. The company stated that an AI bubble is unlikely to occur within the next three years, and the overall demand for AI resources will remain in short supply. From a market performance perspective, U.S. tech stocks and chip stocks collectively strengthened, with major tech stocks rising collectively, including Tesla and Google, which both saw increases of over 6%. The chip sector performed well, with Broadcom’s stock rising 11%, adding $178 billion to its market value; Micron Technology rose nearly 8%, AMD over 5%, and NVIDIA over 2%, while the Philadelphia Semiconductor Index overall rose by 4.6%.
Future Outlook: Formation of a Multi-Polar Competitive Landscape
This AI chip battle signifies that the AI computing ecosystem is transitioning from NVIDIA’s GPU monopoly to a multi-polar competitive landscape. The rise of Google’s TPU not only challenges NVIDIA’s technological leadership but also provides the entire industry with new technological pathways and options. With the explosive growth in AI application demand, the need for computing power is surging, leading to a spike in demand for 800G/1.6T high-speed optical modules, while the supply shortage of upstream optical chips is becoming a new industry bottleneck. This supply-demand contradiction opens a window for domestic optical chip companies with technological breakthrough capabilities, significantly increasing the opportunities for domestic manufacturers. The competitive landscape of the AI chip market is undergoing fundamental changes, shifting from a single supplier dominance to the parallel development of multiple technological routes. This will not only accelerate technological innovation but also provide more choices for downstream applications, ultimately promoting the healthy development of the entire AI industry. In this process, the comprehensive competition of technological strength, ecosystem construction, and market strategy will become the key factors determining success or failure.