Core Insights - The competition between Google's TPU and NVIDIA's GPU is intensifying, with reports indicating that Google's Gemini 3, trained on TPU, outperforms OpenAI's ChatGPT 5, which is trained on NVIDIA's GPU [1][3] - The stock market has reacted to this competition, with NVIDIA's shares dropping by 12.59% while Google's shares rose by 12.85% since November [1] - The rise of Google's TPU may present both opportunities and challenges for Chinese companies embedded in the global computing power supply chain [1] Custom vs. General Chips - The battle between Google TPU and NVIDIA GPU is framed as a competition between customized chips and general-purpose chips, focusing on efficiency and cost rather than a direct rivalry [2] - Historical parallels are drawn to other industries where both types of products coexist, suggesting that TPU's core demand is cost reduction [2] Technical Architecture Differences - Google's TPU is seen as superior in performance and cost, but NVIDIA's GPU offers better ecosystem openness and compatibility [3] - Despite TPU's advantages, NVIDIA's GPUs remain the preferred choice for many manufacturers due to their strong compatibility with existing technologies [3] Future Market Dynamics - The competition is likened to a relay race, with both companies rapidly iterating their chip technologies [4] - Predictions indicate that by 2029-2030, the market share between customized chips and GPUs may reach a 50-50 split, although NVIDIA is expected to maintain dominance until around 2026 [4] Impact on Supply Chain - The competition for computing power is driving higher demands for data transmission efficiency, benefiting hardware supply chains, particularly in the light module and PCB sectors [5][6] - If Google's TPU gains market share, it could lead to significant growth in the light module market, with estimates suggesting TPU v7 may require 3.3 times more light modules than NVIDIA's Rubin [7] Investment Sentiment - While there is optimism about TPU's cost advantages, some investors express caution, noting that a shift to lower-cost TPUs could lead to valuation pressures in the hardware supply chain [8] - The current AI landscape is characterized by a lack of standout applications, with the focus still on computing power rather than software solutions [9] Broader Industry Implications - AI is reshaping traditional industries, with key areas of focus including humanoid robots, smart driving, and AI in drug development [10] - The ongoing debate about whether the AI sector is experiencing a bubble is influenced by comparisons to the 2000 internet bubble, though current indicators suggest a healthier industry with strong revenue growth [11][12] Valuation Perspectives - Current AI leaders have lower projected P/E ratios compared to the peak of the internet bubble, indicating a more sustainable growth outlook [12] - The potential for AI applications to emerge as market leaders remains uncertain, with the need for significant breakthroughs to validate current valuations [13]