Group 1 - The founder and CEO of DRW, Don Wilson, suggests that global spending on GPUs may surpass that on oil in the next decade, highlighting the increasing importance of GPUs as a core resource for AI training [2][3] - The demand for GPUs is expected to explode, with the International Energy Agency projecting that electricity demand for AI data centers in the U.S. will reach 123 million kilowatts by 2035, which is 30 times the level in 2024 [3][2] - The supply of GPUs is uncertain due to factors such as TSMC's production capacity, U.S. export controls, and NVIDIA's product release schedule, leading to potential volatility in the market [3][4] Group 2 - The financialization of GPUs could lead to the creation of futures contracts and indices similar to those for oil, copper, and gold, allowing companies to hedge against price fluctuations [3][4] - Historical trends show that financialized commodities often experience bubbles and crashes, raising concerns about the potential for similar outcomes in the GPU market [4][5] - Unlike oil, which can be stored long-term, GPUs have a short lifecycle due to rapid technological advancements, making them more akin to perishable goods [4][5] Group 3 - Long-term investment success in commodities typically comes from companies that hold advantageous positions in the supply chain, such as manufacturers like TSMC and designers like NVIDIA, rather than from speculative trading in GPU futures [5][6] - The concept of "computing power capitalism" suggests a shift in resource perception from tangible materials like coal and oil to intangible assets like data, algorithms, and computing power [5][6] - The market will likely find ways to financialize new demands, but investors should focus on identifying companies and industries that will benefit from the emerging "computing power capitalism" rather than speculating on GPU futures [6]