[para. 1] The global competition to lead in artificial intelligence (AI) is fueling an extraordinary surge in semiconductor investments, with the United States and China adopting distinct yet aggressive strategies. The U.S. pursues technological dominance through substantial capital investment and market-driven procurement, while China responds to increasing U.S. sanctions by working towards self-sufficiency and rapidly developing its domestic chipmaking sector. This rivalry has transformed chipmakers in both countries into highly sought-after investment opportunities, with enormous implications for future technological and geopolitical landscapes. The primary question that remains is which country’s approach will first achieve artificial general intelligence (AGI) — a development likely to shape global power dynamics for decades[para. 1][para. 2][para. 3][para. 4].[para. 2] The American approach is characterized by large-scale capital deployment. Over the past month, OpenAI concluded massive chip procurement deals with Broadcom, AMD, and Nvidia, requiring a total of 26 gigawatts — an amount of energy comparable to powering nearly three entire New York Cities at peak usage. The U.S. leverages deep capital markets, leading to soaring share prices for major chipmakers: Broadcom surged nearly 10% after its deal announcement, AMD jumped 37%, and Nvidia’s value rose 3.93%. Nvidia’s market capitalization reached $4.42 trillion, making it the world’s most valuable public company, while AMD is now worth $378 billion and Broadcom $1.63 trillion, up 50% this year. OpenAI’s ambitions are further evidenced by its $500 billion Stargate infrastructure initiative to build massive data centers and secure high-speed DRAM supply from Samsung and SK Hynix. However, these aggressive investments raise concerns about “circular deals,” potentially inflating both actual and expected demand and creating the risk of a speculative bubble not unlike that of the dot-com era. The AI-driven bull market has seen the S&P 500 soar nearly 90% since October 2022, powered by the capital expenditures of tech giants like Apple, Microsoft, Alphabet, Amazon, and Nvidia[para. 2][para. 5][para. 6][para. 7][para. 8].[para. 3] In China, the pursuit of AI and semiconductor self-sufficiency has become imperative due to tightening U.S. export restrictions. A report by Bernstein had estimated China’s AI chip demand for 2025 at $39.5 billion, but U.S. regulatory obstacles have widened the anticipated supply gap to over $10 billion as a resumed supply of Nvidia H20 chips failed to materialize. Major domestic investment is being funneled into homegrown firms, with publicly listed Cambricon Technologies reporting its revenue rising by 43 times to $404 million in the first half of the year after a major order from ByteDance, resulting in its shares becoming the most expensive in China for a time. Startups like Moore Threads are fast-tracking IPOs, and even companies with small stakes in these ventures have seen dramatic share increases. Despite investor excitement, these companies remain unprofitable; for instance, from 2022 to 2024, Moore Threads lost 4.6 billion yuan (versus 3.8 billion in R&D), and MetaX lost 2.72 billion yuan in the same period. Domestic foundries face significant production bottlenecks, with AI chip yields below 20%. Founders of leading startups often have deep experience from U.S. firms such as AMD and Nvidia[para. 9][para. 10][para. 11][para. 12][para. 13][para. 14][para. 15][para. 16][para. 17][para. 18][para. 19][para. 20][para. 21][para. 22][para. 23][para. 24][para. 25][para. 26][para. 27].[para. 4] Facing advanced foundry restrictions, Chinese tech firms are turning to alternative systems—a large-scale, cluster-based approach using thousands of less advanced domestic chips to rival top U.S. performance, albeit at greater energy cost. Huawei leads this model, orchestrating supercomputers like the Atlas 900, which delivers 300 PFLOPS by linking 384 Ascend 910C chips. While individual Ascend chips have just one-third the capacity of Nvidia’s best GPUs, system-level performance can be nearly double according to Semi Analysis—though with 2.5 times the electricity requirement. The Chinese ecosystem is also building independent software stacks, such as Huawei’s CANN, which is open-sourced to encourage adoption. Other tech giants, such as Baidu and Alibaba, are developing CUDA-compatible chips to lower barriers for application development. Despite the rapid build-up of infrastructure, actual demand has lagged, with under 30% utilization at new computing centers, leading to expectations of industry consolidation. Ultimately, success for these companies hinges on real market-driven applications, rather than supply-side expansion alone[para. 28][para. 29][para. 30][para. 31][para. 32][para. 33][para. 34][para. 35][para. 36].AI generated, for reference only