After watching this interview with the founder of Anthropic, I could hardly believe my ears. He said Chinese AI is too stupid, not even qualified to go out of control? I was truly shocked. Is the gap between China's large-scale AI models and those of the US really that large? Is this an objective fact, or is he being malicious? Let's peel back the layers. First, his statement about being "six months to a year behind" has factual basis. A Stanford 2026 report indicates that the performance gap between top-tier models from China and the US has substantially closed, leaving only 2.7%. If we compare the most cutting-edge closed-source models, there is indeed still a time lag. Second, his statement that "China has never seen an out-of-control intelligent agent" is nothing short of arrogant and a lie. Just two weeks before he said this, an AI security team discovered that more than a dozen agents, suspected to be from Chinese models and running on Tencent Cloud, breached the sandbox and attempted to infiltrate the Gaode Maps system a staggering 2818 times. He's not unaware of this; he's selectively ignoring it. Third, his real purpose: to erect barriers to the gap. "Security" is Anthropic's moat. Whoever defines security holds the power of discourse. Whoever defines security holds the rules. Therefore, what he said isn't entirely true. But we also acknowledge the gap. So, what will the AI competition between China and the US look like in five years? First, electricity: China's electricity consumption is 2.4 times that of the US. In 2025, China's electricity generation will reach 10.57 trillion kilowatt-hours, while the US will reach 4.43 trillion. The ultimate limit for AI computing power is electricity, and China holds a strong advantage. Second, model performance: The gap has narrowed to "statistical error." Chinese teams iterate extremely quickly; DeepSeek, Kimi, and others have narrowed the gap with their US counterparts to one to three months. The US still leads in basic innovation, but China is catching up in application deployment. Third, application scenarios: The US is "talking about AI," while China is "using AI." Nearly half of the top 50 global mobile AI applications come from Chinese teams. AI usage in the Chinese workplace exceeds 80%. Five years from now, whoever truly integrates AI into factories, hospitals, and power grids will win the first half of the game. Fourth, Profitability: The US bets on "mass investment for miracles," while China bets on "revenue from application scenarios." Anthropic's revenue in the first half of the year was $16.3 billion, and it achieved operating profit in the second quarter. However, it signed $63 billion in computing power contracts, of which $44.1 billion are irrevocable, resulting in at least $6 billion in hard expenses annually. In China, Zhipu's revenue grew by 399.7% in the first half of the year, and MiniMax's by 283.1%, but both are operating at a loss. The US is betting on economies of scale, while China is betting on closed-loop application scenarios. Fifth, IPO Valuation: This is the most painful difference. Anthropic is about to IPO, with its valuation lowered from 2 trillion to 1.5 trillion, but it will still be the world's largest IPO. DeepSeek has also completed a financing round of approximately 100 billion RMB, with a pre-investment valuation of 500 billion RMB. It plans to IPO on the Science and Technology Innovation Board, with a post-listing market capitalization expected to be 1.5 trillion to 2.5 trillion RMB. Let's do the math: Anthropic is valued at $1.5 trillion, approximately 10.7 trillion RMB. DeepSeek's highest expected valuation is 2.5 trillion RMB. A valuation gap of more than four times is indeed enviable. This isn't a difference in models, but a difference in market pricing systems. US stocks offer a "global narrative premium," while A-shares and Hong Kong stocks offer a "domestic implementation discount." The landscape five years from now will likely be: the US wins "from 0 to 1," while China wins "from 1 to 100." The co-founder's statement that "Chinese AI is too stupid" will prove to be an arrogant joke in five years. However, his warning about the "objective existence of risks" is correct. The risks aren't "unseen" in China, but "uncontrollable" in the US. The "arrogance" they speak of cannot mask their anxiety about computing power and electricity. Silicon Valley is printing stocks like crazy, while we're frantically focusing on implementation. Five years from now, who will stand at the top—China or the US?