Connor Pu • Analysis • 2025.10

Kimi-k2, another
DeepSeek moment?

The Strength of Kimi k2 The Kimi k2 thinking model is genuinely impressive. Since its launch in July, it has undergone two significant updates in September and November, rolling out a 256k context window, improved coding capabilities, and "thinking" with interleaved reasoning.

[Visual evidence: Kimi k2 leading in benchmarks. AI constructing high-complexity, high-precision websites. AI rendering a 3D version of Tokyo.]

Nvidia's Outlook With Nvidia's earnings report dropping on November 19, I remain bullish on NVDA for the long term. This confidence holds regardless of whether we look at the legacy market—where general-purpose computing is transitioning to accelerated computing—or the incremental demand from new markets like AI models, applications, 6G, quantum computing, autonomous driving, robotics, and digital factories. Built upon the competitive moat of hardware chip manufacturing + CUDA software + NVLink chip interconnects, Nvidia possesses massive cash flow and significant potential to scale up and scale out.

The "DeepSeek Moment" Question However, the pressing question is: Will Kimi k2 trigger a recurrence of the "DeepSeek Moment" from early 2025, causing a crash in US stocks? My answer is no.

Factually, both Kimi k2 and DeepSeek serve as Chinese open-source LLMs that, at extremely low costs, have surpassed or achieved parity with the most cutting-edge US closed-source LLMs in many domains. When DeepSeek was released in February 2025, its exposure exploded after a one-week lag, and the US stock market plummeted immediately. The critical difference, however, lies in the timing: between early 2025 and late 2025, the mainstream consensus—both in the US and globally—regarding China's AI capabilities has undergone a seismic shift.

The Shattering of Consensus Before DeepSeek (DS) emerged, Trump had just taken office, and new AI policies had not yet been updated. Social expectations aligned with the rhetoric of Eric Schmidt, the former Google CEO and chief AI policy advisor during the Biden administration, who claimed, "The US is 2–3 years ahead of China in AI." The arrival of DS completely shattered this consensus. Schmidt later admitted his misjudgment, shifting his expectation to: "China is only 2–3 months behind the US in AI, and even ahead in certain aspects."

This raised a terrifying question for investors: "Is the high valuation of the US AI supply chain priced in too aggressively?" This realization was the primary driver of the broad market correction. (In this mainstream climate, the "China AI Threat Theory" also emerged and was exploited; CEOs across the industry used the threat narrative to elevate their AI business to the level of national security to secure investment and government backing. For example, Jensen Huang stated that China is "nanoseconds" behind the US, a move designed to lobby the government to relax regulations so Nvidia could sell more chips to China and establish a monopoly.)

Policy Shifts: Biden vs. Trump Even the AI policy during the Biden era was built entirely on the consensus that China was lagging by 2–3 years. Following the CHIPS Act's ban on high-end Nvidia exports, US policy favored heavy regulation. The strategy was to designate a select few AI companies to develop and close-source massive amounts of cutting-edge technology, developing AI slowly and fully within a "safe" framework. Bluntly put, the goal was to ensure that the products and technologies of AI companies served the American elite, helping the establishment solidify a stable monopoly on AI technology and applications both domestically and globally.

I vividly remember early 2025, when entrepreneurs like Sam Altman and Elon Musk accompanied Trump to visit the Saudi Crown Prince face-to-face. Sam made a specific hand gesture, symbolizing a strategy: internally develop highly disruptive new technologies while simultaneously controlling and suppressing the speed at which uncertainty ferments in society. This guarantees the monopoly status of the US and its corporations, eventually allowing for a "dimensional strike" (strategic overmatch) against other nations after a period of time. This is a playbook Biden was very familiar with; as a veteran politician from the Cold War era, he understood that the Gulf War—where the US launched a technological "dimensional strike" against Iraq (then the world's 6th largest military)—was backed by exactly this kind of policy.

The New Paradigm But the policy changed after Trump took office. It wasn't just a change of political parties and interest groups; the consensus and expectations regarding AI also shifted. David Sacks, representing the Silicon Valley right-wing faction, introduced the "AI Action Act" as Secretary. Facing the reality of China catching up, the new approach involves strengthening export restrictions on AI chips and related supply chains while simultaneously fully deregulating competition among companies. The government has even implicitly stepped in to drive AI investment, binding the economy to AI.

By now, in late 2025, the US has accepted the "China AI Threat" consensus. Regarding how Kimi achieved such high quality, China's strategy can be summarized by this quote: "The US has the strongest chips and AI infrastructure; they spent $1 trillion to build a 99/100 model. The 'East' (China) simply waited, and once the path was clear, they used inferior AI infrastructure and spent $1 billion to build a 90/100 model." —The US AI industry has tacitly accepted this consensus.

Under these expectations, if a Chinese open-source LLM suddenly appears at an extremely low cost and rivals top-tier US LLMs, it falls within the realm of social expectations and will not cause panic.

The real disaster for US stocks arrives when the 'East' uses $1 billion to create something that comprehensively surpasses what the US built with $1 trillion. This has not happened yet.

Therefore, the so-called "DeepSeek Moment" implies more than just China developing an open-source LLM at low cost that rivals the US. The essence of the DeepSeek Moment was the shattering of society's expectations and consensus regarding China's standing in AI.