The story of Moonshot AI begins on March 1, 2023 — a symbolic date marking the 50th anniversary of Pink Floyd's legendary album, which served as a source of inspiration for the company's founder, Yang Zhilin. What looked like a small IT office reflecting its creator's musical tastes has today become a key player in the global AI race. The company is preparing for its final funding round before a Hong Kong listing, and by my estimates, its valuation could reach an impressive $50 billion.

The Founder's Journey: From Tsinghua to Returning Home

Yang Zhilin is a typical representative of the new wave of Chinese technocrats. After earning a PhD from Carnegie Mellon University and interning at Google Brain and Meta, he made a choice many would consider unconventional — he returned to China. His path through developing the Huawei PanGu model and the Wu Dao project at the Beijing Academy of Artificial Intelligence demonstrates a systemic trend: China's best minds are increasingly choosing to work at home rather than emigrate to Silicon Valley. Yang's academic advisor, Ruslan Salakhutdinov, calls him "absolutely brilliant" — and this is not mere flattery but recognition that talent of this caliber is in demand even before defending a dissertation.

Phenomenal Growth and Real Numbers

Moonshot AI's valuation trajectory reads like a venture capital dream: from $4.3 billion at the end of 2025 to $10 billion by February, then $18 billion in March, $20 billion in May, and $35 billion in July. The key metric — ARR — doubled to $200 million in a single month and reached $300 million by June. Behind these figures stands a team of just 300 people, making the company one of the most efficient in the industry. More than 70% of revenue comes from corporate API sales, with the remainder from international subscriptions to Kimi.

However, the success of K3 has also exposed vulnerabilities. Within 48 hours of the release, the company had to suspend onboarding of new users due to a shortage of computing power. This is a classic gap between ambitions on paper and the physical constraints of infrastructure. Although the surge in demand confirms that the product exceeded market expectations, the question of scalability remains open.

Openness as a Strategic Weapon

Moonshot AI's key pivot is the shift to open weights. This is not altruism but a forced yet simultaneously brilliant strategy. The release of K3 with 2.8 trillion parameters — the largest open LLM in history — has created a problem for Washington that cannot be solved through legislation. Banning a model copied onto thousands of servers worldwide is technically impossible. At the same time, independent tests show hallucination rates rising to 51%, calling into question K3's unconditional superiority over competitors.

My assessment: Moonshot AI's openness is a double-edged strike. On one hand, it is a battle for global developers; on the other, a geopolitical maneuver that forces the United States into a difficult choice between restrictions and preserving an open ecosystem. The letter from 25 companies led by Nvidia and Meta is clear confirmation that the industry does not support prohibitive measures.

Sanctions as a Catalyst for Innovation

Paradoxically, U.S. export restrictions have acted as an accelerator of Chinese progress. Restrictions on HBM memory and advanced chips have forced Chinese labs to seek architectural workarounds instead of simply scaling up capacity. The result: hybrid attention architectures, aggressive context caching with a 90% discount on repeated tokens, and optimized reasoning generation.

The gap is narrowing but not disappearing. China is catching up through engineering ingenuity rather than technological superiority. However, the pace of change is such that policymakers cannot keep up. Open Chinese models with state-backed capital are already being used by developers worldwide — and this process is irreversible.

My verdict: Moonshot AI is not just a success story of a private company but an indicator of a paradigm shift in the global AI race. A company that dismantled its VIE structure and secured support from China's National Social Security Fund is a hybrid of private initiative and national strategy. The question is not whether it can maintain its pace, but how quickly Western regulators will learn to respond to a reality where open models blur the boundaries between commerce, politics, and technology.