China's Moonshot AI Unveils Kimi K3, Triggering a Second DeepSeek Shock in Silicon Valley

At the same time, Moonshot AI acknowledged that the new system still trails the absolute proprietary U.S. top models overall — namely Anthropic's Claude Fable 5 and OpenAI's brand-new GPT-5.6 Sol. The gap is shrinking rapidly, however. On specialized tests in the areas of programming and complex applications, Kimi K3 outperforms models such as Claude Opus 4.8 and GPT 5.5, the company said.

Lead in Web Programming

The lead is particularly pronounced in website and frontend programming. In the LM Arena, an independent ranking for AI models, Kimi K3 reached first place in six out of seven frontend domains — including Brand & Marketing, Reference-Based Design, and Data & Analytics. Overall, the model climbed from eighteenth to first place, a leap of seventeen positions. The competition from Silicon Valley came under visible pressure as a result.

Observers are drawing immediate parallels to the so-called "DeepSeek moment" in early 2025, when a then-little-known Chinese research lab attracted international attention with DeepSeek R1 and caused U.S. tech stocks to plunge by $1 trillion. With the release of Kimi K3, stock prices have now begun to slide again, which industry circles are discussing as a "second DeepSeek shock."

Price Advantage as a Strategic Lever

Alongside performance, price plays a central role. According to current information, Kimi K3 costs roughly half as much as ChatGPT and significantly less than Anthropic's top product Claude Fable 5. Analysts believe this combination of performance, customizability, and cost efficiency could fundamentally reshape the global market for AI models, provided the model can be exported.

Moonshot AI announced that the software behind Kimi K3 will be made fully available to developers worldwide on July 27. The model is expected to be fully open source by the end of July 2026. Whether Kimi K3 permanently shifts market dynamics will, according to industry observers, only become clear after the release of the source code and the first independent practical tests.

Right on cue for the unveiling of Kimi K3, China's head of state Xi Jinping opened the World AI Conference in Shanghai. It is the president's first speech at the conference, which has been held annually since 2018, and underscores the increased importance of artificial intelligence for the ruling Communist Party. In his address, he emphasized according to reports: "Wir sollten das Risikobewusstsein stärken und gewährleisten, dass KI sicher und kontrollierbar ist".

Geopolitical Tensions and Espionage Allegations

The presentation comes at a time of major geopolitical tensions. Western observers accuse Chinese developers of using a method called "distillation" to extract the capabilities of American models in order to train their own systems faster and more cheaply. This involves simulating millions of interactions with U.S. models and using the results to train Chinese models. Moonshot AI has not publicly commented on these allegations.

Databricks Continues to Forgo an IPO

Parallel to the announcement from China, the U.S. data and AI company Databricks is making headlines. The company raised fresh capital again and increased its valuation to $188 billion, but continues to forgo an IPO. According to a Wall Street Journal report, the current round has a volume of approximately $3 billion.

CEO and co-founder Ali Ghodsi justified the strategy with a change in customer behavior: companies are moving from "tokenmaxxing" to "valuemaxxing." "Statt für jede Aufgabe das teuerste Modell einzusetzen, geht es darum, das beste Ergebnis pro Dollar zu erzielen und frei das passende Modell für den jeweiligen Anwendungsfall zu wählen", Ghodsi said. Databricks reports an annualized revenue rate of $5.4 billion with year-over-year growth of more than 50 percent and is free cash flow positive.

More than 20,000 organizations worldwide use the Databricks platform according to the company, including adidas, Bayer, Mastercard, and 70 percent of Fortune 500 companies. Ghodsi emphasized that Databricks is "IPO-ready." An IPO is considered a certainty but could not take place until 2027 or later. The main reason he cited was not a need for capital, but rather a "Markt-Transaktionsmechanismus" through which employees could sell their shares. Ghodsi most recently stated in June that 2026 is "ein schreckliches Jahr für einen Börsengang."

The background to the delay is, according to industry assessments, a crowded IPO market: with SpaceX, Anthropic, and OpenAI, several mega-IPOs with a combined valuation in the trillion-dollar range are pushing into the public markets this year and absorbing enormous amounts of investor capital and attention there. Databricks only raised around $5 billion at a valuation of $134 billion at the beginning of 2026; at the end of 2024, the valuation after the $10 billion Series J was still at $62 billion.

Moonshot AI's own valuation is reported at $31.5 billion in a current funding round. The start-up is thus one of the highest-valued AI companies in China and visibly benefits from strategic support from the Alibaba Group as well as political backing from the government in Beijing.

Outlook on the Global AI Market

Industry observers see the Kimi K3 release as a turning point in the global competition for artificial intelligence. Only future models such as GPT-6 or Claude Opus 5 could widen the technological gap to Chinese systems again. Until then, the consensus view is that Silicon Valley has lost its previously comfortable lead in cutting-edge research.

For the coming weeks, market observers expect increased volatility on international tech stock exchanges. Should Kimi K3 actually be released as open source as announced and deliver in practice what its makers promise, the competitive pressure on established providers in the United States and Europe is likely to become noticeably greater. In general text generation as well, Kimi K3 performs well in consistent tests and surpasses all models from OpenAI, further underscoring the magnitude of the shift in the AI market.