Kimi K3 от Moonshot AI: новый «момент DeepSeek» и спор о чипах против архитектуры

Chinese lab Moonshot AI has released a large language model called Kimi K3, and in July 2026 it reignited a debate in Silicon Valley over whether China, despite limited access to advanced chips, can catch up with flagship US models — as reported by South China Morning Post.

What is known about Moonshot AI

Moonshot AI is a Chinese startup based in Beijing behind the Kimi AI assistant; the company had previously released models in the same lineup, including Kimi K2. Kimi K3 continues that line and is positioned as a response to flagship Western models. South China Morning Post does not disclose detailed technical specifications of the new model in its report, focusing instead on the significance of the launch for the global debate over AI chips and architecture.

Why is this being called a second "DeepSeek moment"?

Kimi K3 is being compared to the breakthrough of DeepSeek, which shocked Silicon Valley in early 2025. At the time, the Chinese model demonstrated a competitive level of performance at markedly lower cost than American labs, calling into question the idea that leadership in AI can only be bought with massive computing power. The launch of Kimi K3, as South China Morning Post notes, has brought the same question back onto the agenda — but now in a new round of the 2026 race.

Chips versus architecture

The core of the debate is whether limited access to chips can be offset by model architecture rather than by the number of accelerators. Because of US export controls, tightened since 2022, Chinese developers cannot freely purchase the most powerful Nvidia accelerators, and the question is whether they are closing that gap through engineering solutions in the design of the models themselves. Proponents of this view believe that Chinese teams' success in boosting performance through architectural innovation is changing the economics of the entire industry and undermining the rationale for extremely expensive Western computing.

"Whether

China can overcome limited access to advanced chips and match the performance of flagship US AI models" — this is how South China Morning Post frames the industry's key question.

Why trillions of dollars are at stake

According to South China Morning Post, trillions of dollars may hinge on the answer to this question. Some investors and engineers in the US see Chinese developers' achievements as a threat to the central justification for the enormous capital spending of American companies — hundreds of billions of dollars invested in chips and data centers. If comparable results can be achieved more cheaply and on weaker hardware, then the valuations of companies whose worth is built on compute scarcity come into question. That is why yet another strong Chinese model is being perceived in the Valley not as a routine release, but as a signal of risk to an entire investment logic.

What this means

Kimi K3 is another sign that the technological gap between the US and China in AI is measured not only by access to chips. It is not yet known whether the model will repeat DeepSeek's market impact, but the debate itself over what matters more — hardware or architecture — has once again become central to the industry in 2026.