After the mid-July release of Kimi K3, the overseas tech community quickly shifted the discussion from model parameters and rankings to a more easily spreadable question: Why didn't the U.S. retain Yang Zhi?
This question has drawn market attention, not just because of the founder's typical background. Kimi K3 announced the path of open weights, with the official announcement that the full model weights will be open-sourced by July 27, 2026. For enterprises, this relates to whether they can deploy models in a private environment, control data, and costs. For investors, it will touch on the business barriers of closed-source model companies like OpenAI and Anthropic.
The controversy escalated because two prominent figures gave different explanations. Vinod Khosla, the founder of Khosla Ventures, criticized U.S. immigration policy for driving away top global AI talent in the wake of Kimi K3's success. Yang Zhi's mentor at CMU, Russ Salakhutdinov, later countered on Platform X, stating that Yang Zhi had many opportunities to stay in the U.S., including contact with Apple executives, but ultimately chose to return to China for entrepreneurship.
The value of Kimi K3 lies not in proving that Chinese AI has already taken a comprehensive lead, nor in proving that the U.S. has lost the talent war. It brings to light a more practical issue: open models, the entrepreneurial environment, and talent choices are reshaping the global pricing reference for AI.
K3 Makes Overseas Take China's Open Models Seriously
Kimi K3 first attracted attention because it demonstrated visible performance in high-value tasks. The Kimi official blog stated that K3 adopts a MoE (Mixture of Experts) architecture with 28 trillion parameters, activating 16 experts from a pool of 896 each time. In simple terms, the model is very large but only calls upon a portion of its capabilities at a time to balance effectiveness and reasoning costs.
The Big Model competition is no longer just about chatbot experiences. The market is more interested in whether AI can write code, break down multi-step tasks, engage in long-term planning and error correction, known as Agent tasks (autonomously completing multi-step work). These scenarios are closer to software development, enterprise automation, and the future of work processes.
Kimi officials stated that K3 supports multimillion-level contexts and has native visual capabilities, performing well in some coding and Agent evaluation tasks. Limitations need to be considered as well. The company's own stance mentioned that K3 still lags behind some of the cutting-edge closed-source models overseas, and some leaderboard scores still require more independent validation.
Nevertheless, K3 has done enough to prompt overseas developers and investment circles to reassess Chinese open models. In the past, many Chinese models were seen as low-cost alternatives. Now, the market is starting to discuss whether they can take a leading position in higher-value business scenarios such as programming, Agent tasks, and long contexts.
Open sourcing amplifies this signal. Closed-source models are more like cloud services accessed through APIs. Open sourcing allows enterprises to deploy and customize models locally or in private environments after obtaining the weights. For businesses concerned about data leaks, cost control, or vendor lock-in, this is primarily a procurement choice, not a technological stance.
The Case of Yang Zhilin Cannot Be Simplified as America Losing Him
Yang Zhilin's experience could easily be framed within America's talent anxiety. Public information shows that he graduated from Tsinghua University with a bachelor's degree, later earned a Ph.D. from CMU, worked at Google Brain and Meta AI, and then returned to China to found Moonshot AI with support from Alibaba, Tencent, and others.
Khosla's assessment tapped into a long-standing pain point in the American tech scene. H-1B visas, green cards, and geopolitical constraints make the path for international students to stay in the U.S. more uncertain. Top labs heavily rely on global talent, and the more U.S. policies create friction, the more likely the marginal talent flow will change.
However, Russ Salakhutdinov's perspective altered the cause-and-effect link in this case. As Yang Zhilin's mentor, he provided an insider's view: Yang Zhilin did not lack opportunities to stay in the U.S. but believed that if he didn't return to China to start a business, he would regret it for a lifetime. Directly attributing Kimi K3 to U.S. immigration policy causing talent outflow lacks sufficient evidence.
A more fitting explanation is that both forces coexist. U.S. immigration and international exchange environments indeed subject foreign talent to more uncertainty. Simultaneously, China's local entrepreneurial opportunities, capital support, and open model approach are forming a strong enough attraction. The former explains why America is anxious, and the latter explains why this choice may not merely be a passive departure.
For investors, a single case may not launch a systemic loss of talent to the U.S. But it is enough to illustrate that the top choice for AI talent is no longer automatically staying at a U.S. tech giant or starting a venture within the U.S. academic system.
Open Weight Challenges Closed-Source Commercial Control
The larger significance of Kimi K3 is to link the talent narrative with the model path narrative. A returnee entrepreneur has created an open model with global visibility, and the overseas discussion naturally extends from who trained the model to who controls the future distribution and deployment of AI.
The core advantage of closed-source model companies still lies in the strongest model capabilities, a unified API, developer ecosystem, and enterprise customer relationships. Once open weight models approach in programming, agents, and long-context scenarios, it will change the corporate negotiation structure. Enterprises may not completely migrate from closed-source models, but they can use open models to replace some workloads, reduce costs, and retain more data sovereignty.
Tencent's Hy3 Mixnet, Alibaba's Qianwen, and overseas open models are being put on the same track because companies are looking for a second set of infrastructure. Tencent's quarterly report mentioned that Hy3 preview has become one of the most widely used models on OpenRouter since April 28, 2026.
These cases do not prove that open weights have defeated closed source, but they indicate that developers and companies' concerns about data control are on the rise. Microsoft CEO Nadella has warned companies about the risk of providing data to proprietary models, while Hugging Face has long advocated that open models can reduce centralization of power. They are not endorsing a specific Chinese model but reflecting the real concerns of enterprise customers.
For asset pricing, this will affect two types of narratives. Companies like Alibaba and Tencent that own models, cloud, and application distribution may benefit from the expansion of the open model ecosystem. Closed-source cutting-edge companies like OpenAI and Anthropic still have a competitive advantage, but the valuation narrative needs to continue to prove that the lead margin, enterprise stickiness, and security compliance capabilities are enough to offset the open model catch-up.
Adoption Rate Will Determine the Revaluation Slope
What Kimi K3 is now providing is a strong early signal, not yet an industry inflection point. Leading specific evaluations, shifts in overseas public opinion, and the release of open weight plans indicate that the market is beginning to look for new benchmarks, but they do not equate to enterprise revenue, long-term retention, and production-level alternatives.
What needs to be verified is whether open weight models can transition from developer trials to enterprise deployment. Whether enterprises are willing to move core workflows to models like Kimi K3, Hy3, or Qianwen depends on stability, inference cost, security audits, toolchain compatibility, and service support, not just leaderboard rankings.
The same goes for talent. Yang Zhilin has proven that entrepreneurship in China can produce globally visible results. Whether this will lead to a larger scale talent migration will depend on potential adjustments to U.S. policies, the ability of Chinese companies to sustain a world-class research environment, and the actual choices of the next wave of top Ph.D. candidates.
One of the most noteworthy aspects of Kimi K3 for investors is that it has made previous assessments no longer stable. Top-tier AI capabilities may not necessarily be defined solely by closed-source giants, and top talent may not necessarily need to exclusively operate within the U.S. system to achieve maximum impact. The initial hype around the rankings will subside, and the true extent of this reassessment will depend on whether the full weighting is actually opened up as planned, if companies genuinely adopt the technology, and if developers continue to engage.
Welcome to join the official BlockBeats community:
Telegram Subscription Group: https://t.me/theblockbeats
Telegram Discussion Group: https://t.me/BlockBeats_App
Official Twitter Account: https://twitter.com/BlockBeatsAsia
