Yang Zhilin's Moonshot AI became the trigger for a new private-market valuation debate on July 20, when Bloomberg reported that IG market analyst Tony Sycamore estimated Kimi K3 had erased a combined $314 billion from valuation estimates for OpenAI and Anthropic.
The figure is Sycamore's assessment of two unlisted companies. It does not represent disclosed investor markdowns, completed transactions or losses observed in public markets. Bloomberg's report also does not provide his previous estimates, the revised values assigned to each company or the assumptions used in the calculation, so the $314 billion figure cannot be independently reconstructed from the published details.
RuntimeWire recently covered Moonshot's subscription and GPU capacity strain; the new event here is Sycamore's valuation estimate and Bloomberg's account of how K3 is changing investor assumptions.
What the $314 billion estimate measures
Bloomberg framed K3 as a market shock that made investors revisit assumptions behind the global technology rally and prompted comparisons with DeepSeek's market impact. Moonshot says K3 can compete with leading OpenAI and Anthropic systems at a fraction of their cost, the claim at the center of Sycamore's valuation call.
The size of that call is striking beside Moonshot's own valuation. TechCrunch reported in May that the Beijing company raised about $2 billion at a $20 billion valuation. Sycamore's estimated reduction for OpenAI and Anthropic is roughly 15.7 times that figure.
That comparison shows how much value investors have assigned to the revenue and margins of leading proprietary model providers. If lower-priced open-weight systems become credible substitutes, customers gain bargaining power and closed-model vendors may face pressure on usage prices. The estimate does not establish that such a shift has occurred, and Bloomberg's report does not identify transactions that would validate the revised marks.
Moonshot's own disclosures leave important parts of the competitive case unresolved. Moonshot says Kimi K3 is a 2.8-trillion-parameter mixture-of-experts model with a 1-million-token context window, while acknowledging that its overall performance still trails Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol. Moonshot has promised to release the full weights by July 27, which would allow outside developers to inspect and deploy the complete model.
Yang, Moonshot's co-founder and chief executive, studied computer science at Tsinghua University and earned a doctorate from Carnegie Mellon University before working with Google Brain and Facebook AI Research. He co-authored Transformer-XL and XLNet and later returned to China to build a startup, according to The Wire China's profile. He founded Moonshot in 2023 with Zhang Yutao, Zhou Xinyu and Wu Yuxin.
K3 now gives Yang a chance to test whether Moonshot can convert technical claims into durable pricing pressure on much larger US rivals. Sycamore has attached a $314 billion figure to that possibility. Without disclosed methodology, independent K3 evaluations or observable private-market transactions, it remains a market analyst's estimate rather than a verified change in the value of OpenAI and Anthropic.