Microsoft Corp., Amazon.com Inc., and Alphabet Inc. are in early talks with China's Moonshot AI regarding revenue-sharing agreements for its Kimi K3 model. These negotiations occur weeks after US Treasury Secretary Scott Bessent warned he might add the Chinese firm to a trade blacklist.

Moonshot AI is seeking up to 30% of the revenue generated by K3-related services hosted on Azure, Amazon Web Services, and Google Cloud. The discussions remain preliminary and may not result in finalized deals. Reuters reported that an agreement could mark the first major revenue-sharing deal between a Chinese AI company and a U.S. cloud provider.

Geopolitical Tensions and Accusations

The talks unfold against a backdrop of heightened friction over AI technology transfer. Treasury Secretary Bessent recently warned that Chinese AI developers could face sanctions and Entity List designations due to alleged industrial-scale distillation attacks on US models. He stated, "Open source is not open season on American IP."

US officials have accused Moonshot of using outputs from Anthropic's Fable model to build Kimi K3 and improperly acquiring restricted Nvidia Corp. chips. Moonshot denies these claims, asserting that K3's improvements stem from original changes to its underlying architecture. No sanctions have been imposed yet.

Commercial Drivers for US Cloud Providers

The interest from Microsoft, Amazon, and Google is driven by commercial opportunity. Moonshot charges $3 per million uncached input tokens and $15 per million output tokens, pricing well below many Western frontier models.

Kimi K3 features 2.8 trillion parameters, making self-hosting impractical for most customers due to computing costs. This creates a revenue stream for hyperscalers, who can sell access to the competitive model via their existing infrastructure while sharing revenue with Moonshot. Artificial Analysis rates K3 alongside OpenAI's GPT-5.5 and Anthropic's Claude Opus 4.8.

Market Sentiment and IPO Plans

Prediction-market traders on Polymarket assign Moonshot an 11% chance of holding the world's No. 1-ranked AI model by December 31, based on approximately $130,000 in trading volume. However, the firm has only a 1% chance of finishing 2026 with the best-ranked model, trailing Anthropic, xAI, Google, and OpenAI.

These cloud negotiations coincide with Moonshot's preparations for a potential Hong Kong IPO. Reports indicate the company is discussing a valuation approaching $50 billion, increasing pressure to convert K3's technical momentum into commercial revenue.

What the Numbers Show

The proposed 30% revenue share highlights Moonshot's confidence in Kimi K3's market fit despite geopolitical risks. By tying compensation to revenue rather than flat licensing fees, Moonshot aligns its earnings with adoption rates on US platforms. This structure suggests the firm believes the model's competitive pricing ($3/$15 per million tokens) and high parameter count (2.8 trillion) will drive significant usage among customers unable to host such large models independently.

Moonshot AI is reportedly beginning discussions on a final round of fundraising in August, targeting a valuation of nearly $50 billion before a planned listing in Hong Kong. The Chinese AI lab is likely to close a financing round that kickstarted this summer at a valuation of $31.5 billion, according to a report by Bloomberg on Tuesday. A separate report by the South China Morning Post, citing a deleted WeChat post by Moonshot AI’s financial adviser HF Capital, indicated the $50 billion valuation represents an over 11-fold surge from Moonshot’s valuation at the end of 2025 and a rapid jump from its $20 billion valuation after the May funding round.

Once the funding round is finalized, Moonshot plans to immediately start discussions with potential financiers about a follow-on capital-raising. This could be the final capital infusion for Moonshot ahead of its initial public offering, likely this year. Bloomberg earlier reported that Moonshot earned $300 million of annual recurring revenue in June, a rise from $200 million in April. Since the launch of its Kimi K3 model last week, daily sales have multiplied at least six times.

The report comes as Treasury Secretary Scott Bessent indicated that the Trump administration is set to investigate alleged intellectual property theft by Chinese artificial intelligence models and impose sanctions if required. Bessent termed this kind of IP theft "distillation," which refers to a method of AI training where a less competent model is developed using outputs from a more advanced model. Earlier reports suggested that the Trump administration is eyeing restrictions on advanced Chinese AI models, a move that could strengthen and safeguard the market domination of American AI firms like OpenAI and Anthropic.

Market Impact and Sector Shifts

Moonshot's Kimi K3 model has triggered specific demand for high bandwidth memory and faster networking solutions, creating clear winners in the semiconductor sector. Memory makers such as Micron Technology Inc, SK Hynix Inc, and Samsung Electronics Co Ltd are positioned to benefit, alongside networking firms like Marvell Technology Inc and optical component manufacturers including Corning Inc and Coherent Corp. Conversely, frontier model makers like OpenAI, Anthropic, and Alphabet Inc face new competition, while the outlook for GPU makers remains nuanced due to capacity constraints.

Regulatory Debate and Industry Response

Professional Capital Management CEO Anthony Pompliano argued on Monday that banning Chinese AI models would be counterproductive, suggesting American companies should instead focus on building superior technology. He stated on X that banning Chinese models is the "most un-American thing we could do." Former White House AI and crypto czar David Sacks also opposed using regulatory uncertainty to discourage adoption, arguing that U.S. companies should compete on merit. Meanwhile, CNBC commentator Jim Cramer expressed his skepticism towards Chinese tech companies and AI models, indicating that they might be a serious threat to the U.S. market if they are allowed to access U.S. corporate data.