{"slug": "moonshot-ai-closes-round-at-31-5-billion-and-is-already-chasing-50-billion", "title": "Moonshot AI closes round at $31.5 billion and is already chasing $50 billion", "summary": "Moonshot AI, the Beijing-based lab behind the Kimi K3 model, closed a funding round at a $31.5 billion valuation in late July and is already negotiating a raise at up to $50 billion ahead of a planned Hong Kong IPO, according to Bloomberg. The company's open-weight model, Kimi K3, shipped on launch day inside Cursor, a coding tool being acquired by SpaceX for $60 billion, and now powers Western developer tools while undercutting US rivals on price. Moonshot's annual recurring revenue hit $300 million in June, up from $200 million in April, as Chinese open-weight models gain ground with US enterprises due to pricing gaps of up to 9x on token costs.", "body_md": "*The Chinese AI lab behind Kimi K3 just closed a major funding round and is already negotiating its next one, as its open-weight model quietly powers Western developer tools and undercuts US rivals on price.*\n\nMost US founders couldn't have told you Moonshot AI's name three weeks ago. They probably know it now. The Beijing-based lab closed a financing round at a $31.5 billion valuation in late July, and as Bloomberg reported on July 21, it is already in talks to raise again at up to $50 billion ahead of a planned Hong Kong IPO. For context, Moonshot was valued at $4.3 billion at the end of 2025. It crossed $20 billion in May. The speed of that climb isn't hype arithmetic; it's what happens when your model ships on launch day inside a product that SpaceX is acquiring for $60 billion.\n\nThat product is Cursor. On July 26, the coding tool shipped Kimi K3 on the day Moonshot released it, routed through US-hosted inference providers Together AI and Modal. Kimi K3 is the largest open-weight model ever trained, at 2.8 trillion parameters, with a one-million-token context window built for long-horizon coding and deep reasoning. Moonshot acknowledged it still trails Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on overall benchmarks, but it closed the gap by four points on key evals. More practically, US enterprise teams now have a compliance path to run it domestically, with no Chinese infrastructure in the loop.\n\nThe Cursor integration didn't happen overnight. Cursor acknowledged in March that it had used a Kimi model as a foundation when building its product. That history made the launch-day ship feel less like a partnership of convenience and more like the surface moment of something that had been true for a while. Kimi was already there. Developers are now finding out.\n\nMoonshot's backing reflects China's AI investment surge. Existing investors include Alibaba, Tencent, HongShan (the rebranded Sequoia China), IDG Capital, ZhenFund, and 5Y Capital. Its May Series D at $20 billion was led by Meituan's venture arm Long-Z Investments, with state-backed China Mobile and Tsinghua Capital also participating. The company's annual recurring revenue hit $300 million in June, up from $200 million just two months earlier in April. Those are real numbers, not projections.\n\nBefore the pre-IPO round begins in August, Moonshot is working to unwind its offshore red-chip corporate structure, a technical restructuring that clears the path for onshore Chinese capital and for a listing in Hong Kong, potentially before year-end. Yang Zhilin, the 33-year-old founder who studied at Carnegie Mellon and worked at Meta and Google, is moving fast on purpose. The window after a model breakthrough and before the next one closes faster than most people expect.\n\n## What this means for OpenAI and Anthropic pricing\n\nThe Kimi K3 launch fits a broader pattern that is already rattling American AI companies. As CNBC reported in July, Chinese open-weight models are gaining ground with US enterprises as OpenAI and Anthropic costs surge. The pricing gap is not marginal. Chinese models undercut their US counterparts by as much as 9x on token costs, and on some workloads the savings run to 90%. DeepSeek V4 Flash lists at $0.14 per million input tokens; GPT-5.2 lists at $1.75. Those aren't rounding errors.\n\nThe adoption numbers that flow from that are striking. OpenRouter data cited by CNBC shows the share of tokens consumed by US companies on Chinese models has sat above 30% every week since February 8, rising as high as 46% in some weeks. The twelve-month average before that was 11%. OpenAI is reportedly considering drastic token price reductions, a move that signals the company sees the Chinese pricing pressure as existential rather than peripheral.\n\nFrankly, the strategic situation for US labs is awkward. They built moats around proprietary closed weights, enterprise contracts, and the assumption that model quality would keep customers paying premium prices. Chinese labs like Moonshot, DeepSeek, and Zhipu are giving away top-tier weights for free or near-free while their quality closes on the frontier. The gap that used to justify OpenAI and Anthropic's pricing is compressing in public, on benchmarks anyone can read, week by week.\n\nNone of this means Moonshot wins. Kimi K3 still trails on overall performance, its IPO is months away and dependent on market conditions, and a $50 billion pre-money target is a big ask for a company that was worth less than $5 billion eighteen months ago. But a lab that reaches $300 million in ARR in under three years, ships the world's largest open-weight model, and lands inside the tools American developers use daily has already done the thing that most Chinese AI startups couldn't: made itself impossible to ignore on the other side of the Pacific.\n\n**Also read:** [Runlayer sues Rippling for allegedly cloning its MCP gateway after a year of engineering collaboration](https://startupfortune.com/runlayer-sues-rippling-for-allegedly-cloning-its-mcp-gateway-after-a-year-of-engineering-collaboration/) • [Micron has sold out its 2027 AI memory production and the cost is coming for everyone](https://startupfortune.com/micron-has-sold-out-its-2027-ai-memory-production-and-the-cost-is-coming-for-everyone/) • [OpenAI resets Sol usage limits and fixes the efficiency gap that caught power users off guard](https://startupfortune.com/openai-resets-sol-usage-limits-and-fixes-the-efficiency-gap-that-caught-power-users-off-guard/)", "url": "https://wpnews.pro/news/moonshot-ai-closes-round-at-31-5-billion-and-is-already-chasing-50-billion", "canonical_source": "https://startupfortune.com/moonshot-ai-closes-round-at-315-billion-and-is-already-chasing-50-billion/", "published_at": "2026-07-29 08:38:54+00:00", "updated_at": "2026-07-29 08:44:48.685969+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-startups", "ai-policy", "ai-products"], "entities": ["Moonshot AI", "Kimi K3", "Cursor", "SpaceX", "Alibaba", "Tencent", "HongShan", "OpenAI"], "alternates": {"html": "https://wpnews.pro/news/moonshot-ai-closes-round-at-31-5-billion-and-is-already-chasing-50-billion", "markdown": "https://wpnews.pro/news/moonshot-ai-closes-round-at-31-5-billion-and-is-already-chasing-50-billion.md", "text": "https://wpnews.pro/news/moonshot-ai-closes-round-at-31-5-billion-and-is-already-chasing-50-billion.txt", "jsonld": "https://wpnews.pro/news/moonshot-ai-closes-round-at-31-5-billion-and-is-already-chasing-50-billion.jsonld"}}