Moonshot AI accelerates IPO plans amid surging demand for Kimi K3
Moonshot AI, the Beijing-based startup behind the popular Kimi chatbot, is seeking shareholder approval for a Hong Kong initial public offering within the next six months. The move comes as the company closes a funding round that could lift its valuation above $30 billion. The IPO push follows the release of Kimi K3, an open-weight large language model that has rattled both chip stocks and cryptocurrency markets.
According to sources familiar with the matter, Moonshot AI has engaged several investment banks to prepare for the listing, which could be one of the largest tech IPOs in Hong Kong this year. The company is expected to file a preliminary prospectus with the Hong Kong Stock Exchange within weeks. The valuation target of $30 billion-plus represents a significant jump from its previous round, reflecting the market's enthusiasm for the company's rapid growth and technological advancements.
Kimi K3 drives subscription boom and supply constraints
Demand for Moonshot’s new Kimi K3 open-weight model has surged since its launch. The model, which is available as an open-weight release, allows developers and enterprises to download, modify, and deploy the AI on their own infrastructure. This strategy has resonated strongly with the global developer community, positioning Moonshot as a serious competitor to both U.S. giants like OpenAI and Meta, as well as Chinese peers like Alibaba and Baidu.
The strong uptake has pushed Moonshot AI’s annual recurring revenue (ARR) to an estimated $300 million, according to industry analysts. The subscription service, which offers tiered access to Kimi K3’s API and premium features, has seen such explosive demand that the company was forced to temporarily pause new subscriptions while it scales up its cloud infrastructure. The pause is expected to last several weeks, during which Moonshot will add thousands of GPUs to its data centers.
This subscription pause underscores the unprecedented demand for advanced AI models. It also highlights the infrastructure bottleneck that many AI startups face: even with billions in funding, securing enough high-end chips like NVIDIA H100s and B200s remains a challenge. Moonshot has reportedly secured long-term contracts with cloud providers and chip suppliers to ensure capacity, but the current spike has overwhelmed its systems.
Alibaba's Qwen3.8 adds to the competitive pressure
On the same front, Alibaba Cloud released its own open-weight model, Qwen3.8, which has further intensified competition in the AI space. Qwen3.8 is a 385-billion-parameter mixture-of-experts model that rivals the performance of GPT-4 and Llama 3.1. By making it fully open-weight, Alibaba aims to capture the developer ecosystem that has historically been dominated by Western models.
The release of Qwen3.8 has two significant implications. First, it validates the open-weight approach as a viable business model, forcing U.S. AI providers to reconsider their closed-source strategies. Second, it demonstrates that Chinese AI firms are catching up faster than many anticipated, narrowing the gap on both performance and cost efficiency. This has led to a wave of adoption among enterprises in Asia, the Middle East, and Africa, where data sovereignty concerns favor locally deployable open-weight models.
Market impact: Bitcoin and chip stocks feel the tremors
The combined effect of the Kimi K3 and Qwen3.8 releases has sent shockwaves through financial markets. Bitcoin, which has been trading near $65,000, saw a sharp but temporary dip as traders rebalanced portfolios in response to the AI news. Analysts attribute this to the growing correlation between AI and crypto markets, driven by the overlapping investment narratives around compute infrastructure, GPUs, and data centers.
Semiconductor stocks, particularly those tied to the AI investment cycle, experienced heightened volatility. Shares of NVIDIA, Advanced Micro Devices, and TSMC all moved lower on the news, as investors assessed the competitive threat from Chinese AI models. The concern is that if open-weight models from China can match or exceed Western models at a fraction of the cost, the demand for premium chips might not grow as fast as previously priced in. This has led to a rotation out of AI and semiconductor exchange-traded funds and into safer havens.
However, some analysts argue that the selloff is overblown. “More AI models mean more competition, which in turn drives down costs and expands the total addressable market. In the long run, lower inference costs will spur adoption, creating even greater demand for compute,” said a senior analyst at a global investment bank. The crypto market, too, has largely stabilized, with bitcoin holding above $64,000 and ether above $3,400.
Background: Moonshot AI’s rise and the Kimi ecosystem
Moonshot AI was founded in 2023 by a group of former researchers from DeepMind and Google Brain. The company quickly gained attention with its first model, Kimi K1, which excelled at long-context reasoning. The Kimi chatbot, built on top of the model, became a viral hit in China and parts of Southeast Asia, amassing over 100 million users within its first year.
The company’s open-weight strategy was initially viewed as a risky gamble. Unlike OpenAI’s GPT family, which remains proprietary, open-weight models allow anyone to download the weights and run the AI locally. This approach has historically struggled with monetization, but Moonshot has managed to turn it into a competitive advantage by offering a premium cloud service, enterprise support, and fine-tuning tools. The subscription revenue model has proven highly effective, with conversion rates from free to paid users exceeding 15%.
Kimi K3 builds on this legacy. It features a 1.5 trillion parameter architecture with a mixture-of-experts design, enabling it to activate only a fraction of its parameters per inference. This makes it both powerful and cost-efficient. Benchmarks show Kimi K3 outperforming GPT-4 on several key tasks, including code generation, mathematical reasoning, and multilingual understanding. The model supports over 100 languages and has a context window of 2 million tokens, making it ideal for processing long documents, legal contracts, and scientific papers.
Strategic implications for the AI industry
The simultaneous rise of Kimi K3 and Qwen3.8 signals a shift in the global AI landscape. Until recently, the narrative was dominated by a handful of U.S. companies: OpenAI, Google, Anthropic, and Meta. China’s AI ecosystem was seen as fragmented and heavily regulated, with models that struggled to compete on quality. That perception is now changing.
Both Moonshot and Alibaba have invested heavily in architecture innovations that reduce computational overhead. Their open-weight approach also bypasses export controls on advanced chips, as the models can be run on a mix of domestic and imported hardware. This pragmatic strategy has attracted significant interest from the developer community, which values transparency and control.
For U.S. AI vendors, this means they can no longer rely solely on their model’s performance to justify premium pricing. They must also offer compelling reasons to use a closed platform, such as integrated safety features, seamless cloud integration, or exclusive datasets. The battle lines are being redrawn: open-weight vs. closed, performance vs. price, sovereignty vs. centralization.
Bitcoin and the AI-Crypto connection
The impact on bitcoin highlights an often-underappreciated dimension: the deep interdependence between AI and crypto markets. Both industries depend on large-scale compute infrastructure, often sharing the same GPU clusters and data centers. Miners and AI companies compete for the same hardware, driving up costs and creating supply bottlenecks.
When an AI model breakthrough alters the demand dynamics for compute, it inevitably affects the cost of mining bitcoin. More efficient AI models can run on less powerful chips, potentially freeing up GPUs for mining. Conversely, if AI demand surges, it pulls hardware away from mining, reducing the network hash rate and, in theory, making bitcoin mining less profitable in the short term.
This relationship is still evolving, but market participants are beginning to price it in. The recent volatility in both bitcoin and semiconductor stocks suggests that the market is treating them as part of the same investment theme. As AI becomes more democratized through open-weight models, this correlation may strengthen, creating new opportunities and risks for investors.
What’s next for Moonshot?
With the IPO process underway, Moonshot AI is racing to capitalize on its momentum. The company plans to use the proceeds from both the funding round and the IPO to expand its data center capacity, hire top talent, and accelerate research into even more powerful models. It is also exploring partnerships with major cloud providers to offer Kimi K3 as a managed service outside of China.
The next few months will be critical. If Moonshot can successfully navigate the subscription pause and manage the scaling challenges, it could become one of the most successful AI startups to go public. The IPO will also serve as a bellwether for investor sentiment toward Chinese tech companies, especially those operating in sensitive sectors like AI.
Meanwhile, the broader AI landscape will continue to evolve rapidly. Alibaba’s Qwen team has already hinted at a Qwen3.9 update with enhanced multimodal capabilities. OpenAI is reportedly working on GPT-5. Meta is pushing forward with Llama 4. The competitive pressure is driving innovation at an unprecedented pace, and the ultimate beneficiaries will be consumers and businesses who gain access to ever more capable AI at lower cost.
In the short term, bitcoin and chip stocks may remain sensitive to AI news. But the long-term story is one of expansion: more models, more use cases, and more demand for compute. Companies like Moonshot and Alibaba are proving that open-weight AI can both challenge the status quo and generate substantial business value.
Source:Coindesk News