Understanding Moonshot AI: China’s AI Challenger

While DeepSeek has become a rising star in the global AI scene, Moonshot AI remains a lesser-known name outside China. However, within the Chinese AI ecosystem, Moonshot AI has been a dominant force in the large language model (LLM) space, often compared to China’s OpenAI due to its cutting-edge research and rapid growth.

Founded by Tsinghua alumnus Yang Zhilin, Moonshot AI gained massive attention in 2024 with its consumer-facing chatbot, Kimi, which at one point held the highest valuation among AI-to-consumer (AI-to-C) companies in China. The company was heavily backed by top venture capital firms and internet giants and was once the industry’s undisputed leader.

But the landscape shifted. In 2025, Moonshot AI is no longer the hottest name in Chinese AI. The rise of DeepSeek and Doubao has disrupted the industry, forcing Moonshot AI to reassess its strategy.

Consistent Pursuit of SOTA Results Amid Uncertainty

Moonshot AI has identified “continuously obtaining State-of-the-Art (SOTA) results” as its top work priority. However, the company has not yet clarified if it will eventually connect with DeepSeek. When questioned about the possibility of open-sourcing its technology in the future, Moonshot AI remained silent as of press time. (For clarity, SOTA stands for State-of-the-Art, referring to the best technology or highest-level model available.)

According to firsthand information obtained by Huxiu, Moonshot AI has been relentlessly pre-training its models, a move that contradicts earlier rumors suggesting otherwise.

---

DeepSeek’s Meteoric Rise and Kimi’s Shifting Rank

DeepSeek’s breakthrough moment came when it overtook Doubao in daily active users (DAU) on January 28, 2025. In this evolving scenario, Kimi—Moonshot AI’s core AI product—has slipped to the third spot in DAU rankings among AI applications, trailing behind DeepSeek and Doubao. In early 2024, Kimi was lauded for topping both user numbers and growth rate charts.

User base sizes further underline these shifts. Data from Xsignal (January 10 to February 8, 2025) reveals:

• DeepSeek DAU: 119.79 million
• Doubao DAU: 25.05 million
• Kimi DAU: 6.88 million

Although Kimi continued to grow its user numbers during the Spring Festival, its growth rate did not match the explosive increases seen with Doubao and DeepSeek.

---

From AI Darling to Playing Catch-Up

At one time, Moonshot AI and its flagship product, Kimi, were hailed as the darlings of the AI circle in 2024. The company boasted the highest valuation among AI-to-consumer (AI-to-C) large-model companies, was often described as China’s equivalent to OpenAI, secured investments from top funds and internet giants, and built a compelling narrative around founder Yang Zhilin—a “post-90s Tsinghua genius” who rapidly scaled the startup to a valuation exceeding 10 billion in just one year.

However, three pivotal events shifted the narrative:

1. ByteDance-backed Doubao began rapid iterations and heavy traffic investments starting in May 2024, setting off a head-to-head competition with Kimi in the C-end market.
2. A brewing conflict with Jinshajiang Venture Capital escalated at the end of 2024, with well-known investor Zhu Xiaohu intensifying public criticism, dragging Moonshot AI into a whirlpool of negative public opinion.
3. The release of DeepSeek-R1 in early 2025 swiftly catapulted DeepSeek into the global spotlight, redefining competitive dynamics in the AI space.

These events have forced Moonshot AI to retreat from its earlier pedestal as the undisputed leader.

---

Divergent Paths: Open-Source vs. Closed-Source

At the heart of the rivalry between DeepSeek and Moonshot AI lie fundamental differences in their approaches:

• Moonshot AI remains closed-source and focuses heavily on pre-training—a process that demands extensive computing power and resources. It does not see distillation technology as central to its strategy and relies on significant traffic investments in the C-end market to drive new user acquisition.
• DeepSeek operates with an open-source model and also adopts pre-training but leverages distillation technology to reduce costs dramatically, without heavily investing in C-end traffic.

In essence, the three main contrasts are:

1. Open-source versus closed-source approaches.
2. Reliance on computing power versus cost-efficient techniques.
3. User acquisition strategies in the consumer market.

These differences compel a reassessment of Moonshot AI’s technical and user growth models.

---

Investor Concerns: Can Kimi Bridge the Gap?

An early investor shared with Huxiu that Moonshot AI now faces two critical questions:

1. Has DeepSeek truly unlocked new market opportunities in the C-end large-model product space?
2. Does Kimi need to modify its product or user strategy to remain competitive against Doubao and DeepSeek in 2025?

Another industry insider bluntly noted that Moonshot AI’s challenges are twofold:

• It cannot attract as much traffic as Doubao.
• Its technology and product mindset appear to be outpaced by DeepSeek.

The suggestion is clear: if Moonshot AI wants to maintain its niche, it may need to explore changes such as open-sourcing its models or rethinking its traffic investment strategy.

---

Financing, Resource Models, and the Investor Perspective

Moonshot AI has completed four rounds of financing since its inception. Its dependency on computing power has even led to a partial “computing power voucher” model—converting computing resources into equity—in its deals with major internet companies and during funding rounds.

In contrast:

• DeepSeek emerged from Huanhua Quantitative Investment, whose fund management scale exceeded 100 billion yuan in 2021. Notably, DeepSeek has not raised external funds so far.
• Doubao is a strategic project within ByteDance, which is anticipated to invest over 100 billion yuan in AI by 2025.

Despite the rising pressures, Moonshot AI remains optimistic about DeepSeek’s trajectory. Co-founder Zhang Yutong took to WeChat Moments and Xiaohongshu, celebrating the rapid development of AI and citing DeepSeek’s open-source model as a sign that AGI (Artificial General Intelligence) may soon be within reach. Some experts even argue that the wave of AI enthusiasm spurred by DeepSeek ultimately benefits Moonshot AI, especially when server issues at DeepSeek drive users back to Kimi.

---

Inside the Moonshot AI Headquarters

Located in Beijing’s Haidian District, the Moonshot AI headquarters exudes a unique blend of technical rigor and creative spirit. A piano placed at the entrance becomes a daily backdrop during lunch breaks, reflecting the company’s idealistic ethos—a stark contrast to competitors who opt for more conventional settings.

Founder Yang Zhilin is known for his intense focus on R&D. Often, he isolates himself for days or even weeks to tackle complex technical problems. The leadership team at Moonshot AI comprises five co-founders:

1. Yang Zhilin (CEO): Oversees technology research and development.
2. Zhang Yutao: Focuses on product development and was a co-founder in Yang’s previous venture, Circular Intelligence.
3. Zhou Xinyu and Wu Yuxin: Primarily responsible for technology and project execution, forming the core tech and product team alongside Zhang Yutao.
4. Zhang Yutong: Manages user growth, financing, and product-related tasks.

The company’s hiring process is notoriously rigorous—many candidates describe getting into Moonshot AI as a formidable challenge, with a preference for talents having experience in international and domestic tech companies and proven product success.

---

R&D Priorities and User Growth Strategy in 2025

At the beginning of 2025, Moonshot AI refocused on enhancing its model capabilities. Under Yang Zhilin’s leadership, R&D efforts have accelerated. The launches or updates of Vincent Image and Vincent Video in November and January, respectively, have cemented Moonshot AI’s multimodal capabilities. Notably, on January 20, Kimi released version k1.5 nearly simultaneous with DeepSeek R1’s launch.

For 2025, the company plans to further bolster its long-text reasoning capabilities alongside its established multimodal functions, underscoring a deep commitment to advancing model performance.

Previously, Moonshot AI had relied on traffic investment to drive user growth in the C-end market. While Huxiu reached out for confirmation on whether this strategy would continue into 2025, no definitive statement was provided. However, insiders suggest that the heavy traffic investments by both Doubao and Kimi in 2024 have significantly improved the domestic AI market’s perception. With retention rates above 50% for both companies, the numbers indicate that early promotion strategies were effective despite external skepticism.

Yet, a persistent challenge remains: profitability.

---

Revenue Challenges and the Road Ahead

Currently, Kimi generates a portion of its revenue through API access in the B-end market, but it hasn’t charged C-end users. Instead, it manages peak-time access through a reward model. Moonshot AI is yet to achieve a balance between its income and expenditures. After completing its B round of financing in August 2024—valuing the company at US$3.3 billion—no new funding has been announced.

Looking ahead, commercialization is a critical issue for Moonshot AI in 2025. The competition between DeepSeek, Doubao, and Kimi in the C-end space is intensifying. If Kimi is to narrow the user gap with DeepSeek, Moonshot AI must discover innovative methods for sustainable user growth or demonstrate strong user willingness to pay to attract further funds and computing support.

While there is some relief regarding funding pressures—with insiders noting that Moonshot AI’s financing situation isn’t dire and that discussions about an acquisition with Alibaba did not take place in 2024—the company still faces the dual challenge of scaling its user base and proving its long-term revenue model.

---

Moonshot AI stands at a crossroads, grappling with a rapidly evolving competitive landscape. Its journey from an industry darling to a company striving to adapt in the face of formidable rivals like DeepSeek is a compelling narrative of ambition, technological prowess, and the relentless pursuit of innovation.