While other AI models were engaged in a "Spring Festival red envelope battle," Kimi made a significant announcement quietly. According to a February 17 report, Moonshot AI's Kimi is close to finalizing a new funding round exceeding $7 billion, just over a month after completing a previous $5 billion financing. This round was jointly led by existing shareholders, including Alibaba, with participation from Tencent. Additionally, the company has initiated another funding round targeting a valuation between $100 and $120 billion.

Based on these figures, Kimi's latest valuation has effectively doubled, surpassing the $10 billion mark. The consecutive funding rounds, totaling over $12 billion, represent the largest amount raised in the large language model sector in the past year.

Just over a month earlier, on December 31, 2025, Moonshot AI's founder, Yang Zhilin, disclosed in an internal letter that the company had successfully closed a Series C funding round of $5 billion, significantly exceeding its target. He noted the company's cash reserves had surpassed 100 billion yuan.

Despite the industry accelerating and a wave of IPOs emerging towards the end of 2025, Moonshot AI explicitly stated it was in no rush to enter the public markets. Yang Zhilin stated in the letter that the company believed it could raise larger amounts of capital from the private market compared to public offerings. He highlighted that the funds raised in their Series B and C rounds already exceeded the proceeds from most IPOs and secondary offerings by listed companies, indicating no immediate pressure to list.

For Kimi, 2025 was a year of continuous technological advancement. Yang Zhilin began his internal message by congratulating the team on achieving "SOTA results." He wrote that the releases of K2 and K2 Thinking marked significant progress on the path to AGI, citing technical milestones such as "China's first trillion-parameter foundation model."

Within Moonshot AI's strategy, technological breakthroughs and product iterations are paving the way for commercialization. Since May, Kimi has frequently launched new Agent functionalities. Yang Zhilin revealed that from September to November, the number of paying users, both domestic and international, saw an average month-over-month growth exceeding 170%. The launch of K2 Thinking also significantly boosted API revenue, with overseas API income increasing fourfold during the same period.

However, the substantial cash reserve of 100 billion yuan is seen as fuel for an even more costly competition ahead. Yang Zhilin clarified that the Series C funds would be used to "aggressively expand GPU resources" to accelerate the training and R&D of the K3 model. Part of the capital is also allocated for a 2026 incentive plan and an option repurchase program.

The internal letter also hinted at the intense competition for AI talent. Yang Zhilin stated that the company would finalize and distribute reward schemes for the K2 Thinking project and subsequent products before the 2026 Spring Festival. The plan is to increase the average employee incentive to 200% of the 2025 level and significantly raise the option repurchase quota.

The heat in the talent market confirms this trend. A report indicated that demand for AI-related positions in the first seven months of 2025 surged tenfold year-over-year, with a continued shortage of algorithm talent. By December, companies had further raised compensation packages to enhance their competitiveness in the talent war.

Looking ahead to 2026, Yang Zhilin outlined three strategic goals. First, the K3 model aims to improve equivalent FLOPs by at least an order of magnitude through technical enhancements and further scaling, aiming to match the pre-training levels of world-leading models. Second, the focus is on vertically integrating model training with agent product development to make K3 a more distinctive model, offering users novel capabilities not defined by other models. Third, the company will concentrate its product and commercialization efforts on agents, prioritizing the pursuit of higher intelligence and greater productivity value over sheer user numbers, targeting an order-of-magnitude increase in revenue scale.

Sustained and massive investment in computing power remains an inescapable cycle for large model companies, where superior models form the foundation of all narratives and commercialization. Despite holding significant capital, Moonshot AI still needs to demonstrate its ability to build a healthy, sustainable, and profitable business model.