Author: Omnitools
In May 2026, it was revealed that Dark Side of the Moon had restructured its organization from B2B to B2C, and its head was hospitalized due to high intensity and cost pressures. Around the same time, the company completed a new round of financing of approximately $2 billion, pushing its valuation above $20 billion. On one hand, it was burning through cash to the point where people couldn't keep up; on the other hand, capital was lining up to enter the market.
This is not an isolated case. CCTV Finance data shows that in the first quarter of 2026, there were nearly 600 financing deals in China's AI sector, totaling over 110 billion yuan, a surge of 185.4% year-on-year. In May alone, two major model companies, Lunar Dark Side and Leap Star, raised over 30 billion yuan. The embodied intelligence sector is equally hot, with over 50 financing deals disclosed in the first quarter, totaling approximately 20 billion yuan.
When a single quarter's financing exceeds 100 billion yuan, and the financing amount of leading companies exceeds dozens of times their annual revenue, in what direction is this money pushing China's AI competition? Based on publicly available financing data, OmniTools has compiled a list of the domestic AI projects that received the most funding in the first quarter of 2026, attempting to answer three questions: Who is taking the money, what are they doing, and what does the trend behind this mean?
Where did the money go? Nearly 600 financing rounds, computing power consumed 30% to 50% of them.
Let's first look at the macro picture. CCTV Finance, citing data from venture capital firms, reported that in the first quarter of 2026, the total financing amount in China's AI sector exceeded 110 billion yuan, with nearly 600 financing events. What does 110 billion yuan mean? It's nearly three times the amount in the same period of 2025. Even in the history of global technology financing, this is a figure worth recording.
The money mainly went into two sectors. Large model companies took the lion's share; in May alone, Lunar Dark Side and Leap Star raised over 30 billion yuan. Embodied Intelligence followed closely behind, disclosing over 50 financing deals in the first quarter, with over 30 companies receiving investments totaling approximately 20 billion yuan, representing a year-on-year increase of nearly 60%.
What deserves even more attention is where the money goes. A CCTV Finance report clearly pointed out three main flows: R&D, computing power, and talent. Leading large-scale model companies typically invest billions of yuan in R&D by 2025; GPU procurement and cloud service leasing account for 30% to 50% of their financing; and attracting top global talent is the third largest expenditure.
The fact that 30% to 50% of funding is going towards computing power is key to understanding this round of financing. It means that out of every billion yuan raised, 300 to 500 million will directly become hardware and cloud service bills. This isn't waste; it's an entry ticket at the current stage of technology. Training a generation of large models requires a cluster of tens of thousands of GPUs, and inference services require a wide coverage of computing nodes—these are not things that can be solved with a light asset approach. The essence of financing is to quickly transform capital into hardware barriers.
At the same time, the figure of billions of yuan in R&D investment needs to be viewed in context: most large-scale companies will have revenues far below this level by 2025. A model with annual revenue of hundreds of millions and R&D investment of billions means that financing is the only lifeblood for maintaining competitiveness.
Large-scale racing track: Three players with billions of dollars in assets, each with a different strategy.
In this round of financing, the three most eye-catching large-scale model companies are Dark Side of the Moon, LeapStar, and DeepSeek. They all have valuations that have exceeded or are approaching tens of billions of dollars, but their strategies are completely different.
The Dark Side of the Moon: Open Source Approach Accelerates Commercialization
The Dark Side of the Moon is the company with the most intensive fundraising in the past six months. According to reports from Securities Times and Investment Circle, from December 2025 to May 2026, the Dark Side of the Moon completed multiple rounds of financing, raising a total of over US$3.9 billion (approximately RMB 37.6 billion), and its valuation soared from US$4.3 billion to US$20 billion. In just six months, its valuation nearly quintupled.
Supporting this valuation is the commercial performance of Kimi K2.5 after its release. According to data cited by the investment community, K2.5's revenue in just 20 days exceeded the total revenue for the entire year of 2025, and it rose from outside the top 100 in the global payments rankings to 9th place. This means that the commercialization of Kimi K2.5 has not just "improved," but has experienced a turning point leap.
K2.5's open-source strategy was a key variable driving this inflection point. This model became the only Chinese open-source model integrated into Cursor and the official main model of OpenClaw, directly penetrating the overseas developer ecosystem. Marc Andreessen, co-founder of Silicon Valley venture capital firm a16z, commented that it "basically replicates GPT-5-level inference capabilities," and investor Chamath Palihapitiya called it the "Kimi K2.5 moment"—the first time an open-source model truly shook the closed-source system.
Leap Forward to Stars: A Trillion-Dollar Base, Betting on Terminal Deployment
Jieyue Xingchen is taking a different path. In January 2026, the company completed a B+ round of financing exceeding 5 billion yuan, setting a record for the largest single financing round in the large-scale model sector within 12 months at the time. In May, a new round of financing of nearly 2.5 billion US dollars will be completed.
Unlike the open-source approach of Lunar Dark Side, Step-2 Star emphasizes the parameter scale of its foundation model and its terminal deployment. Its trillion-parameter Step-2 model is positioned as an enterprise-grade foundation, with multimodal capabilities covering scenarios such as visual understanding and video generation. On the product side, Step 3.5 Flash topped the OpenRouter fastest model list on its first day of release.
More noteworthy is who is investing in Jieyue Xingchen. In its May funding round, companies across the mobile phone supply chain, including Huaqin, Longcheer, OmniVision, and ZTE, collectively participated. This is no coincidence. Yin Qi, co-founder of Megvii, has become chairman of Jieyue Xingchen, and the company has dismantled its red-chip structure, aiming for a Hong Kong IPO. The entry of capital from the mobile phone supply chain signifies that Jieyue Xingchen's strategy is not to be a pure software company, but rather to embed AI capabilities into terminal devices, pursuing a fusion of "model + hardware."
DeepSeek: From No Funding to a $45 Billion Valuation
DeepSeek's first round of financing is one of the most significant events of 2026. According to reports from Securities Times and Caijing.com, DeepSeek has launched its first round of financing, with its valuation soaring to $45 billion and a planned fundraising amount exceeding 50 billion yuan. The National Integrated Circuit Industry Investment Fund (Big Fund) is in talks to lead the investment.
Prior to this, DeepSeek had never raised external funding. Founder Liang Wenfeng personally invested 20 billion yuan to support the company's operations. DeepSeek is known for its open-source models and extreme cost-effectiveness. Its technology roadmap has had a significant downward pressure on API prices across the industry, pushing mainstream commercial models into the "cent-level era."
The signal that "even companies that don't need money are starting to raise funds" is more significant than any funding figure. It indicates that the intensity of the current AI competition has reached a critical point: even with technological and cost advantages, failing to raise funds could lead to falling behind in the industry-wide arms race for computing power and talent. The National Integrated Circuit Industry Investment Fund's (Big Fund) negotiations to lead the investment signify that state-owned capital has officially entered the core of large-scale models. A valuation of $45 billion corresponds to companies that have not yet achieved large-scale commercialization, and the pressure to digest this valuation is self-evident.
Embodied Intelligence: More than 10 companies valued at over 10 billion – is it a hot trend or a bubble?
Embodied intelligence was another major theme in Q1 financing. Data from ITjuzi and Securities Times shows that over 50 financing deals were disclosed in this sector in Q1, totaling approximately 20 billion yuan, representing a year-on-year increase of nearly 60%. Statistics from Guangzhou Daily show that more than 10 companies have already exceeded 10 billion yuan in valuation.
Galaxy General is the highest-valued project in this sector . In March 2026, the company completed a 2.5 billion yuan financing round, valuing the company at over 20 billion yuan. Investors included the third phase of the National Integrated Circuit Industry Investment Fund (Big Fund III), Sinopec, and CITIC. This marked the first investment by Big Fund III in the embodied intelligence field. Galaxy General's core products are general-purpose humanoid robots and the "Galaxy Star Brain" embodied large-scale model system. Its robots collaborate with CATL to achieve 24/7 fully autonomous operation in battery factories.
Qianxun Intelligent has completed two rounds of financing totaling nearly 2 billion yuan , and its valuation exceeded 10 billion yuan within 26 months of its establishment. Investors include Yunfeng Capital, Chaos Capital, and Sequoia Capital. Qianxun's humanoid avatar intelligent production line is already in operation at CATL's Zhongzhou base.
Independent Variable Robotics has completed a RMB 1 billion Series A++ funding round and a strategic round of several hundred million RMB , valuing the company at over RMB 10 billion. It is the only embodied intelligence company to have simultaneously received investment from three internet giants: Alibaba, Meituan, and ByteDance. Founder Wang Qian is one of the world's earliest scholars to introduce attention mechanisms into neural networks. Independent Variable Robotics follows a full-stack self-developed software and hardware approach and has already launched the "Quantum-1" and "Quantum-2" humanoid robot bodies.
Zhijian Power is the most unique project in this round of financing . Founded only in July 2025, the company raised 2 billion yuan in five rounds of financing within eight months, achieving a valuation of over $1 billion, making it the youngest unicorn in the smart device sector. Sequoia Capital, Tencent, and Alibaba all participated in the investment. A valuation exceeding $1 billion in less than a year is extremely rare in the hardware industry.
Lingxin Qiaoshou focuses on perfecting a single product category . The company has completed a nearly 1.5 billion yuan Series B financing round, with a valuation exceeding 10 billion yuan. It is the only manufacturer in the world capable of producing thousands of high-degree-of-freedom dexterous hands per month, holding over 80% of the global market share. Achieving global leadership in a niche hardware category, Lingxin Qiaoshou's path differs from that of humanoid robot companies that manufacture complete systems; it is more like a "hidden champion" in the embodied intelligence industry chain.
VitaPower and Lumin Robotics represent two directions: consumer-grade and industrial capital . VitaPower's Pre-A round of financing reached nearly 500 million yuan, setting a record for the largest single financing in the consumer-grade embodied smart device sector. The first batch of 500 Vbot super robot dogs has already been delivered. Lumin Robotics has raised nearly 1 billion yuan in total financing, led by Mitsubishi Electric. Its founder is a former executive of Dreame, and it expects revenue to reach the 100 million yuan level by 2026.
The embodied intelligence sector is indeed churning out billion-dollar unicorns. However, only a handful of companies have disclosed their revenue. While leading companies like Unitree Robotics reported revenue of 1.708 billion yuan in 2025 and Leju Intelligent reported 258 million yuan and applied for an IPO, demonstrating genuine revenue, most companies with valuations in the hundreds of billions are still in the early validation stage. The gap between valuation and revenue is the biggest uncertainty. When more than ten companies simultaneously surpass 10 billion yuan in valuation, the pressure of a valuation bubble in the industry cannot be ignored. Ultimately, the high expectations of capital need to be realized by visiting customers and checking production lines.
Behind the Ledger: Five Emerging Trends
Based on the above financing data and project review, OmniTools has identified five emerging trends.
Trend 1: Capital structure is shifting from VC-dominated to a mix of industrial capital and state-owned enterprises. The National Integrated Circuit Industry Investment Fund (Big Fund) Phase III invested in Galaxy General Electric, the Big Fund is in talks to lead DeepSeek's Series A funding round, Mitsubishi Electric led Lumin Robotics, and mobile phone supply chain capital invested in Jieyue Xingchen. Industrial capital is entering with orders and application scenarios, while state-owned capital is entering with strategic intentions; the proportion of purely financial investors is decreasing. This means that companies are not only expected to receive investment but also responsible for industrial implementation and compliance.
Trend Two: Valuation ceilings are rising rapidly, and the Matthew effect is intensifying. With $20 billion in investment from The Dark Side of the Moon, $45 billion from DeepSeek, and 21 billion yuan from Galaxy General, leading companies have absorbed the majority of the funding. Meanwhile, over 30 companies in the intelligent technology sector have received investment, but only about 10 have raised more than 1 billion yuan. Funds are concentrated in the hands of top players, and the risk of exiting smaller projects is increasing.
Trend Three: The IPO channel is opening at an accelerated pace. Leju Intelligent's IPO application on the ChiNext board has been accepted, with a planned fundraising of 2.6 billion yuan, making it the first AI company to meet the fourth set of standards for the ChiNext board. Jieyue Xingchen is dismantling its red-chip structure and aiming for a Hong Kong listing, while Zhipu and MiniMax have also initiated their listing processes. The high valuations in the primary market require secondary market support, and 2026 to 2027 may become a concentrated window for Chinese AI companies to go public.
Trend Four: Open Source Approach Gains Differentiated Competitive Advantage. Kimi K2.5 penetrated the overseas developer ecosystem through open source, while DeepSeek lowered industry-wide prices with its open source strategy. Open source is no longer a "free strategy," but a means to gain access to the global developer community and build ecosystem barriers. For downstream customers, this means more cost-effective models to choose from, and API costs continue to decline.
Trend Five: Embodied Intelligence Moves from Proof-of-Concept to Small-Batch Delivery. VitaPower's Vbot delivered its first batch of 500 units, Galaxy General Robots is in actual operation at CATL, and Unitree Robotics plans to ship 5,500 humanoid robots by 2025. No longer just PowerPoint presentations, some products are already in use at customer sites. However, the transition from "small-batch" to "large-scale" production presents significant hurdles, including supply chain, quality control, after-sales service, and cost control.
The 110 billion yuan in funding raised in the first quarter of 2026 is pushing China's AI competition into a new phase. This phase is no longer characterized by "who has the largest model parameters," but rather "who has the most money to sustain them until commercialization." Billions in R&D investment, computing power accounting for 30-50% of total revenue, and valuations reaching tens of billions are emerging in large numbers. These figures are both proof of the industry's vitality and a signal of a high-intensity war of attrition. When the pace of fundraising becomes a core competitive advantage, and GPU reserves determine the speed of model iteration, the final outcome of this round of AI startups may not be determined by a single technological breakthrough, but rather by cash flow and management capabilities—who will remain at the table.
