The First Year of Commercialization for AI Unicorns: The Rise of a New Generation of Entrepreneurial Organizations
Source: Jiemian News Image Library
Core view: Entering 2025, the focus of the AI venture capital and innovation ecosystem is shifting from technological hype to commercialization. AI unicorns are starting to prove their sustainable revenue models. The development and maturity of the business models of AI Agents and "AI-native" unicorns have provided possibilities for new corporate forms and startup models.
The AI industry leads the growth of the global unicorn ecosystem scale
From 2024 to 2025, the financing scale of global AI startups has shown exponential growth: Among the 54 companies with valuations exceeding $1 billion that have emerged so far in 2025, more than half (57%) are AI companies. Capital is pouring into this track crazily. Almost one out of every two venture capital investments is flowing to AI startups. The financing amount of the AI industry in the first half of 2025 alone has exceeded the total for the whole of 2024 (Figure 1).
Figure 1: The financing amount of the AI industry in the first half of 2025 has exceeded that of 2024. Source: CB Insights
Early AI investments mainly focused on the empowerment logic of "AI + industry". Investors were concerned about how to use AI to transform existing business processes. However, after 2024, the investment logic underwent a fundamental change. Capital began to chase the new value that only AI can create - for example, ThinkinMachinesLab (co-founded by former OpenAI Chief Technology Officer Mira Murati) completed a $2 billion seed round of financing at a valuation of $12 billion without launching any products, which is almost unimaginable in traditional venture capital logic (Table 1).
Table 1: The top five new AI unicorns in 2025. Source: Pitchbook
The rise of super unicorns is the most intuitive manifestation of the centralization of this round of AI investment. After completing huge financings one after another in the past year, four AI companies (large model developers OpenAI, Anthropic, and xAI, and data intelligence and AI platform service provider Databricks) are already among the top ten unicorns in the world by valuation. The core value of these companies lies in their mastery of computing power, algorithms, and models. They represent the highest pricing in the current market for the potential of AGI (Artificial General Intelligence) (Table 2).
Table 2: The top five companies in terms of financing scale in the AI field in 2025. Source: Compiled from public information
These market trends reflect the market's expectation of a "paradigm breakthrough" in AI technology. Investors are no longer betting on incremental improvements but on exponential leaps - from weak artificial intelligence to strong artificial intelligence, from task execution to autonomous decision-making, and from tool assistance to intelligent collaboration. The logic behind the high valuations is that once the technological breakthrough is achieved, the market space will expand non-linearly.
In terms of commercialization level, currently, there are about 15 AI companies in the world with an ARR (Annual Recurring Revenue) exceeding $100 million. Three companies have an ARR exceeding $1 billion: large model companies OpenAI ($10 billion), Anthropic ($4 billion), and AI data annotation company ScaleAI ($1.5 billion). AI companies with an ARR between $50 million and $1 billion are basically various AI applications.
AI Agents and AI-native enterprises are gradually realizing their commercialization capabilities
In 2025, the venture capital hotspots in the AI industry have spread across the platform layer and application layer, especially AI Agents, which have spawned disruptive products and experiences.
An AI Agent is a system based on a large language model (LLM) that aims to independently execute tasks on behalf of users through reasoning, planning, and interaction with external tools.
In less than a year, the number of enterprises in the AI Agent field has grown from about 300 to thousands. From e-commerce to industry, Agents are gradually being integrated into the workflows of various vertical industries. And every leap in the underlying model capabilities directly translates into a stepwise increase in the ARR of AI-native startups.
The product value propositions of these startups are completely based on AI capabilities. Cursor's code completion, Harvey's legal document generation, and Abridge's medical record automation - the core functions of these products could not be realized in the era without AI. The core value of their products increases with the improvement of the model's performance, rather than operational efficiency.
For example, in September 2024, legal AI startup Harvey announced that OpenAI's o1 inference model, combined with domain-specific knowledge and data, enabled it to build a legal Agent. The company raised $300 million in financing at a valuation of $3 billion in February 2025. Its sales team has doubled in size in the past six months, and it has reached the revenue threshold of $100 million (Table 3).
(Optional) Image note: Cursor is an AI Agent launched by Anysphere. Source: CB Insights description
Among the new unicorns in 2025, one-fifth are building AI Agents. These unicorns have shown amazing growth rates. Anysphere reached a valuation of $9.9 billion in just three years since its establishment, and the ARR of its product Cursor has reached $500 million. Lovable reached an ARR of $100 million and a valuation of $3.5 billion in just two years. These figures would have taken 7 - 10 years to achieve in the traditional SaaS era.
Behind the rapid growth is the extreme compression of the product development cycle by AI technology. Traditional software companies require a large number of engineers, a long development cycle, and repeated testing and iteration. However, AI-native enterprises can quickly build prototypes using basic models, automate a large amount of development work through generative AI, and achieve natural growth through intelligent products. This "AI-accelerated AI entrepreneurship" forms a positive flywheel: faster product iteration → better user experience → faster revenue growth → more capital support → stronger technology investment.
In terms of the commercialization model, AI services are shifting from early software subscriptions to result-oriented payment. For AI Agent functions that can completely autonomously execute complex tasks, payment may be made on demand based on the task success rate or calculation duration. The flexibility of this model can better serve customers of different sizes and needs.
When an AI Agent can autonomously complete high-value tasks (such as autonomously generating legal documents and optimizing complex supply chains), the charge will be based on the quality of the results delivered and the business impact, rather than just the usage duration or the number of users. This result-oriented business model solves the problem that the traditional software subscription model cannot match the non-linear value provided by AI.
"Digital colleagues" are moving from concept to reality
AI Agent startups raised $3.8 billion in financing in 2024 (almost three times the total in 2023). All leading large model developers among large technology companies are developing general AI Agents or providing tools for them. More autonomous AI Agents will have a profound impact on enterprises, from changing the employee structure (establishing new hybrid teams composed of humans and AI Agents) to maximizing operational efficiency through full automation of daily tasks.
Since most enterprises tend to choose mature suppliers, large technology companies have significant advantages in AI Agent development. Similarly, enterprise software giants such as Salesforce (Agentforce) and ServiceNow (AI Agent Marketplace) have launched Agent platforms for their existing customer base.
However, startups at all levels of the technology stack are establishing their market positions by solving specific technical challenges and breaking through the boundaries of Agent capabilities. Smaller, specialized participants also have many opportunities. In the AI era, the depth of specialization and the choice of niche are equally important.
Take the AI + office unicorn Glean, which has an ARR of $100 million, as an example. Its core products include Copilot products such as Glean Assistant, Glean Agents, and Glean Search. By deeply understanding the enterprise data structure and permission system, Glean has built a "Google inside the enterprise". Different from general large models such as ChatGPT, Glean realizes AI search within the enterprise based on the integration of enterprise data + RAG (Retrieval Augmented Generation) technology.
Looking to the future, it is worth paying attention to how Agents will appear in a more breakthrough form. Early signs of this are reflected in "AI-native" products - these tools and platforms are built around AI capabilities from the very beginning, rather than adding AI functions to traditional products.
General AI assistants and enterprise workflow automation: Horizontal AI Agent startups currently account for the highest proportion in the entire Agent market landscape. This segment mainly includes startups targeting enterprises, providing general applications across industries, spanning different enterprise systems (including ERP, CRM, HRM, etc.), and covering various work functions such as human resources/recruitment, marketing, and security operations. Companies focusing on productivity and personal assistants, including OpenAI and its Operator Agent, are directly targeting consumers and employees.
The AI Agent fields with the strongest development momentum and the most intense competition are customer service and software development (including coding and code review and testing Agents). Agents can bring great value to well-defined workflows. Taking software development as an example, the capabilities of Agents have evolved from code assistance tools to being able to handle the entire process of autonomous software development from requirements analysis, architecture design to deployment monitoring.
Enterprise workflow automation focuses on tasks that are highly repetitive, clearly defined, but still require manual processing. From invoice processing to customer service, from inventory management to compliance review, AI Agents are taking over these "unloved but necessary" tasks. These Agents can not only execute single tasks but also understand the business logic across systems and achieve end-to-end process automation.
Creative and development assistance tools: AI programming Agents are far ahead in terms of commercialization. Six software development Agents rank among the top, including market leaders such as Anysphere's Cursor (ARR $500 million) and Replit (ARR $150 million). According to statistics from venture capital data institution CB Insights, customer service AI Agents have the highest valuation premium, with an average of 219 times the revenue multiple. This valuation gap reflects investors' confidence in this track and the expectation that enterprises will quickly replace human teams with AI.
The penetration rate of AI + programming scenarios is high. Programmers in domestic and foreign technology companies generally use AI programming tools to improve quality and efficiency. A standout example is Cursor, the product of Anysphere, a new unicorn in 2025. Cursor has strong context management capabilities. Its excellent Tab key completion function can accurately predict the user's next edit and provide multi-line code suggestions at once, enhancing the immersive code creation experience and being widely praised by users.
Vertical domain Agents: Vertical domain Agents targeting specific industries are constantly emerging. Startups are carving out niche markets by solving the problems of customers in specific industries, especially in highly regulated and data-sensitive fields, such as Harvey mentioned above. Harvey's success in the legal field proves that when AI masters domain knowledge, understands industry workflows, and can generate outputs that meet professional standards, it can take on some of the work of legal assistants and even junior lawyers. AI startups in the medical and financial fields are replicating this model by deeply cultivating "industry data + compliance framework + tool capabilities" to cover the entire workflow.
The reliability of AI Agents remains a major challenge in this field. Once an Agent malfunctions, hallucinates, or behaves abnormally, it will immediately bring business risks. As the capabilities of artificial intelligence improve, we expect more startups to make progress in terms of autonomy. Improvements in reasoning and memory capabilities will bring more complex decision-making, adaptability, and task execution capabilities.
AI's innovation in the organizational form and development model of startups
As the capabilities of basic models improve, Agents are expected to become more autonomous - evolving from static task execution to more adaptable, reasoning-driven systems that support dynamic decision-making. With the continuous development of AI Agents, entrepreneurs will find that generative AI makes entrepreneurship cheap and convenient - anyone can become an entrepreneur, just like anyone can become a self-media blogger. AI concentrates information density and processing capabilities in the hands of a very small number of key talents, endowing individual entrepreneurs with the productivity equivalent of a medium-sized organization. This not only significantly reduces the capital requirements of startups but also exponentially speeds up organizational decision-making and product iteration.
The core business logic of AI-native enterprises lies in building organizational leverage through "human-machine hybrid" teams: a small number of experts collaborate with large models/Agents to complete R & D, customer service, operations, etc. that previously required a large amount of human resources. Once the basic models and Agent architectures are fully developed, their marginal costs will drop rapidly. Each iteration of the model may bring non-linear performance improvements to the product without the need to linearly increase human resources or operational investment like traditional enterprises. Their value growth depends on the capture and utilization of proprietary data, rather than traditional channels or the scale of human resources.
The development of AI Agents and digital employees will drive a new wave of "AI equality" - the technical background of entrepreneurs may not be the most important. What is important is to discover problems that can be solved by AI, which requires entrepreneurs to have a deep understanding of the capabilities and limitations of AI and the ability to build products that are superior and more professional than general models.
On the other hand, although generative AI brings promising prospects for entrepreneurs, it also makes it easier to quickly copy ideas. To gain a foothold in the AI entrepreneurship ecosystem, entrepreneurs need to put more effort into building unique competitive advantages, such as solving problems that general chatbots like ChatGPT and DeepSeek cannot solve, rather than just providing marginal efficiency improvements.
Authors: Teng Binsheng, Professor of Strategy at Cheung Kong Graduate School of Business, Deputy Dean of Strategy Research, and Director of the Global Ecosystem Research Center for New Generation Unicorns / He Jianshi, Researcher at the Global Ecosystem Research Center for New Generation Unicorns at Cheung Kong Graduate School of Business
This article is from "Jiemian News". Republished by 36Kr with permission. Journalists: Teng Binsheng, He Jianshi