BEIJING — The pitch that PandaAI’s founders have been making to Chinese investors is simple: the tools that made AI useful for writing and coding can be aimed at trading. The company, which builds infrastructure for AI-driven trading, said it has closed seed, angel and angel-plus rounds in quick succession, raising a combined sum in the tens of millions of yuan, with the two angel rounds led by L2F, a venture firm backed by founders.
The money will go into a specific stack, according to the company: an AI trading large language model, a library of professional quantitative skills it calls QuantSkills, infrastructure for multi-agent collaboration, and a development environment for building trading agents. PandaAI said the funding will accelerate work on its PandaAIOS platform, its ADE agent development environment, and a protocol it calls A2A designed to let AI agents coordinate trading tasks.
The company is aiming at a market that has grown crowded with startups selling signals, chatbots and strategy tools, but it argues the bottleneck is different. “Most retail traders do not need another indicator,” the company said in its announcement. “They need infrastructure that lets AI models handle the research, the execution and the risk checks, and do it together.”
That framing — AI as the operating layer of trading rather than a feature bolted onto a terminal — has drawn a widening circle of backers in China, where deep-tech venture funds have rotated from chips to models to applications over the past two years. L2F, whose name reflects its origin as a fund run by founders, led both angel rounds; the company did not disclose which investors led the seed round or the size of each individual raise.
The product roadmap is ambitious for a startup that only recently emerged from stealth. PandaAIOS is described as the company’s trading operating system, an environment where models, data feeds and execution systems run together; ADE is the developer toolkit for building trading agents; and A2A is the company’s answer to a question the industry has been circling — how independent AI agents coordinate with each other and with human traders without cascading errors.
QuantSkills, the proprietary library, is meant to encode the judgment that professional quant teams spend years accumulating — position sizing, risk limits, execution logic — into reusable components that models can call at runtime. The company said it has been testing the stack with professional investment teams and financial institutions, alongside individual traders, and will use the new capital to expand beyond its home market.
China’s trading-technology scene has been fertile ground for such ambitions. Retail investors dominate volumes in Chinese equities and futures, creating demand for tools that automate research and execution; institutions, meanwhile, are under pressure to cut costs and standardize workflows. Analysts said the addressable base is large, but so is the regulatory surface: securities regulators in China and elsewhere have been tightening rules around algorithmic and AI-assisted trading, and startups in the space must design for compliance from day one.
PandaAI said its agents are built to operate inside regulatory guardrails, with audit trails and kill switches that let human traders override automated decisions. Whether that is enough to satisfy regulators in the company’s expansion markets remains to be seen, but the design choice reflects a lesson from earlier waves of algorithmic trading startups, several of which ran into trouble when their systems behaved in ways their own engineers could not explain.
The funding round also signals where Chinese venture capital is heading in 2026. After a stretch dominated by models and infrastructure, money is flowing back toward applications that put AI to work in specific industries. Trading is a natural target: the domain has clear metrics, existing data, and users who pay for tools that improve returns or cut risk.
The company’s ambition runs larger than a trading terminal. Its founders have described a future in which every trader, from a retail day trader in Shanghai to a portfolio manager in Singapore, works alongside a team of AI agents — one scanning news, one modeling risk, one watching execution quality. PandaAI wants to be the platform those agents run on.
The competitive field is filling in around PandaAI’s chosen lane. Global platforms have begun bundling AI agents into trading terminals, while a crop of Chinese startups sells strategy bots and signal generators to retail investors. PandaAI’s bet is that the durable value sits lower in the stack, in the infrastructure that lets many agents work together with auditability and control. Founders and early investors in Chinese fintech said that argument has gained credibility as earlier application-layer startups found their features easy to copy.
The company also benefits from a regulatory environment that has moved from hostility toward algorithmic trading to a more structured embrace. China’s exchanges have built out systems to monitor and govern automated activity, and the country’s AI adoption in finance has proceeded behind explicit policy encouragement. Analysts said the opening is real but conditional: regulators will tolerate AI-assisted trading that is explainable, logged and interruptible, and PandaAI’s architecture is designed with those three properties in mind.
For now, the three completed rounds buy time and headcount. The company said the funds will support hiring, product development and the push into overseas markets. In a sector where startups routinely die from underfunding or overpromising, PandaAI has done something rare: it has raised the money before making the promises.
