When Binance founder Changpeng ‘CZ’ Zhao tweets, markets listen.
This time, it wasn’t about Bitcoin or regulation — it was about an AI that’s out-trading humans.
His comment came after DeepSeek, a rapidly rising Chinese AI firm, reportedly topped an AI trading challenge, outperforming other popular models like Gemini and Claude — and sparking a wave of discussion about whether AI can out-think or even move financial markets by itself.
What is DeepSeek, and why is everyone talking about it?
DeepSeek is a Hangzhou-based AI company founded in 2023 by former hedge fund executive Liang Wenfeng and backed by quantitative investment firm High-Flyer Quant.
According to Investopedia, DeepSeek’s large-language models — including DeepSeek-R1 and V3 — are designed to deliver top-tier reasoning and code generation performance at a fraction of the cost of Western rivals like OpenAI or Anthropic.
While the company is primarily an AI lab, its technology has recently been used in live crypto trading, where models autonomously generate and execute trading decisions.
In one public “AI trading arena,” six AI models were each allocated $10,000 to trade crypto in real time.
According to 99Bitcoins, DeepSeek’s bot achieved roughly 40% gains, outperforming competitors that included Gemini and other LLM-based bots.
How AI trading works — and why CZ is skeptical
Traditional trading wisdom says profits depend on uniqueness: your edge must be something no one else has.
Once a strategy becomes popular, it stops working. As CZ put it, “Otherwise, you are just buying and selling at the same time as others.”
But AI may be changing that rule.
Instead of one secret formula, DeepSeek’s system processes massive datasets — from market prices and order flow to news sentiment — and uses reinforcement learning to make rapid decisions.
According to PANewsLab, DeepSeek’s models can generate trading signals, risk parameters, and even execution code in seconds, retraining continuously based on new data. This creates a paradox: if enough people adopt the same AI, its collective trading could move prices in its favor — what CZ calls the “crowd effect.”
In other words, the AI’s predictions could become self-fulfilling simply through scale.
Why DeepSeek might be winning right now
Analysts suggest several reasons why DeepSeek has stood out:
- First-mover advantage: Fewer traders are using it, leaving more exploitable inefficiencies.
- Speed and data breadth: Its models can process more market variables in real time than most manual or rule-based bots.
- Open innovation: DeepSeek encourages fine-tuning and community optimization, allowing fast iteration on trading logic.
- Leverage: In one reported setup, DeepSeek used 10–15× leverage in its crypto trades (Tech.News.am).
Together, that can magnify both profits and risk — and may explain why its performance looks dramatic in short-term challenges.
The risks of AI herd trading
CZ’s question goes to the heart of the problem: what happens if everyone trades like the same AI?
- Crowded trades: If too many bots buy or sell the same asset at once, liquidity can vanish instantly.
- Flash crashes: A sudden shift in sentiment could cause a cascade of AI-triggered sell orders.
- Edge erosion: Once a strategy becomes public or replicated, profitability declines rapidly.
- Adversarial dynamics: Other traders may learn to exploit predictable AI-driven patterns.
As Bitget Research noted, the next frontier of AI trading will likely shift from model access to data exclusivity — the quality and uniqueness of the data that feeds these AIs.
Bigger picture: AI is becoming the market
Business Insider reported that even U.S. tech stocks — including Nvidia — took a hit as investors feared Chinese AI firms could disrupt the global AI landscape.
In finance, this points to a new kind of competition: not between humans, but between algorithms — where whoever adapts fastest wins.
DeepSeek’s trading results may be impressive, but CZ’s skepticism is warranted. AI can analyze faster and execute instantly, but the market still rewards originality. If everyone uses the same machine to trade, the machine becomes the market — and that can cut both ways.
For now, DeepSeek has the spotlight — but whether it stays profitable once the crowd catches on remains the trillion-dollar question.
