Let me state the conclusion first: the essence of trading is a game based on information data analysis! The core is information asymmetry—if you know more than I do or can calculate faster than I can, you make money. Chess, stock trading, cryptocurrency trading, and sports betting all follow the same underlying logic. In the past, it was a battle between people; now, AI has entered the arena, but the essence hasn't changed—the intensity of the game has. The core of this transformation is that the asset management industry is undergoing a thorough 'blood change': the past engine was human experience, and the current engine is becoming computational power.
What can AI do to win? This 🉐 data speaks for itself!
At the beginning of 2026, the Aster platform held a 'human-machine battle' live competition. During two weeks of intense market fluctuations, the results were quite interesting: the human trader team overall lost 32.22%, and 43% of them were liquidated. In contrast, none of the 30 AI agents faced liquidation, and the overall loss was only 4.48%.
Where does this gap come from? It's not that AI can predict, but it doesn't make 'human' mistakes—no panic, no greed, no shaky hands. When the market crashes, people are scared and cut losses, while AI increases positions according to strategy. It's that simple.
Research from Guotai Junan Securities also confirms this: large language models can directly convert unstructured data like financial reports, policies, and news into quantifiable trading signals. While a human tires after reading 10 reports in a day, AI can scan 10,000 and still sift through social media.
Of course, AI is not a god; it also has two short legs.
Firstly, it only recognizes history and becomes confused when confronted with the unfamiliar. Pandemics, wars, sudden policy changes—when these black swans emerge, most AI systems fail.
Secondly, it understands correlation too well but not causation. AI knows 'when A rises, B follows,' but doesn't understand why. When new policies or geopolitical conflicts arise, it becomes clueless. Overfitting, weak interpretability, and failure in extreme scenarios are the three major challenges AI quantification must face. Just a note, there is still a difference between AI trading and AI quantification!
So AI and humans are in human-machine collaboration; it eliminates those manual traders who can only read K-lines and make decisions based on feelings.
As technology progresses, we envision that when AI starts competing against AI, it will be a battle of computing power.
If the market is full of AI trading, what will the competition look like? The answer is: a computing power arms race.
Just look at a few sets of data and you will understand. In February 2026, SpaceX and xAI completed a $1.25 trillion merger to send data centers into space powered by orbital solar energy. OpenAI is even more aggressive, signing a procurement agreement for over a trillion dollars of computing power at once, equivalent to the energy of 20 nuclear power plants. NVIDIA invested $100 billion in OpenAI, while AMD took a 10% stake in exchange for orders. This is not burning money but seizing the bottom cards. When strategies become similar and data is made public, the only differentiator is: who computes faster, who has better models, and who can access exclusive alternative data.
Bai Shuo from Hang Seng Electronics said directly: AI applications have entered an era of competition from standardization to differentiation. Guotai Junan Securities also pointed out that the competition among leading institutions has elevated to an 'AI-native' strategy, with the core being the construction of proprietary, trustworthy AI tech stacks that can manage complex systems.
Fourth, what is the role of humans? The answer lies in human-machine collaboration.
There is a detail in the Aster match: the human team overall lost 32%, but the champion, a living person, made a profit. What does this indicate? In an emotion-driven market, human judgment still has excess return potential.
The WEEX AI trading competition also proves this point: the champion team used a combination strategy of 'AI automated trading + manual precision control.' AI acts as the executor, and humans as the decision-makers—this is the currently viable path.
Hang Seng Electronics has made a very accurate judgment: 'Human-machine collaboration' is the better solution. AI is responsible for routine operations—seizing small opportunities, running high-frequency trades, and managing risk. Humans are responsible for two things: first, setting guardrails to prevent AI from going off the rails in extreme market conditions; second, managing those things that AI cannot understand—geopolitics, policy shifts, and human panic.
Guotai Junan Securities summarizes it well: the future is a collaborative process centered on 'human experts, with AI as the intelligent assistant.' AI evolves from being a supportive tool to the decision-making hub, but the final decision must still be made by humans.
The last sentence
The essence of trading has not changed; it is still an information game. It's just that now the information has been thoroughly digested by AI, and the game has turned into a computing power confrontation. The core engine of asset management is transitioning from human experience to AI driven by computing power. But humans should not compare speed with machines; if they do, they should compare—those things that machines do not understand: causation, intuition, and the pricing power of 'unknown unknowns.'
Grateful for the encounter! Grateful for reading and sharing!
$BNB $ASTER #特朗普新全球关税 #AI对抗AI就是算力博弈
