Original author: Frank, PANews
AI Agent has become an unavoidable hotspot in current on-chain trading, from ai16z to Virtual, and then to Swarms, within just a month, the AI Agent track has once again derived a new sub-ecosystem in the MEME field. Faced with the constantly evolving AI Agent tokens, which ones can break through the encirclement, and which are just fleeting concepts? Perhaps there are multiple angles to consider, but the flow of funds on the chain and the changes in the main forces may still be the most important indicators to consider.
PANews has taken the recently popular Swarms token as the main analysis object, and compared it with the large holder addresses of six high market capitalization AI Agent tokens, attempting to once again 'carve a boat to seek a sword' and glean some insights. The data range for this analysis includes: the initial buying and selling situation of the top 1000 holding addresses of the Swarms token (data time cut off at 24:00 on January 6, 2025), as well as the address overlap situation of six AI-related tokens with market capitalizations exceeding 100 million USD, such as Fartcoin, GRIFFAIN, ZEREBRO, ai16z, arc, and Swarms (data time cut off at 14:00 on January 7, 2025), along with the trading records of the internal market.
Some lay low and quietly lay out at the bottom, while others follow the trend and enter the market.
First of all, from the timeline of when the large holders entered the market, most large holders started to enter after January 2, which was 12 days after the token was created. From a temporal perspective, many large holders of Swarms only began to buy after the Swarms ecosystem became popular, failing to complete the early layout.
However, from the price curve of Swarms, if purchased before December 27, the price generally remained below $0.02, with nearly a 30 times increase from the current highest price of $0.6. Analyzing the initial purchase prices of these addresses, 202 addresses bought within the price range of $0.01 to $0.05, followed by the addresses that bought within the range of $0.3 to $0.4 being the most numerous.
The distribution of these two data points means that the early layout holders of Swarms bought in batches at low prices during the price drop period, and this buying was also relatively dispersed, not concentrated in a unified time frame. The benefit of this is that they can acquire chips at lower prices. Meanwhile, another group of large holders started to enter significantly after the discussion heat of Swarms increased, but the holding prices of these large holders do not have much competitive advantage.
This kind of chip distribution may explain the seemingly large short-term fluctuations in Swarms. If large holders who laid low early sell at high points, the new large holders have a relatively high cost, and once a significant sell-off occurs, it is easy to trigger sensitive nerves on both sides, leading to a sharp drop.
However, from the perspective of chip distribution, the main chips of Swarms are relatively dispersed. In the analysis of the top 1000 holding addresses, there are not many tokens sourced from the same address; most addresses’ initial tokens are mainly from on-chain exchanges. Therefore, there are few cases where early large holders obtained a large amount of chips and then dispersed them to multiple addresses.
Additionally, through the comparison of trading addresses on the internal market, it was found that the addresses that purchased early on the internal market have basically not appeared among the current top 1000 holding addresses. Therefore, the early chips of this token have basically completed their rotation.
From the overall data, the average first buying price of Swarms tokens is $0.17, and the average first selling price is $0.23. The average initial purchase amount per address reached $37,600, with the average initial selling amount being about $28,200. In comparing the buying and selling situation of individual addresses, the average first selling price of these addresses is about 2.43 times the buying price.
The largest holder has made a profit of $25 million without selling.
Compared to other MEME tokens, the above-mentioned average initial purchase amount is significantly higher, primarily influenced by some large holder addresses. The address with the highest initial transfer amount is Dsjzh2oj3HxyPefjQr5qqvbR5NrMnvBgptGLSQ3t8T5i, which transferred approximately $4.13 million from another address on December 31 and subsequently transferred about $500,000, with the current holding value being $27.33 million.
The address that transferred in, 5HfrnyodRraAw63aRVPueD5Er4D1sRKMZBMx9LBbhUAs, began large-scale purchases as early as December 20 at 8:22 AM. It continued to buy in, spending a total of $1.89 million to acquire 54.95 million Swarms at an average price of about $0.034, currently realizing a profit of approximately $25.44 million.
According to the tracking, the earliest associated purchase from this address began at 7:13 AM (the opening time of Swarms was around December 20 at 6:45 AM). It is worth mentioning that the associated funding address for this address began purchasing ai16z tokens as early as October 27, achieving a profit multiple of about 36 times.
Additionally, another address, 5NQTp9jHbzS4N9yKMWxwm8pPZW3RFSFPze3Edwss7iLe, transferred in approximately $3.63 million worth of Swarms tokens on January 4. According to on-chain traces, this address also made dispersed purchases through several addresses around January 2 and ultimately consolidated the tokens into one address, with the current holding value of this address being approximately $5.26 million.
There is another address, H1zFMUjYLzJwcfgXEtwiJ2ykvxmBr7JW6afW29PkcEAe, which also used a similar method, holding approximately $2.27 million, but the initial source of tokens for this address came from the Bitget exchange, followed by multiple purchases on-chain.
The combined initial transfer amount of these three addresses reached approximately $10.53 million. Looking at the purchasing process, the initial strategy involved multiple addresses making dispersed purchases, and after the heat of Swarms increased, all tokens were aggregated into a few addresses, becoming smart money in the eyes of on-chain hunters.
27% of addresses purchase multiple AI Agents; who is driving AI Agents behind the scenes?
In addition to analyzing the token addresses of Swarms, PANews also conducted comparative analysis on the top 1000 holding addresses of six addresses, including Fartcoin, GRIFFAIN, ZEREBRO, ai16z, arc, and swarms. The analysis shows that among the 6000 addresses participating in the analysis, 1647 addresses appeared repeatedly, meaning that about 27% of the addresses purchased multiple AI Agent-related tokens, among which ZEREBRO seems to be the favorite token of AI large holders, with 405 addresses purchasing this token. The second one is arc (368 addresses) and ai16z (334 addresses).
Among these addresses, the address with the highest holdings, DJnHztNmw1H56uYm98PNu5eVZ5yhi9482rZ9zA22TUUz, currently holds AI-themed tokens worth approximately $49.86 million, with the amount held in ai16z alone being around $42.70 million. Moreover, this is not the entirety of the holdings of this address; as early as a month ago, this address had profited tens of millions of dollars by acquiring tokens like ZEREBRO and GRIFFAIN.
Additionally, 3xzTSh7KSFsnhzVvuGWXMmA3xaA89gCCM1MSS1Ga6ka6 also holds about $42.84 million worth of AI-related tokens, with this address's on-chain holding value exceeding $73 million. According to social media information, this address should be an early AI Agent address known as Truth Terminal.
In addition, there are many similar addresses. Among the statistics, the 1647 large holder addresses hold AI-related tokens worth over $1.58 billion, with about 29 addresses holding amounts exceeding $10 million, and these 29 addresses hold amounts of about $690 million.
Rather than saying that AI Agent may be the hottest trend in 2025, it is better to say that AI Agent is essentially a better story material in the eyes of large fund investors.
Analyzing trading behavior is more important than tracking smart money addresses.
As the analysis of on-chain data deepens, tracking smart money seems to have become a popular study. However, from the perspective of large holders, when laying out early chips, they do not wish for too many retail investors to enter the market and seize the bottom-priced chips. Therefore, continuously changing new wallets and making dispersed purchases have become the basic operations of large manipulators.
As a result, blindly chasing smart money will gradually become ineffective and may lead to being maliciously harvested. However, through multiple analyses of the operations of large holders, even using new addresses and making dispersed purchases, there are still difficulties in management and capital aggregation. Therefore, in most cases, large holders still need to consolidate the funds from various wallets into one or a few wallet addresses for easier management and can stimulate more following users to enter the market by making small purchases during peak periods. Secondly, to quickly collect and buy early on, these early layout players had to make large purchases within a certain time frame, although the amounts were relatively dispersed, this regular pattern of buying may still become a signal. After all, their investment amounts often range from hundreds of thousands to millions of dollars, and without a certain determination to operate, they generally would not make crazy purchases.
In summary, for general retail investors, if they have to chase smart money through on-chain tracking, it may be much more effective to pay attention to on-chain behaviors than to chase smart money addresses. Of course, an important premise is to think like large holders about what type of themes will make a good story; otherwise, in the face of endless new tokens, blind chasing is akin to searching for a needle in a haystack.
