Billions of dollars a year, hard buy back, Pump.fun, why is the market still empty

Photo by Four Pilars

Original: AididiaoJP, Foresight News

 

Key points

Repurchases of $328 million were completed within eight months with 99.5 per cent of the daily agreed income, and the same conclusion was reached by the two independent data polymers, although not interoperable. To manipulate data, De Fililama has to be tricked into maintaining a stable ratio of 68-69 per cent of the Dune data indexed by Adam tech, supported by 10,517 million PUMPs in a verifiable wallet。

In August 2026, the “dilution curve” was actually a substitute for supply rather than accumulation, and repurchases could absorb twice the additional supply at current income levels. Community emissions end when the team and investors unlock the lock. Monthly emissions will drop from 10 billion to 9.2 billion。

The real reasons for the suppression of the current valuation multipliers are industry classification (the nature of the original crime unit), the basis of trust (anonymous team, discretionary buy-back) and the flow of funds (the use of buy-back sales by suspected insider traders)。

I. $328 million in refunds

The following analysis shows that the rumours of Pump.fun ' s income fraud have been spreading on Twitter。

As at 15 March 2026, fees.pump.fun data showed cumulative repurchases amounting to $328 million. This means using 22,83518 SOLs to purchase 10,4.5 billion PUMPs, or 10.45 per cent of the total supply, to offset 29,52 per cent of the flow of supplies. During eight months, the amount of the daily buy-back was between 99.5 per cent and 100.5 per cent of the agreed income, with an average of $1.25 million as at February 2026. Revenue counterfeiting is supported by large sums of money: each dollar repurchased, i.e., one dollar of SOL, which is transferred out of a certified wallet to a currency deposited at an auditable address. Forgery of income of $328 million required actual expenditure of $328 million。

The relevant tokens are stored in a chain for authentication (as of March 17, the wallet G8CcfRff held 103.96 billion PUMPs, 8PSmqJy6 held 121.1 billion PUMPs, for a total of 10.517 billion). After the initial execution of the wallet 3vkpy5Y (Solscan tag "Pump Buy Back"), the transfer to the holding wallet was completed and rotated in August 2025, with a current balance of zero。

DeFiLlama recorded income from the agreement of 15 July 2025 to 21 February 2026 totalling $300041880. During the same period, the cumulative repurchase amount was $300,178,162. The two match up to 100.05 per cent, with a total difference of $330 million between the two independent systems of $1.36 million。

The Dune dashboard of Adam tech provides third-level validation. The platform tracks only Solana chain revenue, which is stable at 68-69 per cent of DeFiLlama's multi-chain data, as it did not index Padre revenue in October 2025 on Base, Ether and BNB Chain. This ratio remains stable on a daily basis, indicating that both read the same chain of events independently。

In March 2025, before PumpSwap went online, three data sources had errors of between 1 and 5 per cent. When PumpSwap was launched, the data was divided into three layers: total cost, agreed income, Solana income only. The artificial falsification of income data requires the simultaneous deception of two separate chain indexers, the maintenance of a stable cross-cutting ratio in three product changes, the maintenance of a multi-chain income split in line with the actual expansion of operations and the support of the purchase of coins in a verifiable wallet。

II. Four Statistical Tests

In addition to chain evidence, 747 days of cost data are subject to four standard tests to verify financial data authenticity. While the individual test is not conclusive, the four test points to the same conclusion with a significant increase in credibility。

First, the cost-to-income ratio is the most difficult indicator to forge. Pump.fun collects fees from each joint curve transaction, but does not fully account for the agreed income, partly to LPs, creators and recommended awards. The data are centralized and the total cost-to-net income ratio is reduced from 1.0 to about 0.48, but it is not a gradual decline, but a sharp decline in three phases, each of which involves a change in products that should be recorded on the chain:

• On 20 March 2025, PumpSwap went online with the LP fee split mechanism, the ratio fell from 1.00 to 0.70 within two days
• On May 13, the author ' s income-sharing mechanism went online and the ratio fell from 0.69 to 0.56
• September 2-3, the Ascend project introduced a dynamic cost mechanism whereby stratification pricing allowed creators to receive up to 0.95 per cent of the cost of low market value tokens, leaving only 0.05 per cent, from 0.68 to 0.46

The forgery of this data requires a simultaneous simulation of costs and income series, which undergo three structural adjustments, with the daily ratio varying from 0.40 to 0.55 depending on the currency hierarchy. This complexity makes forgery difficult to achieve. The true picture is that product-overlays naturally lead to data changes, rather than structural breakpoints that match artificial construction with contractual deployment times。

A second test of continuity and numerical distribution characteristics is designed to determine whether the data present an artificial entry feature. It is difficult for humans to generate real random sequences, preferring to avoid long-term continuity, preferring integer numbers and inadvertently favouring specific numbers. Pump.fun data does not have this feature:

The maximum number of consecutive increases or decreases is 6 days, with an average duration of 1.92 days, which meets the natural process expectations with moderate momentum. The distribution of consecutive lengths decreases geometrically: 185 consecutive days, 111 consecutive days, 52 consecutive days, until 7 consecutive days and 6 consecutive days。

The closing figures for daily costs were almost evenly distributed between 0 and 9, with the respective figures accounting for 8.7 per cent-11.2 per cent. 88.8 per cent of non-integrity at the end of the day, and of the 743 non-zero dates, only 7 days end at 00 or 000。

Test weekend effects again. Pump.fun is a bulk platform where users issue tokens on trading days rather than on weekends. The average daily cost of working days was $2.14 million, and the weekends were $1.81 million, continuing a reduction of about 18 per cent, which appeared week after week in two years of data. The Mann-Whitney test shows a p value of 0.003, which is statistically significant. An artificially constructed data will have to deliberately keep weekends low, increasing the complexity of forgery and the risk of detection。

Finally, self-relevance is tested to measure the correlation between income today and income tomorrow. Pump.fun ' s lag self-relevance is 0.78, indicating that today ' s costs are 78 per cent related to yesterday; 0.65 after a week (with a delay of 7 days); and 0.57 after two weeks (14 days behind). This slow and smooth decline reflects the dynamic characteristics of organic platform activities: the concentration of active time, and the continuation of low peak periods. If daily income is generated randomly and the correlation is close to zero in the neighbouring days, data will leapfrog like noise rather than market-like flows. It is difficult to forge a single lag of high self-relevance, but at the same time it is almost impossible to falsify the overall decay structure (the lags are gradually decreasing in one way or another) and to maintain the weekend effect, the continuum and the true digital distribution。

Four independent tests, four consistent conclusions, and three data sources corroborate each other. Revenue data are credible。

Analysis of the residual valuation discount factor

THE FALSE CLAIMS OF INCOME ARE ONE OF THE REASONS WHY PUMP ' S CURRENT VALUATION IS UNDER PRESSURE. THIS POINT HAS BEEN CLARIFIED IN THE PRECEDING ANALYSIS. HOWEVER, TOKENS ARE STILL TRADED AT DISCOUNTS, AND OTHER DISINCENTIVES AND THEIR AUTHENTICITY NEED TO BE EXPLORED。

First, analyze the August team to unlock the lock. Community emissions of 10 billion a month will stop after 240 billion in July, just as the team and investors unlocked the lock, bringing the total to 9.2 billion a month. Monthly emissions fell from 10 billion to 9.2 billion and inflation fell by 8 per cent. Based on the current average daily income of $1.25 million, an average of $38 million per month was repurchased at $0.0021 per item price, absorbing twice the additional supply of approximately $19 million per month. Repurchases continued after August, and the ratio will be further improved。

Income has not declined either. For 14 months, the average monthly cost fluctuated between $2.3 million and $4.8 million per day: 49 per cent in July 2025, 94 per cent in August, 72 per cent in September and 45 per cent in January 2026. Overall, the return around an average of $2.5-3 million per day has stabilized at $64-7 million per week. The so-called "third to first quarter decline" is a partial conclusion based on selective selection of September peak data。

The remaining disincentives are as follows:

The price of the original crime share is the longest. Solidus Labs found that 98.6 per cent of the platform had a "carpet" feature, which had the desired effect: whatever the income, the agency's configuration would not include the "monetary casino" in the portfolio. This is a continuous structural factor that has nothing to do with the quality of income。

Source: Solidus Labs

The alleged sale of insiders is a real and immediate pressure. The Wallet 77DsB received 3.75 billion PUMPs in July 2025 from an address marked by Solscan as the "Digniary Trust Wallet" and allegedly cleared 8.02 million USDCs between 16 and 22 February 2026. The wallet GpCfm transferred 1210 million PUMPs (US$ 2.57 million) to Bitget during the same period. The third wallet deposited 1,557 million PUMPs ($3.54 million) in Bitget on March 6. Although no information was available to confirm actual ownership, at least $14 million was repurchased within 30 days at the same time as the agreement was repurchased at a market price of $0.002 to the exchange, as compared with the previous private wheel price of $0.004. This is questionable regardless of the identity of the wallet owner。

The level of trust is the hardest to price. The founders were anonymous (co-founder Dylan had a record of "spreading carpets" in 2017); repurchases were clearly "discretionable" ("pump.fun may modify or suspend the program at any time"); and Bubblemaps, who had referred to Hayden Davis as being associated with $50 million in private solicitations, was deleted because co-founder Alon called it "defamation". The link does exist, but attribution is disputed and unproven。

None of these factors relate to the fundamentals of the operation. The income is real, the data support it and the unlocking arrangement benefits the holder. The original crime unit label, the anonymous founder, and insider capital movements represent a discount on trust in the $1.25 million that can be generated in a day-to-day certification chain to absorb double the additional supply agreement. The discount on trust will eventually narrow and the income on this scale will not be permanently mismatched。