The Pump.fun Economy

Data freshness warning. Specific numbers cited in this article --- revenue figures, fee rates, graduation percentages, per-employee metrics --- are approximate and based on publicly available on-chain data and third-party analyses from late 2024 through early 2025. These figures change frequently as the protocol evolves, fee structures are updated, and market conditions shift. Treat all specific numbers as illustrative, not authoritative. Verify against current on-chain data before relying on them. The structural analysis and microstructure analogies below are the durable content.

Overview

Pump.fun is a platform on Solana (a high-throughput blockchain designed for fast, low-cost transactions) for launching memecoins — tokens with no intrinsic value proposition beyond social momentum and speculative demand. It has generated extraordinary revenue and become one of the most economically significant applications on the Solana blockchain. But the interesting question is not “how much money did pump.fun make” --- it is “who made money, who lost money, and what market structure produced that outcome.”

How a Token Launch Works: The Two-Contract Architecture

A pump.fun token lives through two distinct phases, each governed by a different smart contract — a program stored on a blockchain that executes automatically when its conditions are met — on the Solana blockchain.

Phase 1 — The bonding curve contract. When a creator launches a token, pump.fun deploys a smart contract that implements a virtual constant-product bonding curve — a deterministic pricing function where the token price increases automatically as more people buy. There is no order book and no counterparty: buyers send SOL (Solana’s native currency) to the contract, which mints new tokens at the current curve price and holds the SOL as reserves. Sellers burn tokens back to the contract and receive SOL from those reserves. This is the primary market — tokens are being created for the first time, and every buyer knows exactly what price they will pay because the pricing function is on-chain and deterministic.

Phase 2 — The AMM pool (secondary market). When enough SOL has accumulated in the bonding curve’s reserves (approximately 85 SOL, roughly $10K–$15K at typical prices), the token graduates. Graduation is an on-chain event where:

1. The bonding curve contract permanently closes — no more primary minting.
2. The accumulated SOL and remaining token supply are withdrawn from the bonding curve contract and deposited into a new constant-product AMM pool — a different kind of smart contract explained in the next section.
3. The token transitions from a single-contract primary market to a standard two-sided trading pool on PumpSwap (pump.fun’s own AMM, launched March 2025). Before PumpSwap, graduated tokens were deposited into Raydium V4 pools — Raydium being the largest third-party AMM on Solana at the time.

So yes, there are two distinct smart contracts involved in a token’s lifecycle: the bonding curve contract (primary issuance) and the AMM pool contract (secondary trading). Liquidity physically migrates from one to the other at graduation — the SOL locked in the first contract is transferred to the second in a single atomic transaction (one that either completes entirely or fails entirely — there is no partial migration). See bonding-curves for the full mathematics of the pricing function and PumpSwap for the economics of pump.fun’s vertical integration of both phases.

What Is an AMM?

An AMM (Automated Market Maker) replaces the traditional order book with a mathematical formula that determines exchange rates from token reserves. See constant-product-amm for the full derivation of the $xy=k$ invariant. LPs (Liquidity Providers) deposit paired tokens to fund the pool and earn trading fees; the risk is impermanent-loss.

Participants as Microstructure Agents

Pump.fun’s ecosystem maps cleanly onto the agent taxonomy from market microstructure — the study of how trading mechanisms, information asymmetries, and transaction costs determine price formation and participant outcomes. Each participant occupies a well-defined role.

Token Creators as Informed Traders

The Glosten-Milgrom model (1985) is a foundational market microstructure framework that explains how the presence of informed traders — those with private information about an asset’s value — forces market makers to widen their bid-ask spreads (the gap between the price at which you can buy and the price at which you can sell) as protection. In the model, an informed trader possesses private information that gives them an edge over the market. The adverse selection parameter $\mu$ (mu) is simply a number between 0 and 1 that measures the fraction of trading flow coming from informed agents. If $\mu =0.3$, it means 30% of incoming trades are from people who know something the market maker doesn’t. When $\mu$ is high, the market maker faces more “toxic” order flow and must charge wider spreads to compensate.

Pump.fun token creators hold a specific kind of informational advantage: they know the token’s marketing plan, whether a community will be cultivated, whether the project is intended to have any longevity, and crucially, their own exit timeline. Many creators are sophisticated repeat players who launch tokens, promote them on social media, buy early on the bonding curve (sometimes through multiple wallets), and sell into the demand they generate. Their informational edge is not about fundamental value (there is none in a typical memecoin) but about demand flow --- they know when the promotional push is coming because they control it.

In Glosten-Milgrom terms, $\mu$ is high: a disproportionate fraction of early trading comes from agents who know more than the marginal buyer.

Late Buyers as Uninformed Traders

Late-stage buyers --- those purchasing after initial promotion has peaked --- are the classic uninformed traders. They observe price momentum and social signals but lack the creator’s information about promotional plans and exit timing. In the market microstructure literature, these traders provide liquidity to the informed; colloquially, they are “exit liquidity.”

The expected P&L (Profit and Loss) for an uninformed buyer entering a pump.fun token after initial momentum is negative once you account for all costs:

Probability notation primer

The formula below uses standard probability notation. $E[\cdot ]$ denotes expected value — the probability-weighted average outcome (see expectation-variance-and-mgfs for the full treatment). $P(\text{event})$ denotes the probability of that event occurring, a number between 0 and

1. The formula reads: “the expected profit equals the purchase price times the probability-weighted average of gains and losses, minus all costs.”

$$E[\text{P\&L}]=P_{\text{buy}}\cdot \left(P(\text{further rise})\cdot E[R^{+}]+P(\text{decline})\cdot E[R^{-}]\right)-\text{fees}-\text{MEV loss}$$

where:

• $P(\text{further rise})$ and $P(\text{decline})$ are the probabilities of the token’s price going up or down after you buy
• $E[R^{+}]$ and $E[R^{-}]$ are the average return in each scenario
• fees include pump.fun’s trading fee, Solana transaction fees, and Jito tips (explained below)
• MEV loss (Maximal Extractable Value — profit extracted from blockchain transaction ordering) captures sandwich attack extraction — a form of front-running where a bot detects a pending trade, buys before it to push the price up, lets the victim trade at the worse price, then sells immediately after

For most late entrants, $E[\text{P\&L}]<0$ — the expected outcome is a loss after accounting for all costs. The fees and MEV extraction create a headwind that the token’s price would need to overcome just to break even.

The Protocol as Exchange Operator

Pump.fun is the exchange. It provides the matching engine (bonding curve + AMM), sets the fee schedule, and profits from volume regardless of participant outcomes. The protocol earns revenue from three sources:

$$\text{Revenue}=\text{Creation fees}+\text{Trading fees}\cdot \text{Volume}+\text{Graduation fees}$$

A creation fee is charged per token launch. Trading fees (1% on PumpSwap as of early 2025) are levied on every swap — every time someone buys or sells a token, 1% of the transaction goes to the protocol. Graduation fees are charged when tokens migrate from the bonding curve contract to the AMM pool.

Like a traditional exchange, pump.fun benefits from volatility and volume regardless of direction. The more tokens launched, the more trading, the more revenue --- whether or not any individual trader profits.

Liquidity Providers as Passive Market Makers

Once a token graduates to PumpSwap (or previously Raydium), it trades in a constant-product AMM pool. LPs deposit paired assets (SOL and the memecoin) to fund the pool and earn a share of every trading fee.

The catch is impermanent-loss: when a memecoin’s price collapses, the pool accumulates worthless tokens and loses SOL. For the full derivation see impermanent-loss; for break-even analysis see lp-profitability.

An LP’s net return on pump.fun tokens breaks down as:

$$R_{\text{LP}}=\text{Fee income}-\text{Impermanent loss}-\text{JIT extraction}$$

JIT (Just-In-Time) liquidity is a MEV strategy that further hurts passive LPs. Here is how it works: a bot observes a large pending swap (say, someone is about to buy $10K of a memecoin). The bot quickly deposits a large amount of liquidity into the pool just before the swap executes, capturing most of the trading fees from that swap. Immediately after, the bot withdraws its liquidity. The bot earned fees on a single large trade while bearing almost no impermanent loss (since it removed its liquidity before the price could move further). Meanwhile, the passive LP who had their money in the pool all along earned a smaller share of that fee.

For memecoins with short lifespans and severe price declines, impermanent loss dominates fee income, and most LPs lose money in net terms. See lp-profitability for the break-even analysis.

MEV Bots as Predatory Intermediaries

MEV bots are specialized programs that exploit their ability to control transaction ordering for profit. On a blockchain, transactions are not processed instantly — they sit in a queue (the mempool) waiting to be included in the next block. MEV (Maximal Extractable Value) bots monitor this queue and manipulate the order of transactions to extract value. In the pump.fun ecosystem, several types operate:

• Sandwich bots are the most predatory. The attack works in four steps: (1) the bot sees your pending buy order in the mempool, (2) it inserts its own buy before yours, pushing the price up, (3) your buy executes at the now-higher price, (4) the bot immediately sells, pocketing the difference. You paid more than you should have; the bot captured the difference. See sandwich-attacks for the formal mechanics.
• Arbitrage bots capture price discrepancies between the bonding curve and AMM pools — if the same token trades at different prices on the two contracts, the bot buys on the cheaper venue and sells on the more expensive one within the same block.
• JIT bots extract fee income from passive LPs as described above.

These actors are profitable in aggregate. Their returns come directly from other participants’ losses --- this is not value creation but value redistribution.

Validators as Toll Booth

Solana validators (the computers that produce blocks and confirm transactions) running the Jito client receive tips from MEV bots competing for transaction ordering. Jito Labs is a company that built MEV infrastructure for Solana — it provides a system where MEV bots can submit bundles (groups of transactions that must execute together in a specific order) and bid for inclusion in the next block. This is analogous to what Flashbots built for Ethereum: Flashbots is a research organization that created a structured auction system for Ethereum MEV, replacing chaotic on-chain gas bidding wars where bots would spam increasingly expensive transactions just to get priority.

In the pump.fun ecosystem, validators profit from the MEV supply chain without participating in it directly. Searchers (the MEV bots) pay tips to validators to guarantee their transaction bundles are included in blocks in the desired order. Validators are the passive beneficiary of ordering competition, analogous to how exchange operators in traditional finance profit from co-location fees (charges for placing trading servers physically close to the exchange’s matching engine to reduce latency) and market data sales.

Revenue and Economics

Reminder: these figures are approximate and likely outdated. Verify against current data.

By late 2024 / early 2025, pump.fun had reportedly generated cumulative revenue in the range of $400—600M, with operating margins estimated above 90%. The platform’s lean team (reportedly under 10 people at various points) implied revenue-per-employee figures that rivaled or exceeded major tech companies.

The margin structure reflects the protocol’s design: minimal infrastructure cost (Solana handles execution), no market-making obligations, no customer support in the traditional sense, and revenue that scales linearly with volume. This is the exchange operator’s dream --- pure intermediation with negligible marginal cost.

Graduation Economics

On the bonding curve, tokens trade with deterministic pricing — every buyer knows exactly what they will pay because the price is a mathematical function of the cumulative supply already minted. When the bonding curve’s SOL reserves reach the graduation threshold (~85 SOL), the protocol collects a migration fee and deploys a standard constant-product AMM pool on PumpSwap.

Reported graduation rates have been low --- estimates range from 1—3% of tokens launched --- meaning the vast majority of tokens never reach the secondary market. They are born, traded on the bonding curve, and die when interest fades and selling pressure drains the reserves.

For the protocol, even failed tokens generate creation fees and bonding curve trading fees. The graduation fee is bonus revenue on the minority that succeed.

Net Flow Analysis: A Negative-Sum Game

Consider the total flow of funds through the pump.fun ecosystem. If you add up every participant’s profit and loss, the total is always negative:

$$\sum _i{\text{P\&L}}_i=-\text{Protocol fees}-\text{MEV extraction}-\text{Validator tips}-\text{Gas costs (transaction fees)}$$

The ${\sum }_i$ notation means “sum over all participants $i$.” This equation is an accounting identity, not an estimate — every dollar of fees and extraction is a dollar that left the system. Pump.fun token trading is a negative-sum game: the total money out always exceeds the total money in, by exactly the amount extracted by fees and MEV.

Who wins and who loses:

Participant	Expected P&L	Source
Protocol (pump.fun)	Strongly positive	Fees on all activity
Token creators (sophisticated)	Positive	Informational edge, early entry
MEV bots	Positive	Systematic extraction
Validators	Positive	Jito tips
Token creators (naive)	Negative	Creation fee lost, token goes to zero
Late buyers	Strongly negative	Adverse selection, fees, MEV
Passive LPs	Negative	Impermanent loss exceeds fees

The distribution is not symmetric. A small number of actors (protocol, skilled creators, bots, validators) extract consistently, while a large number of retail participants lose. This is the structure of adverse selection: informed agents profit at the expense of uninformed agents, and the intermediary takes a cut from everyone.

The Transfer Equation

A simplified model of value transfer per token lifecycle — where does the money that retail traders lose actually go?

$$\text{Retail losses}=\text{Creator profit}+\text{Protocol fees}+\text{MEV extraction}+\text{LP impermanent loss}+\text{Validator tips}$$

Every dollar lost by a retail buyer ends up in one of these five buckets. The right side sums to a large fraction of total retail capital deployed. Empirical analyses of pump.fun trading data consistently show that the majority of wallets interacting with pump.fun tokens lose money on a per-wallet basis.

Structural Insights

Why This Structure Persists

The pump.fun economy persists because participants have heterogeneous motivations. Not everyone is maximizing expected financial return:

• Entertainment value: Many participants treat memecoin trading as gambling or entertainment, willingly accepting negative expected value for the variance and social engagement
• Survivorship bias: The small number of tokens that produce 100x+ returns are highly visible; the vast majority that go to zero are not
• Informational asymmetry about the game itself: Many participants do not understand the fee structure, MEV extraction, or the negative-sum nature of the market

Comparison to Traditional Market Structures

Feature	Pump.fun	Traditional Exchange
Listing standards	None	Extensive due diligence
Best execution duty	None	Legally required
Front-running	Profitable strategy	Illegal
Informed trading	Encouraged (creators)	Regulated (insider trading)
Fee transparency	On-chain, visible	Published fee schedules
Investor protection	None	Regulatory framework

The comparison is not to argue that pump.fun should be regulated like a stock exchange. It is to make visible the structural protections that traditional markets provide --- and what happens in their absence. Every protection removed creates an extraction opportunity for more sophisticated participants.

Socratic Questions

1. Pump.fun is a negative-sum game for traders, yet volume remains high. Does this invalidate the efficient market hypothesis (EMH — the theory that asset prices reflect all available information), or does EMH simply not apply when participants are not expected-value maximizers?
2. Token creators on pump.fun have a structural informational advantage. In traditional markets, this would be insider trading. Is there a principled distinction, or is the label purely regulatory?
3. If pump.fun’s margins exceed 90%, the protocol captures enormous surplus. In a competitive market, new entrants would compress margins. Why hasn’t this happened (or has it)?
4. The LP role on memecoin AMMs is almost certainly negative expected value (negative-EV) after impermanent loss. Why do LPs provide liquidity? Are they mispricing risk, or are there non-financial motivations?
5. MEV extraction on pump.fun tokens may exceed the platform’s own fees as a cost to traders. If pump.fun internalized MEV protection (e.g., private order flow, batch auctions), would it capture more volume? Is there a business incentive to protect users from MEV?

See also

• Bonding Curves --- the deterministic pricing mechanism behind pump.fun’s primary market
• PumpSwap --- pump.fun’s proprietary AMM and the vertical integration of issuance and trading
• Constant-Product AMM --- the $xy=k$ math underlying both PumpSwap and Raydium pools
• Impermanent Loss --- full derivation of the cost LPs bear in constant-product pools
• LP Profitability --- break-even analysis for liquidity providers
• MEV Fundamentals --- the extraction supply chain
• Sandwich Attacks --- how MEV bots exploit pump.fun traders specifically
• Protection Strategies --- what individual traders can do
• DEX Design Constraints --- why blockchains use AMMs instead of order books
• Market Microstructure --- the theoretical framework underlying this analysis
• Glosten-Milgrom Model --- the adverse selection model applied throughout this note