Pump.fun routes 47.7% of its revenue to buying back and burning PUMP. Sky routes 8.3% to SKY. That is a 5.7x difference in payout ratio and a 10.7x difference in yield. Run both through TI's four-input buyback framework and neither one works, for opposite reasons. They do not sit at opposite ends of a quality scale. They fail the same test from opposite directions, and the inputs they fail on are the ones that decide whether a buyback does anything at all.
TI's token buybacks framework holds that a headline buyback yield is close to meaningless on its own. Four inputs decide whether the program does anything: whether it is net of insider supply, net of product surprise, net of revenue durability tier, and net of cumulative buyback intensity. The framework already scores HYPE at four of four, SKY at two, AAVE at three into a bad window, and JUP at zero despite carrying the highest gross yield in the cohort.
TI first ran Pump.fun through this framework in a June 2026 appendix to "JUP Has the Highest Buyback Yield in DeFi. The Token Is Down 80%.", when PUMP carried a $510M market cap and traded within 5% of its all-time low. It has since rallied to $782M and sits 73% above that low, which makes it worth rescoring. What is new here is not PUMP's presence in the cohort but the pairing: setting it against Sky isolates the two inputs each one fails, and those inputs turn out to be different ones. If yield alone mattered, PUMP would be the best buyback in crypto. It returns nearly half of revenue to holders, has retired an amount equal to 39.8% of its current market cap in twelve months, and offers a 9.9% yield even on fully diluted value.
The placement is the uncomfortable part. PUMP does not land near SKY on that plot. It lands in the same quadrant as JUP, the program TI's framework scored at zero of four, and for the same structural reason: the highest gross yield in the cohort, funded by the least durable revenue, paid against a float that is still expanding. High payout ratios cluster with weak capacity because a protocol with cyclical revenue has every incentive to return cash while the cycle is good.
A buyback is a claim on future revenue paid for with present revenue. The yield tells you the size of the claim. The four inputs tell you whether the claim is worth anything.
Start with what is genuinely working. Pump.fun routed $13.5M to token holders in the trailing 30 days and $311M over the trailing year, per DefiLlama holdersRevenue. That is 47.7% of the $28.2M in combined 30-day revenue across the launchpad, PumpSwap and Terminal, which corroborates a reported 50% allocation with observed on-chain flow rather than a press release.
Cumulative intensity: passes, decisively. $311M against a $782M market cap is 39.8% of the float retired in a year. The framework's own threshold language is that a 2% yield does not move a $1B float and a 7% yield does. PUMP is at 21% on circulating supply. Whatever else is true, this program is large enough to matter.
Revenue durability: fails. The framework sorts revenue into tiers and puts memecoin-correlated flow in the bottom band, for the reason JUP demonstrated: when Solana memecoin volume collapsed, the revenue funding the buyback fell with it. Pump.fun is the purest expression of that tier. Q2 2026 revenue fell 42.9% year over year and 25.1% quarter over quarter, and trailing 30-day revenue sits 81% below the January 2025 peak of $149.8M.
Insider supply: fails. Only 47.1% of PUMP is circulating, 399.6B of 847.9B tokens. The buyback is retiring float while more than half of total supply has yet to reach the market. This is why the yield figure most commonly quoted is misleading in a specific direction: 21% is the yield on circulating market cap, 9.9% is the yield on fully diluted value, and the gap between them is the dilution the buyback has to outrun.
Product surprise: mixed. PUMP rose 32.2% over the past week, which is a genuine signal that the mid-July unlock cleared a known overhang rather than introducing new selling. Against that, revenue has fallen in five of the last six months. The market got a better supply outcome than feared and a worse revenue outcome than hoped.
One offset deserves credit. PumpSwap and Terminal now produce roughly 34% of total revenue, up from near zero in early 2025. The terminal-decline case now has to argue against three products rather than one. That does not make the revenue durable, but it does make it less purely reflexive than the launchpad alone.
Sky is the mirror image, and its scorecard is the more interesting of the two because it has been treated as the conservative option.
Insider supply: passes cleanly. 99.6% of SKY is circulating, 23.36B of 23.46B tokens. Market cap and fully diluted value are within half a percent of each other. Every dollar spent on the buyback retires float that actually exists. There is no dilution to outrun.
Cumulative intensity: fails. Sky routed $1.12M to holders in the trailing 30 days against $13.5M of revenue, an 8.3% payout ratio that corroborates the reduction to 7.5% made in March 2026. Annualized, that is roughly $14M against a $1.48B market cap, a 0.92% yield. By the framework's own standard this is below the level at which a buyback moves anything. The trailing year figure is better at $82M, or 5.6% of market cap, but that number includes the period before the allocation was cut.
Revenue durability: passes on tier, weakens on trend. This is where the comparison stops being comfortable. Stablecoin issuance is the framework's example of sticky top-tier revenue, and Sky is the archetype. But Sky's monthly revenue has fallen from $29.2M in August 2025 to $12.7M in June 2026, a decline of 56%. Q2 2026 was down 16.3% year over year and 18.6% quarter over quarter.
Sky's revenue is more durable than Pump.fun's by tier and by rate of decline. It is not stable. A framework input that reads as a binary pass conceals a business that has more than halved in eleven months.
Product surprise: fails. A protocol whose revenue has halved while its buyback was cut by an order of magnitude has not exceeded what the market priced in.
The headline comparison is 47.7% versus 8.3%, and it invites the conclusion that Pump.fun is generous and Sky is stingy. The framework says something more specific: each protocol has exactly what the other lacks, and neither has both.
Sky has the capacity to run a meaningful buyback and does not. Its float is fully circulating, its revenue sits in the durable tier, and a dollar spent retires a real dollar of supply. It routes 8.3% of revenue, which produces a yield below the level at which the framework expects any effect.
Pump.fun has the commitment and lacks the capacity. It routes nearly half of revenue, at a scale that has retired 39.8% of its market cap in a year, from a revenue base that fell 42.9% year over year, against a float where more than half of supply is still to come.
Commitment and capacity are independent. The market prices them as one thing, and reads a high payout ratio as evidence of a healthy business rather than a claim on an uncertain one.
There is a second-order point about credibility that cuts against intuition. Sky reduced its buyback discretionarily through governance, from 75% to 7.5%. Pump.fun reduced its allocation too, from 100% to 50%, but reporting describes that commitment as locked in an irreversible contract for a year. TI has verified the flow, not the contract terms. If those terms hold, the protocol with the smaller percentage cut ended up with the more credible commitment, because it gave up the ability to change its mind.
That distinction matters more than the payout ratio. A discretionary 8.3% can become 0% at the next vote. A contractual 50% cannot, for as long as the contract binds. What neither mechanism can do is guarantee the revenue underneath it.
Three refinements fall out of running these two through the four inputs.
The durability tier needs a direction. Sky passes the durability input on category and would pass it on a checklist. Its revenue is down 56% from its August 2025 peak. The input as written asks what kind of revenue funds the buyback. It should also ask where that revenue is heading, because a top-tier revenue stream in decline funds a shrinking buyback just as reliably as a bottom-tier one.
Insider supply belongs in the framework as a yield gap. The most useful single number in this comparison is the spread between yield on circulating supply and yield on fully diluted value. For Sky the two are identical at 0.92%. For Pump.fun they are 21% and 9.9%. That spread is the dilution overhang stated in the same units as the buyback, and it is more legible than a supply percentage.
Commitment structure belongs in the framework. None of the four inputs currently capture whether a buyback can be revoked. Sky's cut is the evidence that this matters: the program did not fail, it was reduced by governance. A framework that scores only size and funding quality will keep missing that failure mode.
Neither token is in TI's signal registry. Both are research-covered only, and nothing here argues for promoting either.
PUMP. The buyback is real, large and verifiable, and the market is right that a sub-10% FDV yield on a business with three products is not obviously expensive. The bear case is not that the buyback is fake. It is that a 47.7% payout ratio on a revenue base down 42.9% year over year is a claim whose size is set by the thing most likely to shrink. The unlock overhang is the second problem, and it is the one the commonly quoted yield hides.
SKY. The cleanest float in this comparison and revenue in the durable tier, paired with a buyback too small to matter and a revenue line down 56% from peak. The risk here is not collapse. It is that the conservative case rests on a durability assumption the recent trend does not support.
Methodology note. Every figure is computed at write time from DefiLlama fees and holdersRevenue endpoints and CoinGecko market data rather than carried from secondary reporting. Payout ratios are observed flows rather than stated policies, which is why they corroborate rather than repeat each protocol's own numbers. The buyback contract mechanics described for Pump.fun, including the one-year lock, are reported and not independently verified by TI.