JUP Has the Highest Buyback Yield in DeFi. The Token Is Down 80%.

Four tokens, four active revenue-funded buyback programs, four wildly different outcomes. The token with the highest buyback yield in the set (18.83%) is down 80%. The token with the most modest yield (6.78%) is up 533%. Buyback yield by itself predicts almost nothing. The dispersion is the lesson, and TI's Net Buyback Yield framework explains it directly.

1. The Dispersion That Breaks the Simple Story

For most of 2024 and 2025, the value-accrual debate in DeFi reduced to a single argument: turn the fee switch on, route protocol revenue to token holders through buybacks, and the token reprices. That argument is now testable. Per Delphi Consulting's State of Token Markets (June 2026), four major DeFi tokens have run real, revenue-funded buyback programs long enough to produce a clean read on whether the simple story holds.

It does not. The four programs are mechanically similar (each retires tokens with protocol revenue) but the outcomes have nothing in common. Hyperliquid's HYPE is up 533% since its Assistance Fund buyback began in November 2024. Sky Protocol's SKY is up 25% since its Smart Burn Engine activated late 2024. Aave's AAVE is down 25% since the Aavenomics buyback launched in April 2025. Jupiter's JUP is down 80% since its 50%-of-fees buyback locked in February 2025. The buyback yields tell an even sharper story: JUP, the worst performer, has the highest trailing-twelve-month buyback yield in the set at 18.83%. HYPE, the best performer, sits at 6.78%, fourth-highest if anything.

This is the dispersion that breaks the simple story. The right question is not "does the protocol buy back its token?" but "what is the buyback's size relative to everything else the market is pricing into the token?" TI's Net Buyback Yield framework formalized this in May 2026 around the HYPE-vs-PUMP comparison; the four-token dispersion above is the cleanest cross-asset validation TI has seen.

2. HYPE: Why the Moderate Yield Was Enough

HYPE's buyback is structurally direct: roughly 97% of protocol fees route to the Assistance Fund, which buys HYPE on-chain and burns it. The mechanism is contract-mediated, not discretionary, and has been running since November 2024. Cumulative buybacks have reached $1.18B per Delphi's measurement window. At a current market cap near $16.24B, the TTM yield is 6.78%, lower than two other protocols in this set.

The yield did not have to be high because everything else the market was pricing was helping, not hurting. Per TI's HYPE research page: core contributor vesting is the only remaining supply pressure (insiders are otherwise fully allocated), the perp DEX product captured 70%+ of decentralized perp share, traditional-asset volume (oil, silver, S&P 500 perps) expanded the fee base in real time, and the U.S. regulatory variable shifted from "uncertain" to "incrementally cracking" with the May 2026 CFTC actions on Kalshi and Coinbase/Deribit. The buyback met a market with limited insider selling and exceptional product traction. NBY net of these inputs was meaningfully positive, and the price followed.

3. SKY: Stable Revenue, Mature Category, Modest Beta

Sky Protocol's Smart Burn Engine activated in late 2024 and routes protocol surplus into open-market SKY buybacks. Cumulative spend is $120M, TTM yield 4.25%. SKY is up 25% since program start, the second-best return in the set. The framing: stablecoin-issuer revenue is one of the most durable cashflow bases in DeFi (Tier 1 in the fee-durability framework), and SKY's buyback compounds against that base.

What kept the return modest rather than spectacular is the mirror image of HYPE's tailwinds. SKY is a mature category with limited surprise upside; the stablecoin-issuer thesis is well-understood and largely priced. Treasury yield compression in 2025-2026 reduced the absolute revenue Sky could route into the engine. The buyback worked the way the mechanism is designed to work; the category just does not produce 5x outcomes from this starting point. Per Delphi's measurement, SKY's return underperformed BTC over the reference window even though the buyback program executed as designed.

4. AAVE: High Discipline, Wrong Window

The Aavenomics buyback launched April 2025 at a planned pace of $1M/week of open-market buys, funded by protocol fee revenue. The AWW (Aavenomics) proposal that activated this also redirects 100% of protocol revenue to AAVE holders, which TI's Fee Switch Natural Experiment covered in detail. Cumulative spend is $42M, TTM yield 2.86%, and AAVE is down 25% since the program started.

The dispersion explanation is not that the buyback failed. It is that the buyback met the worst possible market window. Three forces overwhelmed the bid. First, the April 2026 rsETH / LayerZero contagion produced up to $230M in potential bad debt on Aave V3, exposing the structural concentration in Aave's pooled-liquidity model (the same-pool report from May 6 covered this). Second, the V4 migration to hub-and-spoke + Risk Premiums is the structural answer to that concentration but remains in progress, so the V4 thesis is unresolved and the V3 thesis is now riskier than it looked pre-rsETH. Third, the AWW buyback program activated April 2025 but only recently passed enough cumulative spend to be a meaningful percentage of float. The yield is real; the surrounding context is not yet pricing it as a structural fix.

5. JUP: The Highest Yield, the Worst Outcome

Jupiter's buyback program is the cleanest case study of why yield alone is not predictive. The protocol allocates 50% of platform fees to JUP buybacks, with the purchased tokens locked for three years. Cumulative buybacks reached $149M by the Delphi measurement date, and the TTM yield is 18.83%, the highest in this set by a wide margin. JUP is down 80% since the program started in February 2025.

Two structural forces dominated. The first is vesting. Per TI's prior NBY work and asset-risk-scores, JUP has been releasing roughly 53M tokens per month through mid-2026, which annualizes to ~$104M of dilution at the current price near $0.21. The buyback runs at a comparable cadence but the dilution side is roughly the same size, so the net float retirement is near zero. The second is Solana memecoin volume cyclicality. Jupiter's revenue is heavily tied to Solana DEX activity, which collapsed from the late-2025 peak as the memecoin cycle wound down (the same shock that compressed Aerodrome's fee base). The revenue side of the buyback ratio fell while the unlock side did not.

The Net Buyback Yield framework predicted exactly this pattern: gross yield is interesting, net yield is what matters, and net yield is what the price tracks. JUP's headline 18.83% yield is real but the dilution-adjusted version is not, and a token cannot defy the math of its own float schedule.

6. Extending Net Buyback Yield: The Four Inputs That Decide

Net Buyback Yield, as TI defined it in the HYPE-vs-PUMP report (May 22 2026), is gross buyback minus sell-through times insider unlocks, divided by market cap. That formula correctly identifies why JUP, despite a 19% headline yield, is functionally yielding near zero on a dilution-adjusted basis. The four-token data confirms the formula but also extends it. Three additional inputs determine whether positive NBY translates into a price re-rating, or stays positive without delivering:

The dispersion is fully explained when buyback yield is viewed alongside these three inputs. HYPE is positive on all three. SKY is positive on two of three. AAVE is positive on one and negative on one (with one neutral). JUP is negative on all three despite the highest yield. Buyback yield by itself is a one-variable model; what TI's framework, including this extension, gives you is the three additional variables that determine whether the yield has anything left to deliver after the market accounts for everything else.

7. What Would Falsify the Framework

The extended NBY read is testable on a six-to-twelve month window. The framework fails if any of the following occur:

• HYPE falsifier: if HYPE's price returns to flat-or-negative against BTC while the Assistance Fund continues buying back at 6%+ yield and insider supply stays vested, the "product surprise + supply discipline + sticky revenue" stack failed and the framework needs a fourth input we have not yet named. Window: by end of 2026.
• JUP falsifier (bullish for the token): if JUP outperforms BTC over the next two quarters despite the dilution exceeding the buyback, the framework's claim that "net of unlock matters more than gross yield" fails. Most plausible mechanism: a Solana memecoin volume recovery that materially expands the buyback bid while unlocks are unchanged.
• AAVE falsifier: if AAVE V4 ships hub-and-spoke + Risk Premiums and AAVE price still does not respond within two quarters, the AWW mechanism is structurally limited by something the framework does not capture (perhaps token-supply elasticity to floor-yield expectations).
• SKY falsifier: if stablecoin issuer revenue compresses by 30%+ over two quarters and the buyback still maintains the price floor, sticky revenue is more rate-sensitive than the framework predicts.

8. The Honest Read

Four DeFi protocols ran the same playbook. Four outcomes that have nothing in common. The buyback is the mechanism the value-accrual debate spent two years arguing for, and it is real and operational at all four protocols. None of that means the buyback is a moat. The moat lives in the four inputs above: dilution discipline, product surprise, revenue durability, and the cumulative buyback intensity to overpower the headwinds present in any given window.

For TI's framework reads on the four protocols. HYPE scores positively on all three additional inputs (supply discipline, product surprise, durable revenue tier) alongside its 6.78% headline buyback yield; the structural read from the HYPE research page holds, and the buyback delivered against a market that was already paying for the product. SKY sits on durable Tier 1-2 stablecoin revenue but lacks the product-surprise tailwind that converted HYPE's buyback into a re-rating; the buyback works as designed and the category simply does not produce alpha from this revenue base. AAVE has the supply discipline and the durable revenue, but met the unfavorable window: the April 2026 rsETH event, the V4 transition still in progress, and an AWW program that only recently passed enough cumulative spend to be material against the float. The V4 migration is the variable that resolves whether the wrong window was structural or transient. JUP scores negatively on all three additional inputs despite the highest headline yield in the set, and the price action reflects exactly that.

The portfolio question is not "which token has the highest buyback yield." That question, taken alone, would have led an allocator to JUP in 2025 and to AAVE over HYPE in 2026, both of which would have underperformed BTC. The right question is the one the framework asks: which token has the highest net buyback yield after dilution, with product surprise pointing up rather than down, on a revenue base that does not depend on cycles outside the protocol's control? On those criteria, the answer in this set is HYPE, and it is HYPE for reasons that the buyback alone cannot claim credit for.

9. Update (2026-06-05): PUMP , The Same Pattern, Greater Extreme

The DeFi Report published a Pump Fun deep-dive in May 2026 arguing that PUMP's buyback program represents the highest yield in DeFi alongside what they characterized as the most aggressive repurchase commitment in the sector. The token was, and is, down roughly 80%+. We pulled the data independently to test whether PUMP belongs in this report's four-token set, and the answer is yes , not because the framework changes, but because PUMP is the cleanest reductio yet of the central thesis.

The numbers, verified. Per DefiLlama (which sources buybacks directly from the protocol's fees.pump.fun API), cumulative PUMP buybacks since the program launched on July 14 2025 total $295M through June 5 2026. This is about 24% below The DeFi Report's $389M figure; the gap may reconcile through methodology timing or the USDC-pair under-display issue noted below. PUMP currently trades at $0.00145 (within 5% of its $0.00138 all-time low), down 83.5% from its $0.00881 ATH on September 14 2025. Market cap is $510M; FDV is $1.25B.

By every reasonable annualization, PUMP's forward Net Buyback Yield is multiples higher than JUP's, which was already the highest in the original four-token set. Trailing 30-day annualized: 36.5%. Trailing 90-day annualized: 52.1%. Trailing 180-day: 58.3%. The price has not responded. Every methodology that could understate the magnitude has been applied, and the yield is still the highest documented in any major DeFi token, by a margin that is not close.

Where PUMP slots into the four-input framework

The pattern is identical to JUP, only sharper. Both score negatively on all three extension inputs. Both depend on the same Solana memecoin volume cycle for their revenue base. Both saw the price collapse despite the buyback firing on schedule. The distinction is that PUMP's buyback rate is multiples larger, and the price collapse is equivalent. This is the most direct disproof yet of the simple value-accrual story: at high enough resolution, even an enormous buyback yield does not put a floor under the price if the three other inputs all point down.

The near-term test: July 2026 team cliff

Per The DeFi Report (citing Tokenomist), team tokens are locked until an initial cliff unlock in July 2026 , about four weeks from this update. PUMP's circulating supply is 350B against a max supply of 1T; the post-cliff vesting schedule will determine how much new dilution lands against the existing buyback. If the cliff produces a step-change in float that the buyback cannot absorb, that is the framework working as designed (negative net-of-insider-supply input dominates). If the price holds against the cliff, the framework owes an explanation for why this specific dilution event was different from JUP's continuous ~53M/month release.

What we do not claim. The DeFi Report's article includes a contrarian framing of PUMP's risk/reward at 1.4x sales with 50% of revenue going to buybacks. That valuation argument is separate from the framework question this report addresses. TI's framework establishes that PUMP's buyback yield, however large, does not by itself protect the token from a price decline of the magnitude already realized. Whether the post-decline level represents a re-entry point requires a different analysis , specifically a fee-durability tier exercise that accounts for memecoin-volume cyclicality across the next full cycle. That analysis is not in this report.

What the addition of PUMP changes about the original conclusion. Nothing. HYPE remains the answer to the portfolio question because HYPE scores positively on all three extension inputs. PUMP, like JUP, scores negatively on all three. The framework predicts JUP's outcome and PUMP's outcome with the same machinery. The interesting fact about PUMP is the magnitude: a 36-58% forward NBY producing an 83.5% drawdown is the cleanest evidence we have that the four inputs matter more than the yield headline.