Grayscale Research published two Top 10 tables this cycle. Nine tokens appear on both. The tenth slot differs — SKY on one, CRV on the other. That single swap carries more information than the other nineteen rows combined, and it got buried under the headline number.
There is a second problem nobody flagged. The report timestamp reads October 22, 2026. Either the dataset is forward-dated, or a year was fat-fingered somewhere upstream. Either way, every claim in that ranking about "current" emissions and price levels has to be downgraded to directional. I don't size positions on directional. I size them on verified.
Two metrics, one asset pool. Token holder yield measures what actually reaches holders after dilution. Adjusted earnings yield measures protocol economics against market capitalization. The framing comes from Grayscale's research desk under Zach Pandl, and it is a better framework than most: it forces the question of net transfer instead of gross revenue.
The overlap is nine of ten, which strongly implies a shared universe of assets. That is the assumption I am working from, and it is the assumption that makes the arithmetic below possible.
Holder yield table: HYPE, LIT, CAKE, AERO, UNI, RAY, PENDLE, AAVE, JUP, CRV. Adjusted earnings table: CAKE, LIT, HYPE, RAY, AAVE, JUP, SKY, UNI, AERO, PENDLE.
Bear market context matters here. When price is the only variable moving, everyone hunts for a cash-flow story. Yield rankings are that story. They are also the easiest product to sell into a market that has forgotten how to underwrite.
I spent three months in 2018 auditing the 0x v2 contracts and found seven reentrancy paths. The lesson was not that code is fragile. It was narrower and more useful: a contract can only pay out along an execution path that actually exists. Revenue is a claim. Payment is a path. Every yield number is downstream of that distinction.
Divide one list by the other. Holder yield over earnings yield is the share of each unit of economic earnings that actually returns to the holder — value capture efficiency.
HYPE: 8.0 percent over 9.8 percent, 82 percent capture.
LIT: 7.9 percent over 14.1 percent, 56 percent capture.
CAKE: 2.6 percent over 14.4 percent, 18 percent capture.
JUP: negative 2.6 percent over positive 3.5 percent, negative capture.
CAKE ranks first on earnings yield at 14.4 percent and captures 18 percent of it. The protocol is genuinely profitable. The holder is not the beneficiary. The residual sits in treasury, LP incentives and market-maker subsidies — legitimate uses, none of them yours. The number one slot on that table is a headline, not a payout.
HYPE is the only asset clearing both dimensions, and the reason is structural rather than cyclical. Protocol fees route into an on-chain Assistance Fund buyback. No committee votes on cadence. No governance proposal sets the size. Automated execution and governance-adjustable execution are different trust tiers wearing the same label. CAKE burns on a periodic schedule with veCAKE gauges setting direction. AERO distributes fees to veAERO lockers. RAY, PENDLE and AAVE all run buyback or fee-share paths with discretionary parameters. Those are promises, and promises get re-voted.
Then there is the set difference. SKY appears in the adjusted earnings Top 10 but not the holder yield Top 10, while CRV sits at the bottom of the holder yield list. If both tables share a universe, SKY's holder yield is below CRV's — worse than negative 5.5 percent. That is the ex-MakerDAO entity running a Smart Burn Engine alongside migration incentives large enough to swamp it. High earnings, maximal dilution, one ticker. A protocol can be profitable and still function as a transfer mechanism away from its own holders.
UNI at fifth on holder yield is the anomaly I would verify first. For most of its history UNI has been the canonical example of a governance token with zero value capture. Its presence in the top five implies the fee switch or a burn mechanism is live. If it is, that is a paradigm shift in UNI's fundamentals. If it is not, the methodology has a hole. Either answer is worth more than the ranking itself.
Run the risk math. A 9.8 to 14.4 percent earnings yield against assets that annualize 60 to 100 percent volatility, versus a roughly 4 percent risk-free rate, is 5 to 10 points of compensation for a 15x volatility differential. That is not a spread. That is a lottery ticket priced as a bond.
Retail reads the leaderboard. Smart money reads the ratio. Data speaks louder than sentiment, and the sentiment here is simply that CAKE pays 14.4 percent.
JUP is the cleanest illustration of the gap. Roughly half of protocol fees go to buybacks, adjusted earnings print positive at 3.5 percent, and holder yield is still negative 2.6 percent. The repurchase is real. The emissions are larger. Buyback announcements carry far more public-relations value than actual value transfer, and the market prices the announcement every time.
There is stale-data risk layered on top. If the 2026 timestamp is a data-entry error and the snapshot is older than it appears, HYPE's and LIT's emission phases may have moved materially. LIT's 56 percent capture ratio almost certainly includes first-year incentive emissions — the blend of fee share and subsidy that every new protocol runs before the subsidy ends. The second-place slot may be the most overrated number on the page.
I have watched this movie. In 2020 I put fifty thousand dollars into Uniswap v2 ETH/USDC and learned that impermanent loss erodes principal faster than APY compounds. The dashboard yield was not the wallet yield. Holder yield versus earnings yield is the same accounting trick, scaled from a pool to a protocol. Liquidity dries up when trust breaks, and the trust here is a governance parameter that can be voted away in a single cycle.
Three things to watch, in order of information value: on-chain confirmation of UNI's fee switch, SKY's burn engine versus its migration emissions, and whether HYPE's buyback stays automated or migrates behind a governance vote.
Panic sells, logic buys. In a market where price is the only variable, cash-flow rankings become the product. The question is whether they measure returns or marketing — and whether the buyer can tell the difference before the parameter changes.