The House Edge

Appearance

8. Solana: AMMs, Auctions, and the Speed Game ​

The chain where the validator IS the slot leader IS the block constructor — and what that produces when you wire eight cities and three software stacks into one market. ​

Cold open ​

On 19 January 2025 — a Sunday — Solana's network captured $56.8 million in real economic value in a single day, of which approximately two-thirds was Jito tips paid by searchers for inclusion priority and bundle execution.[1] The chain's January 2025 monthly REV was $551.7 million; the 2025 annual total was approximately $1.4 billion. Sixteen months later, in Q1 2026, Solana's REV had collapsed 68% year-over-year to ~$89.5 million, putting the chain at #2 by REV behind Hyperliquid for the first time in its history.[2] The compression is not the story; the compression is what made the structure visible. The chain that produces $56.8 million of value-extraction in a single day, then loses two-thirds of that flow in a year as memecoin trading cools, is a chain whose underlying architecture — who captures the value when it is there, who keeps capturing it when it shrinks — is now legible in a way it was not when the volume was higher. This chapter is about that architecture: about why aggregators won spot, why no order book has, what the Jito auction actually auctions, what the validator-client diversity does and does not change, and why a Frankendancer Harmonic Performance validator co-located in Equinix FR5 in Frankfurt earns approximately double the priority fees per block of a stock Agave validator in the median rack. The Solana the trader sees is Phantom and Jupiter. The Solana that captures her dollar is eight cities, three client stacks, and six firms.

What this chapter answers ​

  • What is the Solana market structure that produces the +101% Harmonic Performance / +3% operational ratio Chapter 7 documented — and why is the gap specifically a Solana phenomenon rather than a generic on-chain one?
  • Why did the spot CLOB experiment end on Solana, and what replaced it?
  • What does the Jito Block Engine actually auction — and what changes under BAM and JIP-24?
  • Where does retail's dollar go on Solana in 2026, and which firms have the most concentrated claim on it?

The setup ​

Solana's architectural premise, stated cleanly: the validator is simultaneously the slot leader (the actor whose four-slot turn it is to produce blocks), the block constructor (the actor whose software orders the transactions inside the block), and — under the Block Assembly Marketplace introduced in July 2025 — the executor of TEE-attested ordering decisions made by separate BAM Node operators. No proposer-builder separation layer sits between the validator and the construction of the block. No consensus-internal matching engine subsumes both. The integration is what makes client choice, scheduler mode, and infrastructure relationships dominate Solana validator economics in a way they cannot dominate the others. On Ethereum the proposer signs a block built by an external builder (Chapter 5); the proposer's revenue is the builder's bid, less the relay's fee. On Hyperliquid the matching engine sits inside HyperBFT consensus and the validator set is permissioned-top-21 (Chapter 9); the validator has no separable block-construction role to optimise. On Solana the validator's software is the block-construction logic, and the choice of which software is the largest single revenue lever the validator has.

The chapter previews three stratified-by-architecture findings that the worked example and the mechanics develop. First: retail loses less in 2026 than in 2024, because the combination of Beam (Jupiter's protected-routing engine), BAM (Jito's TEE-encrypted block construction), the 8 April 2026 protocol patch (Chapter 3), and the Solana Foundation's enforcement against sandwich-enabling private mempools (June 2024, Chapter 3 / Chapter 7) collectively closed the formal-sandwich surface. The cumulative take at every stack layer remains the cost of routing through five firms the trader never sees — but the firms are not sandwiching her any more. Second: market makers won spot trading by displacing the public AMM with prop-AMMs. The CLOB experiment ended; the pool became inventory; the aggregator became the front of the market. Third: validators with infrastructure relationships have the largest dollar magnitude per actor and the lowest visibility. The +101% Harmonic Performance gap Chapter 7 documented is the chain-specific signature of Solana's vertical integration: the gap exists because the validator's revenue ceiling is a software-and-relationships function, not a physical-operations function. On Ethereum the gap collapses into the builder layer. On Hyperliquid it collapses entirely. On Solana it is the chain's defining economic shape.

The worked example ​

Alice — the trader the book has followed since the prologue and Chapter 1 — makes her $10,000 USDC → SOL swap one more time, this time through the Solana 2026 stack as it actually exists in May 2026. The chapter follows the dollar through every layer of routing, sequencing, and settlement she touches without seeing.

The published, estimated, and partially-opaque take at each layer:

LayerActorTake (illustrative)Cumulative
WalletPhantom (Helius rebate share)fraction of the ~$0.50–$1.00 rebate flows back; opaque whether passed to user[3]~$0
RPCHelius (50/50 with wallet)~$0.50–$1.00 captured by Helius; split with Phantom; not separately disclosed[3:1]~$1
AggregatorJupiter (Ultra)~5–10 bps slippage embedded via prop-AMM internalisation[4]~$6
RoutingJupiter Beam$0 to user (Beam is free in Ultra); upside is sandwich-avoidance[5]~$6
BundleJito tip (any backrun)6% of any bundle tip routes to the Jito DAO post-JIP-24nominal
ValidatorFrankendancer Harmonic Performance+101% priority fees per block vs median, captured at the slot-leader layer[6]~$8 priority-fee equivalent
Total~$8–$15 cumulative take vs Alice's ~$73 of formal-sandwich slippage in 2024 (Chapter 3)

Two acknowledgements. First: several of the per-layer figures are illustrative rather than published. The Helius/Phantom split is documented as 50/50 at the RPC layer; whether Phantom passes the rebate through to Alice or retains it as platform revenue is not disclosed by either firm. Beam's pricing is not published as a take; the value Beam delivers is in slippage avoided, not slippage added. The +101% Harmonic Performance per-block priority-fee gap is the Syndica March 2026 measurement (Chapter 5 / Chapter 7) — it is the validator-level revenue uplift, not Alice's specific tip. Second: the table's cumulative number compresses from ~$73 in 2024 to ~$8–$15 in 2026 because the formal-sandwich surface closed. The dollar shrunk. The structural take — five firms touching one trade — did not.

The mechanics, in detail ​

Aggregators ate the interface — and Jupiter took the aggregator layer ​

What Chapter 2 framed as "aggregators are the front of the market" is, in Solana 2026, the chain's structural shape. Jupiter held approximately 93.6% of aggregator volume in Q1 2026.[7] Aggregator-routed flow accounted for approximately 74% of total Solana DEX volume — up from approximately 40% six months prior. Jupiter's 2025 cumulative routed spot volume was approximately $716 billion.[8] The interface-replacing-the-pool dynamic Chapter 2 documented at the venue level operates, at the chain level, as the interface-replacing-the-chain: the surface most Solana retail users see is Phantom on top, Jupiter underneath, and prop-AMM inventory under that.

Jupiter Ultra — the protected-routing product behind the Jupiter user interface — handles approximately 95% of Jupiter's aggregator volume and approximately 50% of all Solana DEX volume. The protection mechanism is Beam, Jupiter's transaction-landing engine: signed transactions route to Jito as private bundles rather than over the public submission paths the chain inherited from its no-mempool design (Chapter 3). Published Ultra V3 slippage measurements: ~+0.006% on Beam-protected paths versus −0.14% on unprotected — a sign-reversal that puts Beam-routed retail on the correct side of slippage on average, which is structurally unusual for retail.[5:1] In April 2026 Beam opened as a public Transaction Submission API — any signed Solana transaction can now route through Beam, not just Jupiter-routed swaps. The structural fact: Beam is converging toward becoming the default transaction-submission surface for Solana retail, with Jupiter as the operator and Jito's bundle market as the destination.

The founding vision, and the CLOB experiment that ended ​

Solana's founding pitch, in 2017, was specific. Anatoly Yakovenko's whitepaper described a blockchain "so fast and so cheap that you can put a working central limit order book on top of it." The stated competitive target was NASDAQ. The point of the technology was to make on-chain matching cheap enough and fast enough to host the kind of order book that runs every major equity exchange in the world.

In May 2025, eight years after the whitepaper, Yakovenko co-authored a post on the Anza blog with Anza's lead economist Max Resnick. The post is titled The Path to Decentralized NASDAQ. The opening line is the cleanest single sentence on Solana's state-of-the-union the chain's own leadership has produced: "Solana was originally founded to build a blockchain that is so fast and so cheap that you can put a working central limit order book on top of it. Five years in, we have not succeeded in that goal."[9] The rest of the essay diagnoses why. The diagnosis matches the structural argument this chapter and Chapter 7 have been building. A single leader controls ordering inside each slot. That leader runs the Jito auction. Market makers cannot reliably cancel stale quotes before opportunistic takers lift them — Resnick and Yakovenko cite the cancel-race statistic that market makers on centralised exchanges win only about 13% of the time, and report that market makers on Solana win it even less. The structural shape that produces the +101% Harmonic Performance per-block revenue gap is the same shape that prevents a working spot CLOB from existing.

The 2026 data ratifies the admission. Phoenix, the Solana CLOB Chapter 2 anchored the spot-order-book story on, produced about $3.7 million in revenue in Q2 2024 and $68,604 in Q1 2026 — a 54-fold decline in six quarters.[10] The venue was renamed Phoenix Legacy in early 2026 and Ellipsis Labs pivoted to Phoenix Perpetuals. Manifest at about $3.9 billion of 30-day spot volume is the only above-noise Solana spot order book remaining; against Solana's roughly $284.5 billion of Q1 2026 DEX spot volume, Manifest captures under 2% of the market.[11] The architectural diagnosis is the one Chapter 2 developed. No public mempool means searchers cannot observe pending orders, so the front-running risk that sustains CLOB-style market making on Ethereum does not exist on Solana the same way. Aggregator-routed flow internalises into prop-AMM inventory rather than crossing on a public book. The order-book design produced too little volume to compensate the market makers who would host it. Those market makers (Ellipsis Labs who built Phoenix, Wintermute, Temporal) now operate prop-AMMs instead, where they internalise adverse selection and keep the spread.

The prop-AMM segment in March 2026: about 55% of Solana DEX volume ($36 billion of $65 billion total), down from 68% in February, competitively contested across operators. The mid-2025 picture Chapter 2 carried has shifted. HumidiFi (reported to be operated by Temporal) held about 65% of the prop-AMM segment in mid-2025 but was overtaken by BisonFi (Forward Industries' Solana dark AMM, launched December 2025) in late January 2026. Both then declined more than 55% in March as memecoin volume cooled and the segment compressed.[12] The operator map in May 2026: HumidiFi → Temporal (reported, Chapter 2); SolFi → Ellipsis Labs (publicly claimed by the Phoenix team, Chapter 2); Tessera V → Wintermute (Wintermute-confirmed to DL News in August 2025); BisonFi → Forward Industries (publicly disclosed); ZeroFi, Goonfi → undisclosed.[13] Forward Industries is the chapter's one new-to-the-book institutional cameo. A publicly-listed US company operating a Solana dark AMM compresses the regulatory-disclosure gap the segment previously enjoyed. The mid-2025 picture was a small number of crypto-native trading firms running quasi-private AMMs with no public-equity-style disclosure obligation. The 2026 picture is the same architecture with one of the operators now subject to SEC reporting requirements at the parent-company level. The regulatory exposure has shifted. The structural shape has not.

What the Jito auction actually auctions ​

The Jito Block Engine is not, mechanically, an auction over "MEV from a public mempool" — Solana's Gulf Stream produces no mempool (Chapter 3), so there is no public MEV pool to auction. The Block Engine auctions three things in one product: (a) inclusion priority for searcher bundles into the next slot's block, against SOL tips paid alongside the transactions; (b) atomic bundle execution with revert protection — the whole-or-nothing primitive that makes multi-leg searcher strategies viable (Chapter 4); and (c) routing access to validators running Jito-derivative clients. The third is the structurally largest: approximately 32% of Solana stake runs Agave Jito and approximately 28% runs JitoBAM, with Jito-Solana variants accounting for roughly 94% of stake by Q2 2025 across all flavours.[14] When a searcher pays a Jito tip, the tip is paid for guaranteed routing into a block built by software that ~94% of the chain's stake runs.

The dollar magnitudes through this auction in 2025 were among the largest single-product capture in on-chain history. Late-2024 Jito tips averaged approximately $2.5 million per day; the single-day high of $14.7 million was recorded on 17 November 2024.[15] Across 2025, Jito tips were 41.6%–66% of Solana monthly REV depending on the month, and accounted for approximately $720 million of the chain's roughly $1.4 billion annual REV.[16] Post-JIP-24 (Chapter 6), the entire 6% Block Engine cut — and the equivalent 6% BAM cut — routes to the Jito DAO treasury rather than to Jito Labs the firm: projected $15–50 million annual run-rate to the DAO, with the company that built the product continuing to operate it but no longer capturing the take rate.[17] The structural argument the chapter develops: the Jito auction is the largest single value-routing mechanism on Solana, and its commercial structure post-JIP-24 is a DAO-owned protocol take rather than a firm-owned product margin.

The architectural shift introduced under BAM matters at a different layer. Jito's Block Assembly Marketplace launched on Solana mainnet on 21 July 2025 with four named launch BAM Node operators: Helius, Triton One, SOL Strategies, and Figment (the same firms Chapter 6 profiled as Solana's infrastructure incumbents). BAM Nodes run inside AMD SEV-SNP TEE enclaves; the sequencing decision is delegated to the BAM Node, the execution remains with the validator. The structural effect: ordering decisions move out of the validator's own scheduler and into a TEE-attested process whose inputs the validator can no longer inspect — which makes selective transaction-content-based extraction (sandwich attacks, content-based censorship) provably difficult to perform without the BAM Node colluding (Chapter 3). By late February 2026, approximately 25% of Solana stake was connected through BAM.[18] BAM Node sites have expanded beyond the four launch operators' initial locations to include Dublin and Dallas; Lithuania is next; Brazil, Hong Kong, Washington DC, Miami, and South Africa are under evaluation. The open-source target is mid-Q2 2026 — a prerequisite for onboarding third-party operators at scale, since the TEE attestation is verifiable only against published code.

The client + geography matrix — the chapter's signature material ​

Two threads, one subsection. The first thread is software; the second is space.

Software. Solana's validator-client distribution in March 2026, stake-weighted: Agave Jito 32%, JitoBAM 28%, Agave Harmonic 17%, Frankendancer 12%, Rakurai 6%, Firedancer 2% — with approximately 86% of stake running an Agave-based client (Agave-derivatives plus their forks).[19] Maintainers: Anza ships Agave (the reference client); Jito Labs ships the Jito-Solana variants and JitoBAM; Jump Crypto ships Frankendancer (hybrid) and Firedancer (the full C/C++ rewrite); Rakurai is a closed-source Agave fork from an ASIC/SoC-background team that raised $3 million in 2025 from Anagram Ventures (Chapter 5). The structural observation: none of these clients is sandwich-default in 2026. The June 2024 Foundation enforcement removed the visible operators (Chapter 3, Chapter 7); Harmonic's four scheduling strategies (FBA + Priority Fee Ordering, FIFO, MREV, Custom) are SFDP-compliant under their default configuration; BAM's TEE architecture makes content-based extraction provably difficult; Beam's protected routing closes the public-submission path retail used to lose on; the 8 April 2026 protocol patch added a chain-level mechanism whose exact SIMD identifier and validator-client release tag are not publicly documented but whose observable effect was a further compression of the residual sandwich surface (Chapter 3 carries the flag).[20]

Harmonic — the block-building marketplace whose strategies dominate the priority-fee capture distribution — is also the architectural concentration point Chapter 6 documented: Ben Coverston is CEO of both Temporal (HumidiFi's reported operator) and Harmonic (the block-building marketplace). The same individual leads the firm that operates the largest prop-AMM and the firm whose validator-side strategies the +101% per-block gap is measured against. The chapter notes this as a structural fact rather than as an accusation: there is no published evidence of cross-firm coordination, and both firms operate as separately incorporated entities with separate funding rounds. The concentration is in the leadership rather than in the cap tables.

Rakurai's "no timing games" positioning matters for the same structural reason. The Figment migration report of 2 March 2026 — the cleanest published institutional-adoption case of the year — committed explicitly that Rakurai does not introduce intentional block-production delays past the 400ms slot target.[21] The migration outcomes: Figment's Staking Reward Rate rose from approximately 6.85% to 7.17% (a +32 basis-point uplift), priority fees rose approximately 60%, MEV capture increased approximately 5×, and tip capture median rose approximately 158%. One software switch, no other operational changes, and an institutional validator's per-block revenue stepped up by a measurable double-digit-basis-points-of-stake-reward magnitude. The structural argument Chapter 5 anchored — that client choice is the moat — has its cleanest 2026 datum in this migration.

Space. Solana stake concentrates in six cities, in this order: Frankfurt 19%, Amsterdam 16%, London 12%, Vilnius 6%, Tokyo 4%, Ashburn (Northern Virginia) 4% — together approximately 61% of all staked SOL. West Europe plus North America host approximately 70% of the chain's stake.[22] Jito's mainnet block engines run in eight cities: Amsterdam, Dublin, Frankfurt, London, New York, Salt Lake City, Singapore, and Tokyo — five of which match the top-six stake cities.

Optimal round-trip latency block-engine-to-validator is under 50 milliseconds; under 100 milliseconds operates well within current bundle throughput. Measured slot latencies: 0.25 ± 0.9 slots Amsterdam-to-Europe; 0.47 ± 1 slots Utah-to-North-America; 0.29 ± 0.46 slots Tokyo-to-Asia.[23] Standard Solana validator colocation in 2026: bare-metal AMD EPYC 9354P plus 512GB RAM plus three NVMe drives (one-time CapEx approximately $8,000–$15,000), with rack and power at approximately $200–$500 per month in Equinix-tier data centres; premium colocation runs $400–$800 per month. Equinix FR5 in Frankfurt and Equinix NY5 in New York are the named anchor data centres the validator community has converged on. The "next to the TPU port" pattern Chapter 3 introduced — a non-voting validator co-located with a paid RPC in the same rack, forwarding the current leader via QUIC over a LAN — produces a 5–10× latency reduction versus remote cloud routing, per RPC Fast's measurements.[24]

The structural argument the geography subsection lands: the +101% Harmonic Performance validator from Chapter 7's decomposition is, in practice, the validator that has all five of the following: (a) it runs Frankendancer (or JitoBAM); (b) with Harmonic's Performance scheduling strategy enabled; (c) co-located in Equinix FR5 or AM3; (d) connected through a BAM Node; (e) routing transactions inbound from a Helius-backed RPC stack. The validator-quintile gap Chapter 7 documented is geographic, software, and relationship-based — not operational. The block-engine footprint maps almost exactly onto the stake footprint because both follow data-centre infrastructure that predates either; engines locate where validators are; validators colocate where engines are; both follow Equinix's European and North American backbone. The shape is an equilibrium, not a one-way causation. But the equilibrium has a sharp consequence: a stock-Agave validator in São Paulo, with no SFDP delegation, no BAM Node relationship, no Frankfurt colocation, and no Harmonic strategy enabled, earns at-or-below network median priority fees regardless of how disciplined its operations are. The access gap is the chain's economic shape. Operational excellence is the floor.

SFDP, Alpenglow, and Constellation: the response from inside ​

The Solana Foundation Delegation Program is the chain's largest single-source stake-delegation pool and the de facto soft-governance lever the Foundation operates against the concentration shape the previous subsection documented. Starting 1 May 2026 (two weeks before this chapter is written) SFDP participants must operate on an ASN and hosting provider that holds less than 25% of network stake. The data-centre concentration rule requires that no single data-centre provider host more than 15% of all staked validators across the chain.[25] The "onboard 1, offboard 3" policy retires validators with less than 1,000 SOL of external stake after 18 months of program participation. About 150 validators are projected to lose Foundation stake under the May 2026 rule. The Foundation has two structural levers. An anti-concentration lever (the data-centre rule directly counteracts the Frankfurt/Amsterdam colocation gravity the previous subsection documented). And an anti-extraction lever (the censorship-resistance rules; the "no sandwich-enabling private mempools" rule; the June 2024 removal of about 32 validators for participating in DeezNode-style arrangements, Chapters 3 and 7). DeezNode itself has not been litigated further. The validator identified as HM5H6…jdMRA has not been publicly removed from SFDP between June 2024 and May 2026, and Tim Garcia's "ongoing enforcement" framing has not produced a published itemised list of subsequent removals (Chapter 7).

Alpenglow, the consensus-layer rewrite Chapters 1 and 5 introduced, activated on testnet on 11 May 2026 and is targeted for mainnet activation in late Q3 or early Q4 2026. The full architectural treatment belongs in Chapter 12. Three Solana-specific shifts matter against the equilibrium just documented. Vote transactions are removed from blockspace; about 70% of all Solana transactions at peak load are validator votes today, and removing them liberates that capacity for fee-paying transactions. The Validator Admission Ticket introduces about 1.6 SOL per epoch entirely burned as the first explicit per-epoch fixed validator cost in the network's history. And Helius's modelling suggests the profitable-validator stake threshold drops from about 4,850 SOL (~$800,000) to about 450 SOL (~$75,000), a roughly tenfold reduction in the stake required to operate profitably.[26] The structural question Alpenglow poses against the geographic-concentration equilibrium this chapter documented: does the VAT plus threshold drop reverse the Frankfurt-Amsterdam concentration shape by making cheaper validators viable in less-served geographies, or does the access gap survive the cheaper-validator economics intact because the +101% Harmonic-Performance uplift remains a software-and-relationships function rather than a stake-economics function?

The honest answer the Anza team gives in public is the second one. Alpenglow widens the validator base. It does not, on its own, narrow the per-block-revenue gap between top-quintile and bottom-quintile validators. That gap exists because a single leader controls ordering inside the slot, which means the Jito auction and the Harmonic strategy the validator runs together determine how much of the block's MEV the operator captures. Alpenglow leaves the single-leader structure in place. To narrow the access gap, the Anza team has published a second proposal that picks up where Alpenglow leaves off.

Constellation, unveiled by Anza CEO Brennan Watt at Solana Accelerate in Miami on 5 May 2026, is the multi-leader architectural answer. Designed by Max Resnick (the same Anza economist who co-authored the Path to Decentralized NASDAQ essay with Anatoly Yakovenko in May 2025) and Quint Kniep, Constellation replaces the single-leader-per-slot structure with approximately sixteen concurrent proposers rotating every ~1.6 seconds. Each proposer assembles transactions into "pslices" on a 50-millisecond cycle. 256 attesters sign attestations documenting what each proposer produced. A Constellation block is invalid if it excludes a transaction that has been attested by 40% or more of the attester set. The leader cannot, in this architecture, decide to censor or reorder a fee-competitive transaction; doing so produces a block that fails consensus.[27]

The structural target Constellation aims at is the exact gap this chapter has documented. If sixteen concurrent proposers ingest transactions and the designated leader merely assembles attested pslices into a block, the leader's discretion shrinks. The +101% Harmonic Performance per-block revenue uplift Syndica measured in March 2026 is, in this architecture, a feature of the current single-leader bottleneck. Multiple concurrent proposers compress it. Constellation also introduces a two-tier fee market that formalises the cancel-race fix the Path to Decentralized NASDAQ essay diagnosed. Inclusion fees go to validators; ordering fees are burned; a "cancels before takes" policy gives market makers a chance to pull stale quotes before opportunistic takers can lift them. The explicit business case in the Resnick / Yakovenko essay is the cancel-race statistic the Anza team published: on centralised exchanges, market makers win the cancel race only about 13% of the time, and on Solana today they win it even less often. Constellation is designed to invert that ratio.

Constellation is targeted for Q3 2026 mainnet activation, contingent on Alpenglow shipping first. The two-step roadmap (Alpenglow → Constellation) is the Anza team's published response to the same diagnosis this chapter and Chapter 7 documented from outside. Whether the chain's leadership ships both upgrades on schedule, and whether the upgrades deliver the access-gap compression the team promises rather than producing a new concentration shape at a different layer of the stack, are the open questions Chapter 12 develops as the book's closing forward look.

How this plays out on each chain — and what Solana's structure isn't ​

Solana is the chain where the validator-as-slot-leader-as-block-constructor produces the access-vs-operational gap as a software-and-geography function rather than a hardware-and-latency function. It is also the chain whose REV compresses fastest as the trading volume it captures contracts: a 68% year-over-year drop in Q1 2026 REV is a structural fact about how concentrated the chain's revenue is in a small set of speculative-volume verticals (memecoins, recurring NFT cycles, the Telegram-bot trading layer).

Hyperliquid is the structural alternative the next chapter develops in full: the chain whose matching engine sits inside HyperBFT consensus, whose validator set is permissioned-top-21, whose HLP is the chain's house. The architectural surface Solana's vertical integration could not fully eliminate — the gap between the validator and the block constructor — Hyperliquid's design does eliminate, by collapsing both into the chain's consensus layer. The empirical consequence the cold open named: Hyperliquid passed Solana in Q1 2026 REV. Whether that ranking persists is a question for Chapter 9; that it happened at all is a structural data point about value capture under different chain architectures.

Ethereum and its L2s are the third structural option, with the most mature exclusive-flow surface (Chapter 7) and the most explicit PBS separation. The cross-chain comparison Chapter 10 develops in full puts the Solana access gap against Ethereum's exclusive-flow gap and Hyperliquid's no-gap architecture. The three-chain picture is the central structural argument of Part III.

Who wins, who loses, why ​

Three-actor stratification, in order of dollar magnitude per actor.

Validators with infrastructure relationships win the most. The "top validator" on Solana in 2026 — a Frankendancer Harmonic Performance instance co-located in Equinix FR5 in Frankfurt, connected through a BAM Node, with a Helius-backed RPC stack handling inbound retail flow — earns approximately +101% priority fees per block versus a stock Agave-default median validator with no relationships, no Frankfurt colocation, and no SFDP delegation. The Figment Rakurai migration of 2 March 2026 demonstrated the magnitude at the cleanest published institutional scale: one software switch produced a +32 basis-point Staking Reward Rate uplift on the firm's Solana stake. The aggregate Gini coefficient for Solana validator profits in 2025 was approximately 0.93 (Chapter 7) — extreme inequality, with the top-100 validators receiving substantial inflation-based rewards and the bottom-100 relying almost entirely on transaction fees and MEV when selected as slot leader. The named beneficiaries are concentrated: Helius (the largest single Solana validator with over 15 million SOL staked); the four BAM Node launch operators (Helius, Triton One, SOL Strategies, Figment); the Frankendancer Harmonic Performance operator set whose names appear across Syndica's reports.

Market makers win spot trading. Prop-AMMs — HumidiFi, SolFi, Tessera V, BisonFi, and the smaller ZeroFi/Goonfi entrants — collectively capture approximately 55% of Solana DEX volume in March 2026 and internalise the spread the public AMM design used to give away. The named operators behind them are a small set of trading firms: Temporal (HumidiFi), Ellipsis Labs (SolFi, also Phoenix Perpetuals), Wintermute (Tessera V), Forward Industries (BisonFi). The structural argument Chapter 2 developed at the venue level operates, at the chain level, as the inversion of the LP-democratisation thesis the public AMM design originally promised: the spread does not flow back to public depositors because there are no public depositors of consequence on Solana spot in 2026; the spread flows back to the firms operating the prop-AMMs.

Retail traders lose less than in 2024 but still pay the structural take. Alice's $73 of 2024 formal-sandwich slippage compressed to approximately $8–$15 of 2026 cumulative take in the worked example — a meaningful improvement that the combination of Beam + BAM + the April 8 patch + Foundation enforcement collectively produced. The compression is real and is the chain's clearest improvement to retail welfare since 2024. The structural take, however, did not disappear: five firms still touch one trade — Phantom, Helius, Jupiter, Jito, and the leader validator — and four of them are firms the trader has never heard of. The Solana that compressed its formal-sandwich surface did so by replacing one visible extraction layer (the public-mempool sandwich) with five less-visible ones (the wallet-RPC rebate, the aggregator routing, the bundle market, the validator-side priority-fee capture, the BAM Node attestation layer). Whether this is an improvement is a judgement call. The chapter offers the comparison and the dollar trace; the verdict belongs to the reader.

The closing structural observation. Solana's 2026 market structure is what happens when a chain's vertical integration eliminates the proposer-builder separation Ethereum has and never accepts the consensus-internal matching Hyperliquid has. The resulting equilibrium concentrates value capture in a small set of firms operating across the wallet-RPC-aggregator-bundle-validator stack, with the access gap between top-quintile and median validators driven by software and infrastructure relationships rather than operational discipline. The chain is not optimising for retail welfare. It is not optimising for validator equality. It is optimising for the throughput-and-low-cost retail experience that has made Phantom + Jupiter the most-used retail interface on any permissionless chain in 2026, and the cost of that optimisation is the structural concentration the chapter has documented.

The Anza team's diagnosis matches the book's. The structural shape the book has documented from outside (an order-of-magnitude access-to-operational ratio for validator revenue; a concentrated infrastructure stack; an exclusive-flow surface that compresses the spot CLOB) is the same shape Resnick and Yakovenko diagnose from inside in the Path to Decentralized NASDAQ essay. The team running the chain accepts that the founding vision has not landed. Alpenglow plus Constellation is their two-step published response. Whether the response ships, on what timeline, and whether it produces the access-gap compression the team promises rather than relocating the concentration to a different layer of the stack, are the open questions Chapter 12 develops. The chapter's verdict is that the structural shape this chapter has documented is known to the people running the chain and is the subject of in-flight architectural work. Whether that work delivers is not yet a fact in evidence.

What changes when… ​

Chapter 9 picks up the structural counterpoint. Hyperliquid is the chain whose architecture eliminates the surface Solana's vertical integration could not — no mempool, no separable block constructor, the matching engine inside HyperBFT consensus, and a permissionless-top-21 validator set whose role is fundamentally different from Solana's slot-leader-as-block-constructor model. The empirical question the cold open raised — how does a chain pass Solana in Q1 2026 REV with a fraction of Solana's transaction throughput? — is Chapter 9's central explanation. The structural question that follows from it — does Hyperliquid's design eliminate the extraction Solana's design merely compresses, or does it concentrate the extraction differently? — is Chapter 9's verdict.

Footnotes and sources ​

  1. Blockworks REV dashboard, Solana Network REV (Real Economic Value), https://blockworks.com/analytics/solana/solana-overview/solana-network-rev-real-economic-value; bit2me News, REV: metrica de ingresos blockchain Solana, https://news.bit2me.com/en/rev-metrica-ingresos-blockchain-solana. Solana 2025 annual REV ~$1.4B; Jito tips 41.6%–66% of monthly REV depending on month; January 2025 monthly peak $551.7M; single-day high $56.8M on 19 January 2025. Accessed 2026-05-14. ↩︎

  2. PANews, Solana Q1 2026 Network Activity Report, https://www.panewslab.com/en/articles/019da9c6-e9d5-76c0-904a-878418fd7fbc. Q1 2026 Solana REV ~$89.5M, down ~68% YoY as memecoin volume cooled; Solana ranks #2 by REV behind Hyperliquid for the first time. Accessed 2026-05-14. ↩︎

  3. Helius, Backrun Rebates Documentation, https://www.helius.dev/docs/sending-transactions/backrun-rebates; Helius, Wallets Use Case, https://www.helius.dev/use-case/wallets. 50/50 RPC-rebate split between Helius and integrator wallet (named wallets: Phantom, Ledger, Bitgo, Exodus, Backpack, Squads, Trust, Solflare). Already cited Chapters 6, 7. Accessed 2026-05-14. ↩︎ ↩︎

  4. Syndica, Deep Dive: Solana DeFi, March 2026, https://blog.syndica.io/deep-dive-solana-defi-march-2026/. Prop-AMM internalisation typically embeds 5–10 bps of effective slippage in retail-routed orders; the precise per-route number varies by depth of liquidity and the aggregator's internal route selection. Already cited Chapter 2. Accessed 2026-05-14. ↩︎

  5. Jupiter, Ultra V3 Documentation, https://developers.jup.ag/docs/ultra. Ultra V3 published slippage measurements: ~+0.006% on Beam-protected paths versus −0.14% on unprotected. Beam opened as a public Transaction Submission API in April 2026 — any signed Solana transaction can now route through Beam. Accessed 2026-05-14. ↩︎ ↩︎

  6. Syndica, Deep Dive: Solana Onchain Activity, March 2026, https://blog.syndica.io/deep-dive-solana-onchain-activity/. Per-block priority-fee captures versus the network median: Frankendancer Harmonic Performance +101%, Frankendancer Harmonic Balanced +39%, Agave Harmonic +36%. Block-time-adjusted figures slightly lower. Already cited Chapters 5, 6, 7. Accessed 2026-05-14. ↩︎

  7. SolanaFloor, Jupiter Reclaims Dominance with 93.6% Market Share in Solana's Aggregator Landscape, https://solanafloor.com/news/jupiter-reclaims-dominance-with-93-6-market-share-in-solana-s-aggregator-landscape. Jupiter aggregator share Q1 2026 ~93.6%. Already cited Chapter 2. Accessed 2026-05-14. ↩︎

  8. SolanaFloor, The Rise of Aggregators in Solana DeFi: Over 70% of DEX Volume Now Routed Through Aggregators, https://solanafloor.com/news/the-rise-of-aggregators-in-solana-de-fi-over-70-of-dex-volume-now-routed-through-aggregators-reaching-a-7-month-high. Aggregator-routed share ~74% of total Solana DEX volume; 2025 cumulative Jupiter routed spot volume ~$716B. Already cited Chapter 2. Accessed 2026-05-14. ↩︎

  9. Max Resnick (Anza) and Anatoly Yakovenko (Solana Labs), The Path to Decentralized Nasdaq, Anza Blog, 8 May 2025, https://www.anza.xyz/blog/the-path-to-decentralized-nasdaq. The opening admission ("Solana was originally founded to build a blockchain that is so fast and so cheap that you can put a working central limit order book on top of it. Five years in, we have not succeeded in that goal.") is the canonical statement of the founding-vision-vs-2026-state-of-the-union framing. The essay diagnoses single-leader ordering as the root cause and outlines a multi-leader response, a two-tier fee market (inclusion fees to validators, ordering fees burned), and a "cancels before takes" ordering policy. The 13%-cancel-race statistic for centralised exchanges (with Solana market makers winning the race even less often) is the essay's published business case. The companion SolanaFloor coverage (https://solanafloor.com/news/anatoly-yakovenko-max-resnick-outline-gameplan-nasdaq-on-solana) summarises the public reception. Accessed 2026-05-14. ↩︎

  10. DefiLlama, Phoenix Protocol Page, https://defillama.com/protocol/phoenix; Ellipsis Labs, Introducing Phoenix Perpetuals, https://www.ellipsislabs.xyz/blog-posts/introducing-phoenix-perpetuals. Phoenix peak Q2 2024 revenue $3.7M, Q1 2026 $68,604 (~54× decline); renamed Phoenix Legacy; Ellipsis Labs pivoted to perpetuals. Already cited Chapter 2. Accessed 2026-05-14. ↩︎

  11. DefiLlama, Manifest Trade Protocol Page, https://defillama.com/protocol/manifest-trade; Pine Analytics, Solana DEX Spot Volume Q1 2026. Manifest 30-day spot volume ~$3.9B; Solana Q1 2026 DEX spot volume ~$284.5B. Already cited Chapter 2. Accessed 2026-05-14. ↩︎

  12. SolanaFloor, Prop-AMMs on Solana: BisonFi Ends HumidiFi's Dominance with $11.5B Volume, https://solanafloor.com/news/prop-amms-on-solana-bison-fi-ends-humidi-fi-s-dominance-with-11-5-b-volume; Syndica, Deep Dive: Solana DeFi, March 2026, https://blog.syndica.io/deep-dive-solana-defi-march-2026/. March 2026 prop-AMM share ~55% of Solana DEX volume ($36B of $65B total); HumidiFi and BisonFi both declined >55% in March; BisonFi launched December 2025 and overtook HumidiFi in late January 2026. Accessed 2026-05-14. ↩︎

  13. DL News, Solana Dark AMMs Make Trading More Efficient, But at a Cost, https://www.dlnews.com/articles/defi/solana-dark-amms-make-trading-more-efficient-but-at-a-cost/. HumidiFi → Temporal (reported); SolFi → Ellipsis Labs (publicly claimed); Tessera V → Wintermute (Wintermute-confirmed to DL News August 2025); BisonFi → Forward Industries (publicly disclosed); ZeroFi, Goonfi → undisclosed. Already cited Chapter 2. Accessed 2026-05-14. ↩︎

  14. Syndica, Deep Dive: Solana Onchain Activity, March 2026, https://blog.syndica.io/deep-dive-solana-onchain-activity/. Stake-weighted client distribution March 2026: Agave Jito 32%, JitoBAM 28%, Agave Harmonic 17%, Frankendancer 12%, Rakurai 6%, Firedancer 2%. Already cited Chapters 5, 6, 7. Accessed 2026-05-14. ↩︎ ↩︎

  15. Tokenomics.com, Jito Tokenomics: How JTO Captures MEV and Staking Revenue on Solana, https://tokenomics.com/articles/jito-tokenomics-how-jto-captures-mev-and-staking-revenue-on-solana. Late-2024 Jito tips averaged ~$2.5M/day; single-day high $14.7M on 17 November 2024. Accessed 2026-05-14. ↩︎

  16. Helius Research, Solana MEV Report, https://www.helius.dev/blog/solana-mev-report. Solana 2025 MEV ~$720M; first year MEV exceeded priority fees as the largest REV component. Already cited Chapter 6. Accessed 2026-05-14. ↩︎

  17. Jito Foundation, JIP-24: Jito DAO Receives All Jito Block Engine Fees and Future BAM Fees, https://forum.jito.network/t/jip-24-jito-dao-receives-all-jito-block-engine-fees-and-future-bam-fees/860. Post-JIP-24: entire 6% Block Engine + 6% BAM cut routes to Jito DAO treasury; projected $15–50M annual run-rate. Already cited Chapter 6. Accessed 2026-05-14. ↩︎

  18. Jito, Introducing BAM, https://bam.dev/blog/introducing-bam/; Chainflow Solana, Summary of Solana Validator Discussions (February 2026), https://chainflowsol.substack.com/p/summary-of-solana-validator-discussions-8be. BAM mainnet launch 21 July 2025; four launch BAM Node operators (Helius, Triton One, SOL Strategies, Figment); ~25% of stake connected by late February 2026; new sites Dublin + Dallas; Lithuania next; open-source target mid-Q2 2026. Already cited Chapter 3. Accessed 2026-05-14. ↩︎

  19. Syndica, Deep Dive: Solana Onchain Activity, March 2026. Detail in [14:1]. Maintainers: Anza (Agave); Jito Labs (Jito-Solana variants + JitoBAM); Jump Crypto (Frankendancer + Firedancer); Rakurai (closed-source Agave fork). Already cited Chapter 5. Accessed 2026-05-14. ↩︎

  20. Edgen Tech, Solana Network Shuts Down Sandwich Attacks on April 8, https://www.edgen.tech/news/post/solana-network-shuts-down-sandwich-attacks-on-april-8-boosting-user-security. The 8 April 2026 protocol patch's specific SIMD identifier, validator-client release tag, and consensus mechanism are not in a canonical primary source as of the chapter's writing date; observed effect was a further compression of the residual sandwich surface. Already flagged Chapter 3. Accessed 2026-05-14. ↩︎

  21. Figment, Figment Upgrades Solana Infrastructure with Rakurai Client, https://www.figment.io/insights/figment-upgrades-solana-infrastructure-with-rakurai-client/; PRNewswire, Rakurai Raises $3M Seed Round to Accelerate Development of High-Throughput Solana Nodes, https://www.prnewswire.com/news-releases/rakurai-raises-3m-seed-round-to-accelerate-development-of-high-throughput-solana-nodes-302395935.html. Figment migration 2 March 2026: SRR 6.85% → 7.17% (+32 bps); priority fees +60%; ~5× MEV; tip capture median +158%, p95 +55%. Rakurai's explicit "no timing-game manipulation past 400ms target" commitment. Already cited Chapter 5. Accessed 2026-05-14. ↩︎

  22. Helius, Solana Decentralization: Facts and Figures, https://www.helius.dev/blog/solana-decentralization-facts-and-figures. Solana stake by city: Frankfurt 19%, Amsterdam 16%, London 12%, Vilnius 6%, Tokyo 4%, Ashburn 4% (top 6 ≈ 61% of all stake); West Europe + N. America ≈ 70%. Accessed 2026-05-14. ↩︎

  23. Jito, Low Latency Transaction Send, https://docs.jito.wtf/lowlatencytxnsend/. Jito mainnet block engines in Amsterdam, Dublin, Frankfurt, London, NY, Salt Lake City, Singapore, Tokyo. Optimal block-engine-to-validator latency <50ms; <100ms operates well within current bundle throughput. Measured slot latencies: 0.25 ± 0.9 Amsterdam→Europe; 0.47 ± 1 Utah→N. America; 0.29 ± 0.46 Tokyo→Asia. Accessed 2026-05-14. ↩︎

  24. RPC Fast, Solana Trading Co-Location and Low-Latency Setups, https://rpcfast.com/blog/solana-trading-co-location-low-latency; The Good Shell, Solana Validator Cost, https://thegoodshell.com/solana-validator-cost/. Standard colocation: AMD EPYC 9354P + 512GB RAM + 3× NVMe (CapEx $8–15K); rack + power $200–500/mo; premium $400–800/mo. Equinix FR5 (Frankfurt) and NY5 (New York) as anchor data centres; "next to the TPU port" pattern produces 5–10× latency reduction vs remote cloud. Accessed 2026-05-14. ↩︎

  25. Solana Foundation, Delegation Criteria, https://solana.org/delegation-criteria; Phemex News, Solana Foundation Updates Validator Delegation Requirements for May 2026, https://phemex.com/news/article/solana-foundation-updates-validator-delegation-requirements-for-may-2026-69090. May 2026 SFDP rules: <25% ASN/hosting-provider stake-share; <15% data-centre concentration; ~150 validators projected to lose Foundation stake. "Onboard 1, offboard 3" retires validators with <1,000 SOL external stake after 18 months. Already cited Chapter 5. Accessed 2026-05-14. ↩︎

  26. CoinDesk, The Biggest Consensus Overhaul in Solana History Is Officially Live for Testing, 11 May 2026, https://www.coindesk.com/tech/2026/05/11/the-biggest-consensus-overhaul-in-solana-history-is-officially-live-for-testing; SIMD-0326 (Alpenglow) and SIMD-0357 (Validator Admission Ticket), https://github.com/solana-foundation/solana-improvement-documents/blob/main/proposals/0326-alpenglow.md; Helius, Alpenglow, https://www.helius.dev/blog/alpenglow. Alpenglow testnet 11 May 2026, mainnet target late Q3 / early Q4 2026. VAT ~1.6 SOL/epoch entirely burned; profitable-validator stake threshold drops from ~4,850 SOL (~$800K) to ~450 SOL (~$75K) per Helius modelling. Already cited Chapters 1, 5. Accessed 2026-05-14. ↩︎

  27. Brennan Watt (Anza) and Max Resnick, Constellation (proposal unveiled at Solana Accelerate, Miami, 5 May 2026). Coverage and technical write-up: Helius, What is Constellation? Multiple Concurrent Proposers on Solana, https://www.helius.dev/blog/constellation; Constellation proposal tracker, https://constellation.anza.xyz/. ~16 concurrent proposers rotating every ~1.6 seconds; 256 attesters; 60% attestation participation required for block validity; any pslice attested by ≥40% of attesters must be included by the leader (else the leader's block is rejected). 50-millisecond economic tick. Two-tier fee market (inclusion fees to validators, ordering fees burned). Designed to work as a preprocessor to Alpenglow; targeted for Q3 2026 mainnet activation contingent on Alpenglow shipping first. Accessed 2026-05-14. ↩︎