Cost of Trust 2.0

Our founding claim in 2018 was structural: trust was about to become contestable. The maintenance of records, the enforcement of agreements, the custody of value, the verification of identity and ownership, the activities that comprise roughly thirty-five percent of the United States labor market, and on the order of twenty-nine trillion dollars of global economic activity, had been intermediated for over a century by for-profit companies and institutions whose moats were jurisdictional licensing, regulatory capture, and the cost to coordinate alternatives. Those moats, we argued, were structurally weaker than they appeared. A new category of organization, the token network, could provide the same trust functions more cheaply, more openly, and at global scale, because it possessed three structural properties that for-profit companies do not.

It was open source: anyone could fork the protocol, inspect the rules, and verify that the system worked as advertised. In a domain where the product itself is trust, transparency is not a feature but the substrate. It was permissionless: anyone could participate, transact, build, or extend without asking. The thesis was that this would unlock global addressable markets that no jurisdictionally-bounded company could reach, and allow parties to agree on new technology standards without fearing platform risk. And it was programmable: trust functions that legacy infrastructure delivered as discrete services could be composed, recombined, and built upon as primitives. The thesis was that this would produce new categories of product that legacy infrastructure could not match.

From these three properties, we made a directional prediction: that open source innovation in high-cost-of-trust markets would outpace closed source, that token networks would outcompete for-profit companies in the activities those companies had historically intermediated, and that the largest outcomes in crypto over the coming decade would concentrate in three specific categories where the structural properties had the most leverage. Decentralized finance, where the legacy intermediaries (banks, custodians, exchanges, brokers) extracted the highest rents on the largest pools of capital, and where the programmable property made composability across companies and geographic boundaries possible. Payments, where global, real-time, low-cost settlement was structurally outside what correspondent banking could deliver, and where the permissionless property made global reach achievable. And Web3, our framing at the time for everything from decentralized infrastructure to consumer applications built on token incentives.

The receipts on those three calls, eight years later, are unambiguous. Decentralized finance became the largest category crypto produced. The onchain financial stack collectively cleared tens of trillions of dollars in cumulative volume, roughly $11 billion in trailing-twelve-month onchain protocol revenue per the 1kx Onchain Revenue Report (June 2025 to May 2026), and produced more than half of our top outcomes when we include the substrate protocols that hosted them. Stablecoins emerged as the successor to payments, establishing themselves by 2025 as the most resilient sector in the crypto landscape. In 2024, stablecoin transfer volume hit $28 trillion, surpassing Visa's $15.7 trillion total volume for the same period, according to Visa Onchain Analytics. This dominance is further evidenced by multi-billion-dollar equity M&A activity such as Stripe's $1.1 billion acquisition of Bridge in October 2024 and Mastercard's $1.8 billion acquisition of BVNK in 2026. Visa and Mastercard are now integrating native stablecoin settlement into their networks, and the technology is achieving deep penetration into emerging markets that remain structurally inaccessible to traditional banking institutions.

Outside the three categories we named, the broader institutional reception of crypto crystallized over the same period. The US spot Bitcoin ETFs, approved in January 2024, brought $23 billion of net inflows in 2025 alone, with BlackRock's IBIT surpassing $100 billion in AUM within twenty-two months of launch. Pensions, endowments, and registered advisors now allocate to Bitcoin as a portfolio asset. Traditional financial balance sheets have started coming onchain in tokenized form: tokenized US Treasuries grew from near-zero in late 2022 to roughly $15.5 billion onchain by May 2026 (RWA.xyz), with Circle's USYC, Ondo, BlackRock's BUIDL, and Franklin Templeton's Benji as the leading institutional issuers.

The third category requires a more careful accounting. Our 2018 framing of Web3 was broader than the term is used today. It included what we now call DePIN (decentralized physical infrastructure networks for compute, storage, wireless, mapping, geospatial), and it included a mass-market consumer thesis: that token incentives and onchain ownership would produce durable consumer applications across social, identity, gaming, and creator economies. Of the two branches, DePIN has emerged as the one showing meaningful traction. As per the 1kx Onchain Revenue Report, it is currently the fastest-growing crypto sub-sector by fees. The category as a whole remains earlier in its monetization curve than DeFi or payments. The honest framing is that DePIN is the branch of the 2018 Web3 thesis that is showing promise, not yet the one that has fully arrived. The mass-market consumer branch, by contrast, has not produced durable outcomes at scale.

Another element of the 2018 thesis aged more partially than the others, and it is worth naming directly. We argued that open source innovation would outpace closed source. The claim was substantially correct at the protocol layer: Bitcoin, Ethereum, and the durable DeFi primitives are open source and have outcompeted every closed-source alternative attempted in their domains. It was less correct at the application layer, where the work of making crypto usable, distributing it to mainstream users, and integrating it with regulatory and consumer infrastructure has been captured significantly by for-profit companies. Coinbase, Stripe via Bridge, the wallet stack, the institutional custody providers, the on-ramp and off-ramp infrastructure. These are equity companies, and they capture material value in the categories our 1.0 thesis named. The corrected reading is narrower and more accurate: open source dominates at the trust-bearing protocol layer; closed source competes effectively at the usability and distribution layer above it. Both layers exist; both produce returns, and our portfolio includes investments in both.

The remainder of our 2018 thesis, three structural properties (open source, permissionless, programmable), the directional prediction (token networks outcompete for-profit companies in high-cost-of-trust markets), and the three named categories (DeFi, payments, Web3), carried us through three market cycles. The thesis was not vague. It was specific, opinionated, and largely correct at the level it operated.

What it did not yet have was the within-category filter that distinguishes which configurations capture venture-scale outcomes from which do not. The point of running an investment firm for eight years is to accumulate the pattern data that produces that filter: to see, with the benefit of completed cycles and realized outcomes, which bets inside the called categories actually paid and which did not. The following patterns emerged with enough clarity over those years to make the sharper formulation possible.

The winners cluster, and they cluster on identifiable mechanisms

When we re-categorized our portfolio by function rather than by the older infrastructure/middleware/consumer taxonomy, more than ten companies and protocols cleared our outcome bar, defined as a realized exit or a current mark at 10x-100x+. They clustered into two groups: financial infrastructure and base-layer protocols. The common thread was that the company or protocol either compressed rent on activity that legacy infrastructure was expensive at, or created activity legacy infrastructure structurally could not offer.

The losers cluster, too, and they cluster cleanly

The portfolio's underperforming category was consumer-facing applications. Our retrospective shows that most of these were investments where blockchain was a substrate for distribution-and-timing bets without a structural reason for the technology to be load-bearing. The product could have been built with a database. The trust problem the blockchain solved was not the binding constraint. The technology was decoration, not architecture.

Two independent records, historical winners and historical losers, produce the same two-mechanism shape. That convergence is what made it possible to write a sharper thesis with conviction rather than after-the-fact rationalization. Cost of Trust 2.0 is the formalization of that pattern.

There are two distinct mechanisms by which blockchains create durable economic value within the addressable surface 1.0 identified. A thesis-fit deal captures one. The largest outcomes capture both.

Mechanism 1: Trust-intermediary rent compression.

Replace a paid trust intermediary; compress the rent.

A blockchain that delivers the same trust outcome at lower cost makes the intermediary's wage the addressable market.

Examples where the mechanism is real and the rent is meaningful: correspondent banking on cross-border B2B payments (up to 340 basis points on lower-value cross-border flows [McKinsey, 2024], plus 2 to 5 days of settlement delay); custody and audit on bearer assets (prime broker fees, insurance, regulatory capital); bookmaker margins on event markets (4 to 10 percent vig); bank net interest margin on lending (200 to 400 basis points in developed markets, wider in emerging). These are recurring, fee-priced rents that a competitor can, in principle, undercut.

Not all trust intermediaries are private. Central banks and governments charge the largest such rents in the global economy as the wage for currency issuance and monetary sovereignty: seigniorage, inflation tax on holdings, FX-control rent in restricted-currency jurisdictions, and the confiscation-risk premium realized in periodic bank holidays and bail-ins. Other public-sector intermediaries charge for land and property registration, notarization, business licensing, and legal certification. These rents are not itemized on a billing statement, but they are functional wages paid to be trusted with the currency, the record, or the credential. Blockchain configurations compress these rents where users can voluntarily opt out, holding value in a sovereign-independent asset or transacting via a sovereign-independent rail. Bitcoin as a portfolio asset is the canonical demonstration; stablecoin holdings by users in capital-controlled jurisdictions are another. The framework also predicts which public-sector rents yield to compression and which do not: where users can opt out (currency holding, cross-border value transfer, savings allocation), compression is real; where they cannot (taxation, regulatory licensing, healthcare mandates), coercive enforcement or lack of alternatives protect the rent.

Examples where the mechanism is weak: government recording fees on land titles (nominal, not market-priced); enterprise database costs in supply chain coordination (diffused across many parties, no single intermediary extracting capturable rent); consumer login (free at point of use; the cost is paid in privacy and data, which users have not historically been willing to pay to recover).

Mechanism 2: Enablement / zero to one.

Make possible what legacy infrastructure cannot produce. The trust still matters (users need to believe their money will pay out, the oracle will resolve correctly, the platform will not seize funds).

Polymarket is the canonical case, and worth stating carefully because a regulatory counter-narrative is available. The proximate reason event-contract markets were absent from the US for two decades was CFTC enforcement, not technology: Intrade was pushed out in 2012, PredictIt's no-action letter was revoked, and Polymarket itself was barred from US users after a 2022 settlement. The market reopened only when Kalshi won its 2024 lawsuit establishing that election event contracts were not categorically prohibited. A skeptic could fairly observe that none of this required blockchain, and that Kalshi (centralized, CFTC-regulated, US-legal) demonstrates the technical feasibility of a regulated alternative. The 2.0 frame holds, but it has to be stated precisely. Even under the now-permissive regime, Kalshi is structurally constrained. Every contract operates within CFTC category-level prohibitions and a 90-day review window (the regulator disapproved Kalshi's Congressional Control contracts in 2023 before the 2024 court reversal on election contracts); it must still avoid any market the CFTC can deem gaming or unlawful, it gates participation behind identity verification and excludes dozens of jurisdictions, and it cannot aggregate fully permissionless global liquidity into a single citable price the way a neutral protocol can. Polymarket's dominance over Kalshi on the 2024 US presidential election (clearing over $3.6 billion in volume on a single event) was driven by exactly the surface the regulated form cannot serve: markets on questions a regulator will take too long to approve, global participation that no licensed venue can operationally handle, and the information externality that made Polymarket odds citable in Bloomberg, the Financial Times, and the New York Times. Enablement under 2.0 does not require that the legacy system be forbidden to produce a product; it requires that the legacy system, in the form it actually operates, cannot produce it.

Perpetual swaps are another. Crypto invented the perpetual swap in 2016. They do not exist in traditional finance (TradFi). The category is now being copied back into the legacy financial system following CFTC clarity, but the original win was the creation of a product type, not the compression of fees on an existing one. Tokenized treasuries on a public chain (BlackRock's BUIDL, Superstate, Ondo) are not cheaper than money market funds. They are programmable, composable with onchain collateral systems, and accessible 24/7. Stablecoin dollar accounts for users in inflationary economies are not undercutting a cheaper alternative; there was no accessible alternative for retail-scale synthetic dollar holdings outside the formal banking system.

Why blockchains specifically

Both mechanisms can be captured, in principle, by centralized players. Stripe compresses payment rent. AWS enables products that did not exist before cloud computing. The thesis is meaningful only if blockchains are the technology that uniquely captures the mechanisms in cases where centralized alternatives structurally cannot. They are, for four reasons.

Programmable bearer settlement: value moves natively as data, with finality, without a custodial chain, and onchain assets compose directly with onchain applications in the same transaction.

Verifiable computation and state: anyone can audit what happened, what the rules are, and what assets exist; solvency is provable rather than asserted.

Structural neutrality: a credibly decentralized protocol cannot retroactively change rules, exclude users, or capture the value layer for itself the way a company-controlled platform can.

Permissionless global participation: anyone with an internet connection can transact, hold assets, deploy capital, or build an application without asking.

These properties map directly onto the two mechanisms. Compressing trust-intermediary rent requires programmable bearer settlement and verifiable state: without the first, value still routes through bank-mediated rails that recapture the margin; without the second, the new intermediary inherits the same trust cost as the old one, and the rent simply moves rather than compresses. Creating what legacy infrastructure cannot requires structural neutrality and permissionless participation. Without structural neutrality, any coordination layer becomes a competitive risk for the parties it coordinates, the failure mode that ended R3 Corda and IBM Food Trust. Without permissionless participation, the global user base that drives the largest enablement wins cannot be served: emerging-market dollar access, Polymarket's geographic reach, and the entire DeFi user base. The strongest outcomes capture all four properties at once. Centralized challengers can, in principle, replicate any single property, and have.

Additionally, the configurations that capture both mechanisms also produce a class of operational skills the legacy form cannot acquire without restructuring: smart-contract risk underwriting (curated lending markets), MEV-aware strategy design (basis trade vaults, perp funding harvesting), permissionless market creation (event contracts), oracle-driven settlement, restaking economics, and cross-chain composability. The four-property substrate generates these skills as a byproduct of the configuration. Centralized challengers can adopt the rails (BlackRock issued BUIDL, Apollo routes through Morpho, Visa partnered with stablecoin issuers) without acquiring the operational-skills moat that determines which configuration on the rails actually wins.

Within this mapping, the L1 chain itself captures both mechanisms by substrate attribution. The chain does not directly compress a specific legacy trust-intermediary rent at the protocol layer; it delivers the four structural properties to every application above. The realized aggregate rent compression on an L1 ecosystem exceeds the L1's own fee capture by orders of magnitude. R3 Corda, Quorum, and the family of permissioned enterprise chains failed precisely because a regulated intermediary cannot deliver the structural-neutrality property; the substrate bets that succeeded sit in the upper-right quadrant for this reason, not as enablement-only wins but as the substrate that makes every application-layer fee and cost compression possible.

The thesis:

Stated explicitly: Cost of trust investments capture value through one of two mechanisms: (1) trust-intermediary rent compression, undercutting the fees and costs current trust intermediaries extract, or (2) enablement, creating trust-dependent outcomes that legacy infrastructure cannot produce. The strongest outcomes compound both.

The framework places the industry's historical record on a single chart. The horizontal axis is trust-intermediary rent compression: how much rent current trust intermediaries extract that a blockchain alternative could undercut. The vertical axis is enablement / zero to one: how much demand the legacy system structurally cannot serve. Bubble size reflects trailing-12-month onchain revenue per category, drawn from the 1kx Onchain Revenue Report (June 2025–May 2026); color reflects the highest early-stage venture return achieved within each category across the industry (data sourced from Messari, DefiLlama, TokenTerminal).

Four observations:

Upper-right: the strongest wins

Stablecoin payments to emerging-market B2B and remittance corridors capture both mechanisms: meaningful rent compression on correspondent banking (up to 340 basis points on lower-value cross-border flows), and structural enablement of dollar access for users excluded from US banking.

The decentralized exchange layer (DEX) sits in the same quadrant: over $4 billion in TTM onchain revenue, compressing centralized-exchange listing fees and stable-pair spreads decisively while enabling permissionless listing and global, composable access. The spread compression is more selective on retail-size volatile pairs, where passive AMM LPs require roughly 30 basis points to compensate for impermanent loss.

DeFi lending belongs here too: it compresses bank net interest margin and recovery cost on collateralized institutional lending while enabling a product category legacy lending cannot produce, namely, composable positions usable as collateral elsewhere in the same transaction, programmable terms, flash loans, instant settlement, and permissionless global access.

The 2022 cycle was the cleanest validation: centralized crypto lenders, including BlockFi, Celsius, Genesis, and Voyager, all failed under fractional reserve and counterparty opacity, while overcollateralized onchain lending protocols all survived without bailouts. DeFi vaults and curation, liquid staking, decentralized prediction markets, decentralized perpetuals, and tokenized private credit also belong in this quadrant. Each compresses a real rent layer (active management fees, centralized staking custody fees, sportsbook edge, CEX derivative fees, and structuring rent on small and mid-market private credit) while requiring crypto-native operational skills the legacy form structurally cannot acquire: smart-contract risk underwriting, MEV-aware strategy design, permissionless market creation, oracle-driven settlement, and restaking economics.

Bitcoin as a portfolio asset also belongs here for analogous reasons. L1 chains also belong here by substrate attribution, for the reason already given above; this is where the substrate bets in our portfolio sit. Asset tokenization belongs here too, by the same logic: it delivers the four properties to every asset class, financializing asset categories that could not exist in legacy markets (governance tokens, prediction-market positions, NFTs, attention markets, compute resources) and compressing the intermediation layer (transfer agents, settlement float, custody, listing) across the existing universe of stocks, bonds, real estate, and commodities.

Upper-left: enablement-driven wins

Centralized perps and derivatives platforms and the crypto exchanges (Coinbase, Binance) sit here, alongside adjacent categories that share the same enablement-only profile: wallets and interfaces, token launchpads, DePIN, and the Digital Collectibles & IP category. The rent compression argument is weak in each case (no incumbent was extracting comparable rent), but the enablement argument is strong. These are products and access channels that did not exist before crypto, or that crypto enables at a scale and accessibility that legacy systems cannot match. The configurations in this quadrant rely on operational skills the legacy form already has (matching engines, custody, KYC, fiat banking integration), which is why Coinbase and Binance compete effectively at the exchange layer despite using architecture closer to legacy exchange tooling than to native onchain primitives. Coinbase and Binance are the cleanest validations that enablement can produce venture-scale winners across the onchain surface, including at the exchange layer that brought digital assets to the broad retail and institutional user base.

Lower-right: pure rent compression is rare in practice

Major onchain categories at venture scale rarely sit in pure rent compression for long. The framework allows for this quadrant analytically (rent compression without an enablement layer underneath), and the empirical record is not empty but narrow.

Configurations in this quadrant are structurally unstable; they exit through one of five paths: regulator-forced compliance, rent layers thinner at scale than projected, growth of an enablement layer that becomes dominant, capture of the compression trade by a non-blockchain tech-enabled disruptor, or legacy adoption of the same rails. The closest current case is stablecoin retail remittance to non-sanctioned emerging-market corridors, where USDT and USDC on Tron, Solana, and Stellar are compressing Western Union and MoneyGram fees from roughly 6.5 percent toward 1 percent on small retail transfers, at a market projected to grow past 45 billion dollars annually by 2027. The compression is real and large, and the legacy form is now adopting the rails: Western Union announced its own dollar stablecoin, USDPT, on Solana for the first half of 2026, MoneyGram has been settling on USDC for over a year, and operators with deep emerging-market corridor relationships and local-banking integration contest the residual margin that follows the compression. The net effect at the user level is permanent rent compression, but the durable winners capture an enablement layer the legacy form structurally cannot run, not just the bare compression.

For state-law-regulated rents (title, notarization, registries), the same dynamic plays out earlier and against crypto: a non-blockchain tech-enabled disruptor (centralized remote-online-notarization platforms, AI-augmented title verification at legacy carriers) captures the compression inside the regulatory frame, where the blockchain's structural neutrality is a disadvantage rather than an asset. Lending is the cleanest counter-case: banks can add onchain credit, but they cannot offer composable, permissionless, programmable positions without becoming a different kind of institution.

Lower-left: neither mechanism.

Land registries, supply chain blockchains, decentralized notarization, real estate tokenization, enterprise consortium chains, decentralized training, decentralized compute, decentralized social, and gaming. Every category here has had major attempts and significant capital deployed. None has produced category-defining outcomes. The framework predicts this and explains why: neither rent capture nor structural enablement applies. A second failure mode appears in categories where substrate enablement is present but not sufficient: state law anchors ownership at the county or jurisdictional level (real estate tokenization, land registries), irreducible operational overhead remains beyond what the substrate can collapse, or a non-blockchain tech-enabled disruptor captures the compression trade first (notarization, title insurance).

We believe that our cost-of-trust framework not only explains the past but is also predictive about the future. The same two mechanisms and the same four properties that account for where value has already clustered are the lens we use to map the next wave of created and disrupted markets.

Why now: AI as the forcing function.

The existing trust infrastructure is being rapidly displaced and disrupted by the rise of artificial intelligence, specifically generative and agentic AI. First, generative AI has collapsed the cost of producing a convincing fake, an email, a voice, a credential, an entire counterparty, toward zero.

Second, agentic AI is creating entirely new surfaces where trust is load-bearing and fragile: autonomous agents need full-permissioned access to documents, accounts, and communication channels to realize their full potential. With the new wave of costs of trust emerging, new trust-manufacturing technologies will be required to meet these challenges.

The forward test

In practice, we run a forward opportunity through three gates and one question. The gates:

Once the initial gates are cleared, the most critical question remains: who will ultimately capture the opening? We have identified five paths that typically erode positions in the rent-compression quadrant: regulatory forced compliance, thinner-than-anticipated rent layers, the dominance of a growing enablement layer, disruption by a non-blockchain competitor who captures the trade first, or adoption of the same rails by a legacy incumbent. These pathways do not merely describe how existing positions decay; they serve as a forward-looking menu of the ways a promising opportunity can fail to achieve durability. The operational-skills moat, the configuration-specific expertise the legacy form cannot acquire without restructuring, is what separates the winner from the adopter.

What changes in the forward use is the evidence. Backward, we could color the map by realized return multiples and size it by trailing revenue. Forward, those numbers do not yet exist, so we read leading indicators instead: the size of the uncontested rent pool, the maturity curve of the enabling technology, where the binding constraint is tightening, and reveals which firms are building on whose rails. We expect the forward map to sit higher and further left than the historical one, because the largest opportunities AI is opening are enablement-led, the zero-to-one creation of markets that did not exist, more than they are the compression of rents that already do.

What this means for deployment at 1kx.

This sharpens our mandate: the forward surface divides into two regions that we will concentrate capital on, separated by maturity.

Mature trust markets: onchain finance. The core. It is where all four properties (structural neutrality, permissionless participation, programmable bearer settlement, and verifiable state) already compound, where the rents are largest and best understood, and where eight years of our record give us the most reliable read on which configurations endure. Yet we believe that the transformation of these segments is only at its beginning, and it remains the larger share of where we will deploy. The following is by no means an exhaustive list of examples:

Emerging trust markets: AI and the applications it unlocks. AI is creating new trust bottlenecks and trust-intermediary rents that did not exist before, thereby creating entirely new cost of trust markets and product categories.

At the infrastructure layer, the trust infrastructure built for humans operating in business hours cannot serve machines operating at all hours. Agentic AI is becoming fully autonomous without certifiable and deterministically verifiable outcomes. Agentic commerce is emerging without the ability to verify the counterparty and reputation of counterparty agents. Trust infrastructure that can unlock agentic execution and commerce is emerging, but remains nascent.

Following are two early examples to outline the space, from the more general to the more concrete.

At the application layer, it is the consumer and frontier products that ride the same stack: verifiable human action and provenance, where the value of proving a real person did a real thing, or that a piece of content is authentic, rises in direct proportion to how cheaply both can now be faked. These applications are not a separate sector. They are the surface expression of the same emerging trust market, and they belong next to the infrastructure that makes them possible. The near-term instances are already legible: verified-incentive advertising, where an advertiser pays only for an action a provably real human took, and content provenance, where a file's origin can be proven rather than asserted.

Cost of Trust 1.0 was a thesis about the size of the prize. Cost of Trust 2.0 is a thesis about where the prize is captured. It is what eight years of running an early-stage venture firm taught us about the difference between a market that exists and a configuration that wins.

The two theses are at different levels of resolution, and are mutually reinforcing. 1.0 named the surface. 2.0 names the mechanisms. The portfolio shape we built under 1.0 is consistent with the diligence sharpening 2.0 enforces.

Turned forward, the same lens does not predict winners, which no lens honestly can; but it tells us where to look: which trust problems are becoming solvable, which technology is ready to solve them, and whether trust was the binding constraint all along.