a16z: 17 Structural Changes in the Crypto Industry

Original Title: 17 things we are excited about for crypto in 2026
Original Author: a16z New Media
Translation: Peggy, BlockBeats

Editor's note: As we enter 2026, the crypto industry is undergoing a deep structural redefinition: from stablecoins and RWA, AI agent ecosystems, to privacy networks, prediction markets, and legal framework reshaping, outlining a pivotal year of both technological and institutional inflection. The crypto industry is shifting from "chain performance competition" to "network effect competition," from "code is law" to "spec is law," and from trading-driven to product-driven; meanwhile, AI is driving a comprehensive evolution of agent economies and prediction systems.

This article compiles 17 forward-looking observations from various a16z teams, providing a framework for understanding the next phase of crypto narratives and industry directions.

The following is the original text:

This week, partners from a16z's Apps, American Dynamism, Bio, Crypto, Growth, Infra, and Speedrun teams released their annual "Big Ideas" trend outlook.

The following content brings together 17 observations from multiple a16z crypto partners (and several guest authors) on future developments, covering topics from intelligent agents and AI, stablecoins and asset tokenization, financial innovation, to privacy and security, prediction markets, SNARKs, and other application scenarios... and extending to future ways of building.

On Stablecoins, RWA Tokenization, Payments, and Finance

Smarter, More Efficient Stablecoin On/Off-Ramp Infrastructure

Last year, stablecoin transaction volume was estimated at $46 trillion, continuously setting new records. To better understand this scale: this figure is more than 20 times that of PayPal; nearly 3 times the transaction volume of Visa, one of the world's largest payment networks; and is rapidly approaching the annual transaction volume of the US electronic clearing network ACH (commonly used for direct deposit payroll, etc.).

Today, you can complete a stablecoin transfer in less than a second and for less than 1 cent. But the real unresolved issue is: how to connect these digital dollars to the financial networks people use every day—that is, the on/off-ramp mechanism for stablecoins.

A wave of new startups is entering this gap, trying to connect stablecoins with local payment systems and fiat currencies. Some use cryptographic proofs to allow users to privately exchange local balances for digital dollars; others integrate with regional payment networks, leveraging QR codes, real-time payment rails, etc., to enable interbank payments... Still others are building truly interoperable global wallet layers and platforms that support direct stablecoin card spending.

Overall, these approaches collectively expand the range of people joining the digital dollar economy and may drive stablecoins to be used more directly in mainstream payment scenarios.

As these on/off-ramp infrastructures mature, allowing digital dollars to directly access local payment systems and merchant tools, new behavioral patterns will emerge:

Cross-border workers can settle wages in real time;

Merchants can accept global dollars without a bank account;

Applications can instantly settle value with global users.

Stablecoins will shift from a "niche financial tool" to the foundational settlement layer of the internet era.

——Jeremy Zhang, a16z crypto engineering team

Thinking About RWA Tokenization and Stablecoins in a More "Crypto-Native" Way

In recent years, banks, fintech companies, and asset managers have increasingly wanted to bring US stocks, commodities, indices, and other traditional assets on-chain. But currently, much of RWA tokenization has a clear "skeuomorphic" tendency: still based on traditional thinking about real-world assets, without leveraging the advantages of being crypto-native.

Meanwhile, synthetic asset forms like perpetual contracts (perps) often provide deeper liquidity and are simpler to implement. The leverage structure of perps is also easier to understand, so I believe they are the crypto-native derivatives with the best product-market fit. Additionally, I think emerging market equities are one of the most worthy asset classes to "perpetualize." For example, 0DTE (zero-day-to-expiry) options markets for certain stocks often have more liquidity than spot markets, making them ideal candidates for perp experiments.

Ultimately, this is about one question: "Perpetualization vs. Tokenization." In any case, we will see more crypto-native forms of RWA tokenization in the coming year.

Similarly, in the stablecoin field, 2026 will see "not just tokenization, but on-chain native origination." Stablecoins fully entered the mainstream in 2025, and their issuance scale continues to grow.

However, stablecoins without robust credit infrastructure are essentially like "narrow banks"—they only hold a small portion of assets considered extremely safe and liquid. Narrow banks are certainly an effective product, but I don't think they will become the long-term backbone of the on-chain economy.

We are currently seeing some new asset managers, asset curators, and protocols beginning to facilitate on-chain loans collateralized by off-chain assets. These loans are usually issued off-chain and then tokenized. But I believe that, aside from making it easier to distribute to on-chain users, there is little advantage to off-chain lending followed by tokenization.

This is why debt assets should originate directly on-chain, rather than being issued off-chain and then tokenized. On-chain origination can reduce loan servicing costs, back-office structure costs, and improve accessibility. The real challenge is compliance and standardization, but teams are already working on these issues.

——Guy Wuollet, a16z crypto general partner

Stablecoins Will Trigger a Technology Upgrade Cycle for Bank Ledgers and Create New Payment Scenarios

Most banks still run on legacy software systems that modern developers can barely recognize: in the 1960s–70s, banks were early adopters of large software systems; second-generation core banking systems appeared around the 1980s–90s (such as Temenos's GLOBUS, InfoSys's Finacle). But these systems have gradually aged, and their upgrade pace lags far behind modern needs.

Therefore, the most critical core ledgers in the banking system—the databases recording deposits, collateral, and various financial obligations—often still run on mainframes, written in COBOL, and rely on batch file interfaces rather than APIs.

The vast majority of global assets are also stored in these same "decades-old" core ledgers. Although these systems have been battle-tested, are regulator-approved, and deeply embedded in complex business processes, they also significantly limit the pace of innovation.

For example, adding new features like real-time payments (RTP) often takes months or even years, and must navigate layers of technical debt and regulatory hurdles.

This is where stablecoins come into play. In recent years, stablecoins have found true product-market fit and entered the mainstream, and this year, traditional financial institutions' acceptance of stablecoins has reached new heights.

Stablecoins, tokenized deposits, tokenized treasuries, and on-chain bonds enable banks, fintech companies, and institutions to build new products and serve new customers. More importantly, they don't require rewriting those old but long-reliable core systems. Thus, stablecoins become a new path for institutional innovation.

——Sam Broner

The Internet Will Become the New "Bank"

As intelligent agents scale up, more and more commercial activities will no longer rely on user clicks but will be completed automatically in the background, requiring value transfer methods to change accordingly.

In a world where systems act based on "intent" rather than step-by-step instructions, when AI agents automatically move funds to recognize needs, fulfill obligations, or trigger outcomes, value must flow as quickly and freely as information. This is where blockchains, smart contracts, and new protocols come into play.

Smart contracts can already complete global dollar settlements in seconds. By 2026, new primitives like x402 will make such settlements programmable and responsive:

Agents can instantly and permissionlessly pay each other for data, GPU time, or API fees—no invoices, reconciliations, or batch processing required;

Developers can embed payment rules, limits, and audit trails directly into software updates—no need to integrate fiat systems, open merchant accounts, or connect to banks;

Prediction markets can self-settle in real time as events unfold—odds update, agents trade, and profits are globally cleared in seconds... no custodians or exchanges needed.

When value can move this way, "payment flows" will no longer be a separate operational layer but a network behavior: banks become part of the internet's underlying pipes, and assets become infrastructure.

If money becomes a "data packet" routable by the internet, then the internet not only supports the financial system—it becomes the financial system itself.

——Christian Crowley and Pyrs Carvolth, a16z crypto go-to-market team

Wealth Management Will Become a Service for Everyone

For a long time, personalized wealth management services were only available to high-net-worth clients, as providing customized advice and portfolio management across asset classes was expensive and complex. But as more assets are tokenized, crypto networks enable these strategies to be executed and rebalanced instantly with AI-generated advice and assistance, at near-zero cost.

This is not just "robo-advisory"; active management will become accessible to everyone, not just passive management.

In 2025, traditional financial institutions increased their allocation to crypto assets (directly or via ETPs), but that was just the beginning. In 2026, we will see more platforms focused on "wealth accumulation" (not just preservation)—especially fintechs (like Revolut, Robinhood) and centralized exchanges (like Coinbase) that can leverage their tech stack advantages.

Meanwhile, DeFi tools like Morpho Vaults can automatically allocate assets to lending markets with the best risk-adjusted returns, becoming a foundational yield allocation in portfolios. Holding other liquid assets in stablecoins instead of fiat, or in tokenized money market funds instead of traditional MMFs, further expands yield possibilities.

Finally, retail investors now have easier access to less liquid private market assets, such as private credit, pre-IPO companies, and private equity. Tokenization improves accessibility while maintaining necessary compliance and reporting requirements.

As various asset classes in a balanced portfolio (from bonds to stocks to private and alternative assets) are gradually tokenized, they can be automatically and intelligently rebalanced without interbank wire transfers.

——Maggie Hsu, a16z crypto go-to-market team

On Agents and AI

From "Know Your Customer" (KYC) to "Know Your Agent" (KYA)

The bottleneck of the agent economy is shifting from intelligence itself to identity.

In financial services, the number of "non-human identities" now exceeds human employees by a ratio of 96:1—yet these identities remain unaccepted "unbanked ghosts." The most lacking foundational capability today is KYA: Know Your Agent.

Just as humans need credit scores to get loans, AI agents need cryptographically signed credentials to transact—credentials that must bind the agent to its principal, behavioral constraints, and liability boundaries. Until this infrastructure emerges, merchants will continue to block agent access at the firewall level.

The industry that spent decades building KYC infrastructure now has only months to solve KYA.

——Sean Neville, Circle co-founder, USDC architect; Catena Labs CEO

We Will Use AI to Accomplish Substantive Research Tasks

As a mathematical economist, in January this year I still struggled to get consumer-grade AI models to understand my research process; by November, I could give the model abstract instructions as if guiding a PhD student... and sometimes get novel and correctly executed answers.

More broadly, we are seeing AI begin to be used for real research activities—especially in reasoning domains, where models not only assist discovery but can independently solve Putnam-level math problems (one of the world's toughest college math competitions).

It is still unclear which disciplines will benefit most and how. But I believe AI will foster and reward a new "polymathic" research style: able to hypothesize between ideas and quickly extrapolate from more exploratory intermediate results.
These answers may not be entirely accurate, but can still point in the right direction (at least in some topological sense).

In a sense, this is like leveraging the model's "hallucination ability": when models are "smart" enough, their random collisions in abstract space may generate meaningless content, but occasionally, as with nonlinear human thinking, trigger real breakthroughs.

Reasoning this way requires a new AI workflow style—not just agent-to-agent collaboration, but "agent-wrapping-agent": multilayer models evaluating early model attempts and continuously distilling truly valuable parts. I use this approach to write papers, others use it for patent searches, creating new art forms, or (unfortunately) designing new smart contract attacks.

But for such "wrapped reasoning agent clusters" to truly serve research, two problems must be solved: interoperability between models, and how to identify and fairly compensate each model's contribution—both of which may be solved by cryptography.

——Scott Kominers, a16z crypto research team; Harvard Business School professor

The "Invisible Tax" Facing Open Networks

The rise of AI agents is imposing an invisible tax on open networks and fundamentally shaking their economic foundation.

This disruption stems from the misalignment between the internet's "context layer" and "execution layer": at present, AI agents extract data from content sites (context layer) that rely on ad revenue to provide convenience to users, but systematically bypass the revenue sources (ads and subscriptions) that support this content.

To prevent the erosion of open networks (and the weakening of the content ecosystem that AI itself depends on), we need to deploy technical and economic mechanisms at scale: this may include a new generation of sponsored content models, micro-attribution systems, or other new funding allocation models.

Current AI licensing agreements have proven unsustainable—they often pay content providers only a fraction of the losses caused by AI's erosion of traffic.

Open networks need a new techno-economic framework that allows value to flow automatically. The most critical shift in the coming year is from static licensing to real-time, pay-per-use compensation models.

This means we need to test and scale systems—possibly based on blockchain-supported nanopayments and granular attribution standards—so that every entity contributing information to a successful agent task is automatically rewarded.

——Liz Harkavy, a16z crypto investment team

On Privacy (and Security)

Privacy Will Become the Most Important "Moat" in Crypto

Privacy is the key capability driving global finance onto the blockchain. It is also a feature almost all existing blockchains lack. For most chains, privacy has long been just an "afterthought."

But now, privacy alone is enough to distinguish one chain from all others. More importantly: privacy can create chain-level lock-in effects—a "privacy version of network effects," especially when performance competition can no longer differentiate chains.

Because of cross-chain protocols, as long as everything is public, migrating from one chain to another is almost costless. But once privacy is added, things change dramatically: moving tokens cross-chain is easy, but moving "secrets" cross-chain is hard.

Any process of moving from a private chain to a public chain allows observers of the blockchain, mempool, or network traffic to infer your identity. Migration between different private chains also leaks various metadata, such as time or amount correlations, making tracking easier.

In contrast, new chains lacking differentiation and whose fees will be driven to zero in competition (because blockspace becomes highly commoditized), privacy chains can instead form stronger network effects.

The reality is: a "general-purpose chain" without a thriving ecosystem, killer apps, or distribution advantages has little reason to attract users or developers, let alone inspire loyalty.

When users are on public chains, as long as chains can freely interact, which chain to join doesn't really matter. But once users join a private chain, the choice of chain becomes very important—because once joined, they are less willing to migrate and expose risks.

This creates a "winner-takes-most" landscape.

And because privacy is extremely critical for most real-world applications, ultimately only a few privacy chains may dominate most of the crypto economy.

——Ali Yahya, a16z crypto general partner

The (Near) Future of Messaging Is Not Just Quantum-Resistant, But Decentralized

On the road to the quantum computing era, many encryption-dependent communication apps (Apple, Signal, WhatsApp) have done a lot of cutting-edge work. But the problem is: all mainstream communication tools today rely on privately operated servers by single organizations.

These servers are vulnerable points for government shutdowns, backdoors, or data surrender demands.

If a country can directly shut down servers; if a company holds the server keys; or if there simply exists a "private server"... then what is the point of quantum-level encryption?

Private servers require "trust me"; but no servers mean "you don't have to trust me."

Communication does not need a centralized company in the middle. What we need is an open protocol that requires trusting no one.

To achieve this, the network must be decentralized: no private servers; no single app; all code open source; best-in-class encryption (including quantum resistance).

In an open network, no individual, company, nonprofit, or country can take away our ability to communicate. Even if a country or company shuts down an app, 500 new versions will appear the next day.

Even if one node is shut down, new nodes will immediately join to replace it, incentivized by mechanisms like blockchain.

When people control information with their own keys, just as they control their "money," everything changes. Apps may come and go, but users always control messages and identity—the user owns the message, not the app.

This is not just about quantum resistance or encryption, but about ownership and decentralization.

Without these two, we are just building an "unbreakable but still stoppable" encryption.

——Shane Mac, XMTP Labs co-founder and CEO

"Secrets-as-a-Service"

Behind every model, every agent, every automated system, lies the same thing: data.

But today, most data pipelines—model inputs and outputs—are opaque, mutable, and unauditable.

This may suffice for some consumer applications, but for industries handling sensitive data (like finance and healthcare), such mechanisms are far from enough.

This is also the main obstacle preventing institutions from fully tokenizing real-world assets.

So, how can we achieve secure, compliant, autonomous, and globally interoperable innovation while maintaining privacy?

We need to start with data access control: who controls sensitive data? How does data move? Who (or what system) can access it?

Without data access control, anyone wishing to protect privacy must rely on centralized services or build complex systems themselves—which is not only time-consuming and expensive, but also hinders traditional financial institutions from fully leveraging the advantages of on-chain data management.

As intelligent agents begin to browse, trade, and make decisions autonomously, users and institutions need not "best-effort trust," but cryptographic-level guarantees.

Therefore, we need "secrets-as-a-service": new technologies providing programmable, native data access rules; client-side encryption;

decentralized key management—clearly defining who can decrypt what data, under what conditions, for how long... all enforced on-chain.

Combined with verifiable data systems, "secrets" will become the underlying public infrastructure of the internet, not just an "application-level patch" added later.

Privacy will become part of the infrastructure, not an add-on feature.

——Adeniyi Abiodun, Mysten Labs co-founder and Chief Product Officer

From "Code Is Law" to "Spec Is Law"

Recent DeFi attacks, even on mature protocols with years of battle-testing, strong teams, and strict audits, have exposed an unsettling reality: today's security practices are still essentially empirical and "case-by-case."

To bring DeFi security to maturity, we must shift from bug patterns to design-level properties, from "best-effort" to "principled" systematic approaches:

Static / Pre-deployment Security (testing, auditing, formal verification)

The future focus is on systematically proving global invariants, not just verifying a few manually selected local properties.
Multiple teams are now building AI-assisted proof tools to help write specs, propose invariants, and automate much of the previously labor-intensive, costly proof engineering work.

Dynamic / Post-deployment Security (runtime monitoring, runtime enforcement, etc.)

After going on-chain, these invariants can become live guardrails for the system: the last line of defense.
These guardrails will be encoded as runtime assertions, requiring every transaction to meet relevant security conditions.

In other words, we no longer assume "all bugs are caught before deployment," but let the code itself enforce core security properties, automatically rolling back any transaction that violates them.

This is not just theoretical, but practical

In fact, almost every past attack could have triggered these checks at execution, stopping the hack.
Thus, the once-popular "code is law" concept is evolving into: "spec is law."

Even entirely new attack methods must satisfy the same set of security properties in system design; thus, the attack surface is compressed to only very small or extremely difficult possibilities.

——Daejun Park, a16z crypto engineering team

On Other Industries and Applications

Prediction Markets Will Become Bigger, Broader, and Smarter

Prediction markets have gone mainstream. In the coming year, as crypto and AI converge, they will become larger in scale, broader in coverage, and more intelligent, while also bringing new challenges for builders to solve together.

First, more types of contracts will be listed. This means that in the future, we will get real-time odds not just for major elections or geopolitical events, but for all sorts of fine-grained outcomes and complex event combinations. As these new contracts continuously disclose information and integrate into the news ecosystem (already happening), society will have to face a question: how do we balance the value of this information, and how do we design more transparent, auditable prediction systems?

Cryptography can provide tools for this.

To handle a larger volume of prediction contracts, we need new "truth alignment" mechanisms to advance contract settlement. Centralized platforms' adjudication mechanisms (did an event occur? how to confirm?) are important, but controversial cases like the Zelensky lawsuit market and Venezuela election market have exposed their limitations.

Therefore, to expand the scale and application value of prediction markets,
new decentralized governance mechanisms and LLM oracles will become important tools for resolving disputes and achieving truth.

The possibilities brought by AI are not limited to LLMs. AI agents can autonomously trade on prediction platforms, scan the world for signals, and seek short-term advantages. This helps us discover new ways of thinking and better predict "what will happen next." (Projects like Prophet Arena have already shown early excitement in this area.)

Beyond serving as "advanced political analysts" to be queried, the emergence strategies of AI agents can even help us reverse-engineer the fundamental predictive factors of complex social events.

Will prediction markets replace polls? No, they will make polls better.

Poll data can even become input for prediction markets. As a political economist, what excites me most is seeing prediction markets and a healthy, diverse polling ecosystem work together. But to achieve this, we need new technology: AI can improve the survey experience; cryptography can prove respondents are real people, not bots, and bring more innovation.

——Andy Hall, a16z crypto research advisor; Stanford University professor of political economy

The Rise of "Staked Media"

The traditional media model (especially the "objectivity" assumption) has shown cracks. The internet allows everyone to speak, and more and more industry insiders, practitioners, and builders are expressing their views directly to the public. Ironically, audiences often respect them not "despite their interests," but because they have interests.

The real new change is not social media, but: crypto tools allow people to make public, verifiable commitments.

When AI reduces the threshold for content creation to nearly zero—any perspective, any identity (real or fictional) can be infinitely replicated—just "saying something" is no longer enough to build trust.

Tokenized assets, programmable lockups, prediction markets, and on-chain history provide a more solid foundation for trust:

Commentators can express opinions and prove they have "put real money on the line";

Podcast hosts can lock up tokens to show they won't "pump and dump";

Analysts can tie predictions to publicly settled markets, building an auditable record.

This is what I call the early form of "staked media": a new type of media that embraces the "skin in the game" concept and provides verifiable evidence.

In this model, credibility no longer comes from "pretending to be neutral," nor from "unsubstantiated claims," but from publicly verifiable risk-taking.

Staked media will not replace existing media, but will complement the current ecosystem.

It provides a new signal: not "trust me, I'm neutral," but "see what risks I'm willing to take, and you can verify if I'm telling the truth."

——Robert Hackett, a16z crypto editorial team

Crypto Provides New Primitives for the World Beyond Blockchain

For years, SNARKs (cryptographic proofs of verifiable computation) were almost exclusively used in blockchain. The reason is simple: generating proofs was too expensive—possibly 1,000,000 times more costly than direct computation.

It made sense when costs could be spread across thousands of verifiers, but was nearly impossible in other scenarios.

All this is about to change.

By 2026, zkVM provers will bring costs down to about 10,000 times, with memory usage of just a few hundred MB: fast enough to run on a phone, cheap enough to deploy anywhere.

Why might 10,000x be a "magic number"? Because the parallel capability of a high-end GPU is about 10,000 times that of a laptop CPU.

By the end of 2026, a single GPU will be able to generate proofs in real time for CPU execution.

This will unlock the long-standing vision in old papers: verifiable cloud computing.

If your workload already runs on cloud CPUs—because you don't need much computing power, lack GPU capability, or for historical reasons—

in the future, you will be able to get cryptographic proofs of computational correctness at reasonable cost.

The prover itself is already GPU-optimized, and your code doesn't need to change.

——Justin Thaler, a16z crypto research team; Georgetown University associate professor of computer science

On Future Building

Trading Is Just a "Transit Station," Not the Endgame for Crypto Companies

Today, except for stablecoins and a few core infrastructures, almost all well-run crypto projects have shifted to trading businesses, or are preparing to. If "all crypto companies eventually become trading platforms," what will the endgame look like?

A large number of players doing the same thing will squeeze each other, leaving only a handful of winners.

Those who pivot to trading too early or too quickly may miss the opportunity to build more defensible, sustainable businesses.

I fully understand founders' constant exploration to make their financial models work, but chasing "seemingly immediate product-market fit (PMF)" also comes at a cost.

Especially in crypto, the unique dynamics of token mechanisms and speculative culture make it easier for founders to pursue "instant gratification" and overlook deeper product issues.

In a sense, this is a "marshmallow test." Trading itself is not a problem; it is an important market function. But it doesn't have to be the endgame.

Those founders who truly focus on the "product" part of PMF are often the ultimate big winners.

——Arianna Simpson, a16z crypto general partner

Only When Legal Architecture Finally Aligns with Technical Architecture Can Blockchain Unleash Its Full Potential

Over the past decade, one of the biggest obstacles to building blockchain networks in the US has been legal uncertainty.

Securities law has been extended and selectively enforced, forcing founders into a regulatory framework originally designed for "companies," not "networks."

For years, "reducing legal risk" replaced "product strategy"; engineers were replaced by lawyers.

This dynamic has led to many strange distortions:

Founders are told to avoid transparency;

Token allocations become legally arbitrary and unnatural;

Governance becomes performative;

Organizational structures prioritize legal risk avoidance;

Tokens are forced to be designed without economic value or business models;

Worse, projects that follow the rules less often move faster.

But now, US crypto market structure legislation is closer to passing than ever before, and is expected to eliminate these distortions next year.

Once passed, the legislation will: incentivize transparency; establish clear standards; and replace today's "enforcement roulette" with clear, structured fundraising, token issuance, and decentralization paths.

After the GENIUS Act passed, stablecoin growth exploded; the changes brought by crypto market structure legislation will be even more profound—this time targeting the network itself.

In other words, this kind of regulation will allow blockchain networks to operate as they were meant to: open, autonomous, composable, credibly neutral, and decentralized.

——Miles Jennings, a16z crypto policy team; General Counsel

[Original Link]