This week, a16z released its annual "Big Ideas" report, featuring insights from partners across its Apps, American Dynamism, Bio, Crypto, Growth, Infra, and Speedrun teams. Below are 17 observations on 2026 industry trends from multiple a16z crypto partners (including several guest contributors) — covering agents and AI, stablecoins, tokenization and finance, privacy and security, extending to prediction markets, SNARKs, and other applications, and finally discussing the direction of industry building.
On Stablecoins, RWA Tokenization, Payments, and Finance
1. Better, More Flexible Stablecoin On/Off Ramps
Last year, stablecoin transaction volume was estimated at $46 trillion, continuing to hit new historical peaks. To put this in perspective, this scale is more than 20 times the transaction volume of PayPal, nearly 3 times that of Visa, one of the world's largest payment networks, and is quickly approaching the transaction volume of the US Automated Clearing House (ACH) (the electronic network that processes financial transactions like direct deposits in the US).
Today, sending a stablecoin takes less than a second, with a fee of less than one cent. But the core problem that remains unsolved is: how to connect these "digital dollars" to the financial system people use every day — the stablecoin on/off ramps.
A new generation of startups is filling this gap, promoting the integration of stablecoins with more widespread payment systems and local currencies: some use cryptographic proof technology to allow users to privately convert local currency balances into digital dollars; some integrate regional networks, leveraging features like QR codes and real-time payment channels for inter-bank transfers; others are building a truly interoperable global wallet layer and card issuance platform, allowing users to spend stablecoins directly at everyday merchants. These solutions collectively expand participation in the digital dollar economy and may accelerate stablecoins becoming a mainstream payment tool.
As on/off ramps mature and digital dollars directly integrate with local payment systems and merchant tools, new application scenarios will emerge: cross-border workers can receive payments in real-time, merchants can receive global dollars without a bank account, and applications can settle value instantly with global users. At that point, stablecoins will completely transform from a "niche financial tool" to an "internet-native settlement layer."
— Jeremy Zhang, a16z Crypto Engineering Team
2. Reimagining RWA Tokenization and Stablecoins with a "Crypto-Native Mindset"
Currently, banks, fintech companies, and asset management institutions show strong interest in "bringing traditional assets on-chain," involving US stocks, commodities, indices, and other traditional assets. But as more traditional assets go on-chain, their tokenization process often falls into the "skeuomorphic trap" — limited to the existing form of the real-world asset, failing to leverage the advantages of crypto-native properties.
Synthetic derivatives like perpetual futures not only provide deeper liquidity but are also easier to implement. Meanwhile, the leverage mechanism of perpetual contracts is easy to understand, which is why I believe it is the crypto-native derivative with the highest "product-market fit." Additionally, emerging market stocks are one of the most suitable asset classes for "perpetualization" (the liquidity of the "zero-day expiry options" market for some stocks already exceeds that of the spot market; perpetualizing them would be a highly valuable experiment).
This is essentially a choice between "fully on-chain vs. tokenization," but regardless, in 2026 we will see more "crypto-native" RWA tokenization solutions.
Similarly, stablecoins entered the mainstream market in 2025, with outstanding issuance continuously growing; in 2026, the stablecoin field will shift from "simple tokenization" to "innovative issuance models." Currently, stablecoins lacking a sound credit infrastructure resemble "narrow banks" — holding only highly secure, specific liquid assets. Although the narrow bank model is reasonable, in the long run, it is difficult to become the core pillar of the on-chain economy.
Currently, several new asset management institutions, asset managers, and protocols are exploring "on-chain asset-backed lending based on off-chain collateral," but these loans are usually initiated off-chain first and then tokenized. I believe the value of tokenization in this model is very limited, only serving users already in the on-chain ecosystem. Therefore, debt assets should be "initiated directly on-chain," not "tokenized after being initiated off-chain" — on-chain initiation can reduce loan servicing costs, back-office architecture costs, and improve accessibility. Although compliance and standardization remain challenges, developers are actively solving these problems.
— Guy Wuollet, a16z Crypto General Partner
3. Stablecoins Drive Bank Ledger Upgrades, Unlocking New Payment Scenarios
The software used by most banks today is almost "unrecognizable" to modern developers: in the 1960s-70s, banks were early adopters of large-scale software systems; in the 1980s-90s, second-generation core banking software emerged (like Temenos' GLOBUS, Infosys' Finacle). But this software has aged, and the pace of updates is extremely slow — today, the banking industry (especially core ledger systems, the critical databases recording deposits, collateral, and other liabilities) still often relies on mainframes running COBOL, using batch file interfaces rather than APIs.
The vast majority of global assets reside in these "decades-old core ledgers." Although these systems are time-tested, regulator-approved, and deeply integrated into complex banking scenarios, they also severely hinder innovation: adding key functions like real-time payments (RTP) can take months or even years, while dealing with layers of technical debt and regulatory complexity.
This is where the value of stablecoins lies: in the past few years, stablecoins not only achieved "product-market fit" and entered the mainstream, but in 2025 traditional finance (TradFi) institutions "fully embraced" stablecoins. Stablecoins, tokenized deposits, tokenized treasury bonds, and on-chain bonds allow banks, fintech companies, and financial institutions to develop new products and serve new customers — more importantly, without forcing these institutions to refactor "aging but decades-stable" legacy systems. Stablecoins provide financial institutions with a "low-risk innovation path."
— Sam Broner
4. The Internet Will Become the "New Generation Bank"
As AI agents (AI Agent) become widely adopted, more business activities will be completed "automatically in the background" (rather than relying on user clicks), which means the "way value (money) flows" must change accordingly.
In a world where "systems act on intent" (rather than step-by-step instructions) — for example, an AI agent identifies a need, fulfills an obligation, or triggers an outcome and automatically transfers value — the flow of value needs to have "the same speed and freedom as the current flow of information." And blockchain, smart contracts, and new protocols are key to achieving this goal.
Today, smart contracts can complete global dollar payments in seconds; in 2026, emerging foundational protocols like x402 will make "settlement programmable and responsive": agents can pay for data, GPU computing power, or API calls instantly and permissionlessly, without invoicing, reconciliation, or batch processing; developers can release software updates with built-in payment rules, limits, and audit trails, without fiat integration, merchant onboarding, or reliance on banks; prediction markets can "settle automatically in real-time" as events develop — odds updating, agent trading, global payout in seconds, without custodians or exchanges.
When value can flow in this way, the "payment process" will no longer be a separate operational layer, but will become a "network behavior": banks will integrate into the internet infrastructure, assets will become infrastructure. If money can flow like "internet-routable data packets," the internet will no longer "support the financial system," but will "itself become the financial system."
— Christian Crowley, Pyrs Carvolth, a16z Crypto Go-to-Market Team
5. Wealth Management Services Accessible to All
Traditionally, personalized wealth management services were only available to a bank's "high-net-worth clients": customized advice and portfolio adjustments across asset classes were costly and operationally complex. But as more asset classes are tokenized, crypto channels enable personalized strategies with "AI recommendation + assisted decision-making" to be "executed instantly and rebalanced at low cost."
This is far more than "robo-advisors": everyone can get "active portfolio management" (not just passive management). In 2025, traditional financial institutions already increased their allocation to crypto assets in investment portfolios (banks suggested allocating 2%-5% directly or through exchange-traded products (ETP)), but this is just the beginning; in 2026, we will see the rise of platforms "aimed at wealth accumulation" (not just wealth preservation) — fintech companies like Revolut, Robinhood, and centralized exchanges like Coinbase will capture this market with their tech stack advantages.
Meanwhile, DeFi tools like Morpho Vaults can automatically allocate assets to the lending market with the "optimal risk-adjusted return," providing a "core yield allocation" for the portfolio. Holding idle working capital in stablecoins (not fiat) and in tokenized money market funds (not traditional money funds) can further expand the yield space.
Finally, tokenization, while meeting compliance and reporting requirements, also makes it easier for retail investors to access "illiquid private market assets" (like private credit, Pre-IPO company equity, private equity). When all types of assets in a balanced portfolio (from bonds to stocks, to private and alternative assets) are tokenized, rebalancing can be done automatically without wire transfers.
— Maggie Hsu, a16z Crypto Go-to-Market Team
On Agents and AI
6. From KYC to KYA
Currently, the bottleneck of the "agent economy" is shifting from "intelligence level" to "identity recognition."
In financial services, the number of "non-human identities" (like AI agents) has reached 96 times that of human employees, but these identities are still "ghosts unable to access the banking system" — the core missing foundational capability is KYA (Know Your Agent).
Just as humans need credit scores to get loans, agents need "cryptographic signed credentials" to complete transactions — credentials need to be linked to the agent's "principal," "constraints," and "liability attribution." If this issue is not resolved, merchants will continue to block agents at the firewall level. The industry that built KYC infrastructure over decades now needs to solve the KYA challenge within months.
— Sean Neville, Co-founder of Circle, Architect of USDC, CEO of Catena Labs
7. AI Will Empower "Substantive Research Tasks"
As a mathematical economist, in January 2025, I could hardly get consumer-grade AI models to understand my workflow; but by November, I could send abstract tasks to AI models like instructing a PhD student — sometimes they even return "innovative and correctly executed" results. Beyond my personal experience, AI applications in research are gradually becoming popular, especially in the "reasoning field": AI not only directly assists discovery but can also "autonomously solve Putnam Mathematical Competition problems" (Putnam problems, considered the world's most difficult university-level math exam).
What still needs exploration is: in which fields are these research assistance functions most valuable, and how are they specifically applied. But I predict AI will spawn and reward a "new polymath research mode" — one that focuses more on the ability to "speculate on connections between ideas" and "quickly derive from highly speculative answers." These answers may not be accurate but can point in the right direction (at least within a specific logical framework). Ironically, this is equivalent to "harnessing the power of model hallucination": when models are smart enough, giving them abstract exploration space may produce meaningless content, but it may also lead to key breakthroughs — just as humans are most creative in a "non-linear, non-explicit goal-oriented" state.
To achieve this reasoning mode, we need to build "new AI workflows" — not just "interaction between agents," but "agents nested within agents": multi-layer models assist researchers in evaluating "the methods of previous models," gradually filtering effective information and eliminating invalid content. I have used this method to write papers; others use it for patent searches, creating new art, and even (regrettably) discovering new attack vectors for smart contracts.
But note: to run "nested reasoning agent clusters" to support research, two key issues need to be solved — "interoperability" between models, and "identifying and fairly compensating the contributions of each model" — and crypto technology can provide solutions for this.
— Scott Duke Kominers, a16z Crypto Research Team, Professor at Harvard Business School
8. The "Invisible Tax" on the Open Web
The rise of AI agents is imposing an "invisible tax" on the open web, fundamentally undermining its economic foundation. This destruction stems from the increasing misalignment between the internet's "context layer" and "execution layer": currently, AI agents extract data from "ad-supported websites" (context layer), providing convenience to users while systematically bypassing the "revenue sources that support content creation" (like ads, subscriptions).
To prevent the decline of the open web (while protecting the diverse content that "fuels AI"), large-scale deployment of "technology + economic" solutions is needed, such as "next-generation sponsored content," "micro-attribution systems," or other new funding models. Existing AI licensing agreements are essentially "financially unsustainable stopgaps" — compensation for content providers is often only a fraction of the revenue they lose due to AI diverting traffic.
The open web needs a "new techno-economic model for automatic value flow." The key shift in 2026 is: from "static licensing" to "real-time, pay-per-use." This means testing and scaling systems based on "blockchain-based micropayments + precise attribution standards" — automatically rewarding "all entities that contribute to the agent completing a task."
— Elizabeth Harkavy, a16z Crypto Investment Team
On Privacy and Security
9. Privacy Will Become Crypto's "Most Important Moat"
Privacy is a key prerequisite for "global finance going on-chain," but currently almost all blockchains lack this functionality — for most chains, privacy is only an "afterthought."
Today, "privacy capability" is enough to make one chain stand out from the competition; more importantly, privacy can "create chain lock-in effects," which can be called "privacy network effects" — especially when "competing on performance alone is no longer enough."
Thanks to cross-chain bridge protocols, migrating between different chains is very easy as long as the data is public; but once privacy is involved, the situation is completely different: "It's easy to transfer tokens cross-chain, but hard to transfer secrets cross-chain." When entering or exiting a "privacy zone," observers of the chain, mempool, or network traffic may identify the user; transferring assets between a "privacy chain and a public chain" or even "between two privacy chains" leaks metadata like transaction timing and amount correlation, increasing the risk of user tracking.
Currently, many "undifferentiated new chains" see fees approach zero due to competition (on-chain space is essentially homogeneous); whereas blockchains with privacy capabilities can build stronger "network effects." The reality is: if a "general-purpose chain" lacks a thriving ecosystem, killer apps, or unique distribution advantages, users and developers have no reason to choose it, build on it, let alone loyalty.
On public chains, users can easily trade with users on other chains, and the choice of chain is irrelevant; but on privacy chains, "which chain you choose" is crucial — once you join a privacy chain, users are reluctant to migrate for fear of exposure, creating a "winner-take-all" pattern. Since privacy is a rigid demand in most real-world scenarios, a few privacy chains may dominate the crypto space.
— Ali Yahya, a16z Crypto General Partner
10. The (Near) Future of Messaging: Not Just Post-Quantum, But Decentralized
As the world prepares for the "quantum computing era," encryption-based instant messaging applications like Apple's, Signal, and WhatsApp have taken the lead, with significant results. But the problem is: all mainstream communication tools rely on "privately operated servers by a single organization" — these servers are easy targets for governments to "shut down, implant backdoors, or forcibly obtain private data."
If a country can shut down servers, companies hold private server keys, or even companies themselves own private servers, then what's the point of "post-quantum encryption"? Private servers require users to "trust me," while "no private servers" means "you don't need to trust me." Communication doesn't need an intermediary (a single company), it needs "open protocols that don't require trusting any entity."
The path to achieve this is "decentralization of the network": no private servers, no single application, fully open-source code, using "top-tier cryptography" (including post-quantum threats). In an open network, no individual, company, non-profit, or country can deprive people of the right to communicate — even if a country or company shuts down one application, 500 new versions will appear the next day; even if a node is shut down, the economic incentives brought by technologies like blockchain will have new nodes immediately take its place.
When people "control messages with keys" (just like controlling money), everything changes: applications may iterate, but users always control their messages and identity — even if they stop using a certain application, message ownership still belongs to the user.
This is not just "post-quantum" and "encryption," but also "ownership" and "decentralization." Without these two, what we build is merely "unbreakable encryption that can be shut down at any time."
— Shane Mac, Co-founder & CEO, XMTP Labs
11. "Secrets-as-a-Service"
Behind every model, agent, and automated system lies a simple foundation: data. But most data transmission channels today — whether data input into models or output from models — are opaque, tamperable, and unauditable. This might not matter much for some consumer applications, but many industries and users, like finance and healthcare, require companies to protect sensitive data privately; simultaneously, this is a major obstacle in the current institutional push for real-world asset tokenization.
So, how to achieve secure, compliant, autonomous, and globally interoperable innovation while ensuring privacy? There are many solutions, here I focus on "data access control": who controls sensitive data? How does data flow? Who (or what entity) has the right to access data?
Without a data access control mechanism, any entity wanting to protect data confidentiality today must either rely on centralized services or build custom systems — an approach that is time-consuming, costly, and hinders traditional financial institutions and other entities from fully utilizing the functions and advantages of on-chain data management. Furthermore, as agent systems begin to autonomously browse information, complete transactions, and make decisions, users and institutions across industries need "crypto-grade guarantees," not "best-effort trust promises."
That's why I believe we need "Secrets-as-a-Service": using new technologies to achieve programmable native data access rules, client-side encryption, and decentralized key management — clearly specifying who can decrypt what data, under what conditions, for how long, with all rules enforced on-chain. Combined with verifiable data systems, "data confidentiality protection" will become part of the internet's basic public infrastructure, not a patch applied after the fact at the application layer, truly making privacy core infrastructure.
— Adeniyi Abiodun, Chief Product Officer & Co-founder, Mysten Labs
12. From "Code is Law" to "Specification is Law"
Recent DeFi hacking incidents affected protocols that were long-tested, had strong teams, rigorous audit processes, and had been running stably for years. These events reveal an unsettling reality: current mainstream security practices largely remain at the level of "empirical judgment" and "case-by-case handling."
To mature DeFi security, two major shifts are needed: from "patching vulnerability patterns" to "guaranteeing design-level properties," from "best-effort protection" to "principled systematic protection." This can be approached from two aspects:
Static / Pre-deployment phase (testing, auditing, formal verification): Systematically prove "global invariants" (the core rules the entire system always follows), not just verify "manually selected local rules." Several teams are developing AI-assisted proof tools that can help write specifications, propose invariant hypotheses, and significantly reduce the proof engineering work previously done manually — work that was extremely costly and difficult to scale in the past.
Dynamic / Post-deployment phase (runtime monitoring, runtime enforcement, etc.): The above "invariant rules" can be transformed into real-time protection barriers, serving as the last line of defense. These barriers are directly encoded as "runtime assertions," and all transactions must satisfy the assertion conditions to execute.
This way, we no longer need to assume "all vulnerabilities have been fixed," but instead use the code itself to enforce critical security properties — any transaction violating these properties is automatically rejected.
This is not theoretical. In fact, almost all hacking attacks to date would have triggered such security checks during execution, potentially stopping the attack. Therefore, the once-popular "code is law" concept is gradually evolving into "specification is law": even when facing new attacks, attackers must abide by the core security properties that maintain system integrity, leaving only attack vectors with minimal impact or extremely high implementation difficulty.
— Daejun Park, a16z Crypto Engineering Team
On Other Industries and Applications
13. Prediction Markets: Larger Scale, Broader Coverage, Higher Intelligence
Prediction markets have entered the mainstream. In 2026, with deeper integration with crypto technology and AI, they will further expand in scale, coverage, and intelligence — but will also bring new important challenges for developers that need urgent resolution.
First, prediction markets will list more contracts. This means we can get real-time odds not only for "major elections, geopolitical events" but also for various niche outcomes and complex cross-event scenarios. As these new contracts continuously release information and integrate into the news ecosystem (a trend already visible), society will face important issues: how to balance the value of this information? How to improve the transparency, auditability, etc., of prediction markets through optimized design (crypto technology can achieve this)?
To handle the significant increase in the number of contracts, new "consensus mechanisms" are needed to settle contracts. Centralized platform settlement (confirming whether an event actually happened, how to verify) is important, but controversial cases like the "Zelensky lawsuit market" and "Venezuela election market" expose its limitations. To solve these edge cases and expand prediction markets into more practical scenarios, new decentralized governance mechanisms and large language model (LLM) oracles can assist in determining the truth of disputed outcomes.
Beyond LLM oracles, AI brings more possibilities for prediction markets. For example, AI agents trading on prediction platforms can widely gather various signals to gain short-term trading advantages, thereby providing new ideas for understanding the world and predicting future trends (projects like Prophet Arena have shown potential in this area). These agents can not only serve as "senior political analysts" for people to consult for insights, but by analyzing their autonomously formed strategies, we can also help discover the core factors influencing complex social events.
Will prediction markets replace polls? The answer is no. Instead, they can improve the quality of polls (polling information can also be fed into prediction markets). As a political scientist, what I most look forward to is the synergistic development of prediction markets and a "rich and active polling ecosystem" — but this requires new technology: AI can optimize the survey experience; crypto technology can provide new ways to prove that poll respondents are real humans, not bots, etc.
— Andrew Hall, a16z Crypto Research Advisor, Professor of Political Economy, Stanford University
14. The Rise of Staked Media
The traditional media model touted "objectivity," but its drawbacks have long been apparent. The internet gave everyone a voice, and now more practitioners, practitioners, and builders are communicating directly with the public — their perspective reflects their "stake" in the world. Ironically, the audience respects them, often not "despite their having a stake," but "precisely because they have a stake."
The new change in this trend is not the rise of social media, but the "emergence of crypto tools" — tools that allow people to make "publicly verifiable commitments." As AI makes generating vast amounts of content cheaper and easier (generating content from any perspective, any identity — real or not), relying solely on the words of humans (or bots) is no longer sufficient for credibility. Tokenized assets, programmable locking, prediction markets, and on-chain history provide a firmer foundation for trust: commentators can prove they "practice what they preach" (backing their views with capital); podcast hosts can lock tokens, proving they won't opportunistically change their stance or "pump and dump"; analysts can tie predictions to "publicly settled markets," forming an auditable track record.
This is the early form of what I call "staked media": media that not only embraces the idea of "having skin in the game" but also provides tangible evidence. In this model, credibility comes neither from "pretending to be neutral" nor from "unfounded claims," but from "publicly transparent, verifiable commitment of interest." Staked media won't replace other forms of media but will complement the existing media ecosystem. It sends a new signal: no longer "trust me, I'm neutral," but "this is the risk I'm willing to take, this is how you can verify I'm telling the truth."
— Robert Hackett, a16z Crypto Editorial Team
15. Crypto Provides "New Foundational Components Beyond Blockchain"
For years, SNARKs — a cryptographic proof technology that verifies computation results without re-executing the calculation — were largely confined to blockchain applications. The main reason was "prohibitively high cost": the workload required to generate a proof of computation could be 1 million times that of directly performing the calculation. This technology only made sense in scenarios where the "cost could be amortized across thousands of verification nodes" (like blockchain); in other scenarios, it was impractical.
But this is about to change. In 2026, the cost of zero-knowledge virtual machine (zkVM) provers will drop to about 10,000x (i.e., the workload to generate a proof is 10,000 times that of direct computation), with memory footprint of only a few hundred megabytes — fast enough to run on mobile phones, and cheap enough for widespread application. The reason 10,000x might be the "critical threshold" is partly because: the parallel processing power 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 of CPU execution in real-time."
This will realize the vision proposed in old research papers: "verifiable cloud computing." If you need to run CPU workloads in the cloud due to "insufficient compute for GPU processing," "lack of technical capability," "legacy system limitations," etc., in the future you can just pay a reasonable extra cost to get a "cryptographic proof of computational correctness." Provers are already optimized for GPUs, your code can be used without additional adaptation.
— Justin Thaler, a16z Crypto Research Team, Associate Professor of Computer Science, Georgetown University
On Industry Building
16. Trading Business: A "Pit Stop" for Crypto Companies, Not the "Finish Line"
Today, apart from the stablecoin field and some core infrastructure companies, almost all performing crypto companies have either pivoted to trading business or are in the process of pivoting. But if "all crypto companies become trading platforms," where does it end? A large number of companies crowding into the same track will not only distract users but also lead to a result of "a few giants monopolizing, most companies eliminated." This means that companies that pivot to trading business too quickly will miss the opportunity to build "more competitive, more sustainable business models."
I fully understand the founders'初衷 to achieve business profitability, but "pursuing short-term product-market fit" also comes at a cost. This problem is particularly acute in crypto: the unique dynamics related to token characteristics and speculative attributes make it easy for founders to choose the "instant gratification" path in the process of "finding product-market fit" — this is essentially similar to the "marshmallow test" (testing delayed gratification ability).
There's nothing wrong with the trading business itself; it's an important market function, but it shouldn't be the "ultimate goal" of a company. Founders who focus on the "product essence of product-market fit" are ultimately more likely to become industry winners.
— Arianna Simpson, a16z Crypto General Partner
17. Unleashing Blockchain's Full Potential: When Legal Architecture Finally Matches Technical Architecture
Over the past decade, one of the biggest obstacles to building blockchain networks in the US has been "legal uncertainty." The expanded scope and uneven enforcement of securities laws forced founders into regulatory frameworks designed for "companies, not networks." For years, "avoiding legal risk" replaced "product strategy," and the importance of engineers gave way to lawyers.
This situation led to many distortions: founders were advised to avoid transparency; token distribution became arbitrary from a legal perspective; governance became a formality; organizational structures were "primarily aimed at avoiding legal risk"; token designs deliberately "avoided carrying economic value" or "had no business model." Worse, those crypto projects that "disregarded the rules and operated in gray areas" often developed faster than "honest and compliant" builders.
But now, the US government is closer than ever to passing "crypto market structure regulatory legislation" — legislation that could eliminate all the above distortions in 2026. If passed, it would incentivize companies to increase transparency, establish clear standards, and replace "random enforcement" with "clear, structured financing, token issuance, and decentralization paths." Previously, after the GENIUS Act passed, stablecoin issuance grew significantly; but crypto market structure-related legislation will bring about more significant change — this change will focus on "blockchain networks."
In other words, this type of regulation will allow blockchain networks to "truly operate as networks": open, autonomous, composable, credibly neutral, and decentralized.
— Miles Jennings, a16z Crypto Policy Team, General Counsel
