8 big ideas for 2026

a16z recently released its list of “big ideas” tech builders may tackle in the year ahead, according to partners across the Apps, American Dynamism, Bio, Crypto, Growth, Infra, and Speedrun teams.

So below is a selection of some of the big ideas from various crypto team members (plus a few guest contributors) for what’s ahead — on topics ranging from agents and AI; stablecoins, tokenization, and finance; privacy and security; and prediction markets and other applications. For more of what we’re excited about for 2026, read the full post.

On building

1. Trading as a way station — not the last stop — for crypto businesses

It seems like every crypto company that’s doing well today, outside of stablecoins and some core infrastructure, has pivoted to or is pivoting to trading. But if “every crypto company becomes a trading platform,” then where does that leave everyone? Having so many players all doing the same thing cannibalizes mindshare for the many, and leaves just a few big winners. This means those that pivoted too quickly to trading missed the opportunity to build a more defensible, more durable business.

While I have a lot of empathy for all the founders out there trying to make their business financials work, chasing the immediate sense of product-market fit has costs, too. This problem is particularly an issue in crypto, where unique dynamics around tokens and speculation can lead founders down the immediate-gratification path on their journey to finding product-market fit. It’s a kind of marshmallow test, if you will.

There’s nothing wrong with trading — it’s an important market function — but it doesn’t have to be the final destination. The founders who focus on the “product” part of product-market fit may end up the bigger winners.

– Arianna Simpson, a16z crypto general partner

On stablecoins, RWA tokenization, payments, & finance

2. Thinking about tokenization of real-world assets, and stablecoins, in a more crypto-native way

We’ve seen strong interest from banks, fintechs, and asset managers to bring U.S. equities, commodities, indices, and other traditional assets onchain. As more traditional assets come onchain, the tokenization is often skeuomorphic — rooted in the current idea of real-world assets, and not taking advantage of crypto-native features.

But synthetic representations like perpetual futures (perps) allow deeper liquidity and are often simpler to implement. Perps also provide easy-to-understand leverage, so they may be the crypto-native derivative with the strongest product-market fit. Emerging market equities may also be one of the most interesting asset classes to perpify. (The zero-days-to-expiration or 0DTE options market for some equities often trades with deeper liquidity than the spot market, and would be a fascinating experiment for perpification.)

It all comes down to the question of “perpification vs. tokenization”; but either way, we should see more crypto-native RWA tokenization in the coming year.

Along similar lines, in 2026 we’ll see more “origination, not just tokenization” when it comes to stablecoins, which went mainstream in 2025; outstanding stablecoin issuance continues to grow.

But stablecoins without strong credit infrastructure look like narrow banks, which hold specific liquid assets that are considered extra-safe. While narrow banking is a valid product, I don’t believe it will be the backbone of the onchain economy in the long term.

We’ve seen a number of new asset managers, curators, and protocols start to facilitate onchain asset-backed lending against offchain collateral. Often these loans originate offchain and then are tokenized. I think tokenization offers few benefits here, other than perhaps distributed to users that are already onchain. That’s why debt assets should be originated onchain, not originated offchain and tokenized. Origination onchain reduces loan servicing costs, back office structuring costs, and increases accessibility. The challenging part here will be compliance and standardization, but builders are already working on solving those problems.

– Guy Wuollet, a16z crypto general partner

3. Stablecoins unlock the bank ledger upgrade cycle — and new payment scenarios

The average bank is running software that is unrecognizable to modern developers: In the 1960s and 1970s, banks were early adopters of large software systems. The second generation of core banking software started in the 1980s and 1990s (for instance, via Temenos’ GLOBUS and InfoSys’ Finacle). But all this software has been aging, and is being upgraded too slowly. As such, the banking industry — especially critical core ledgers, the key databases that track deposits, collateral, and other obligations — still often run on mainframe computers, programmed with COBOL, and with batch file interfaces instead of APIs.

The large majority of global assets live on those same core ledgers that are also decades old. While these systems are battle tested, trusted by regulators, and deeply integrated into complex banking scenarios, they are also holding back innovation. Adding key functionality like real-time payments can take months or more likely years, and requires navigating layers of technical debt and regulatory complexity.

That’s where stablecoins come in. Not only were the last couple years when stablecoins found product-market fit and hit the mainstream, but this year, TradFi institutions embraced them at a whole new level. Stablecoins, tokenized deposits, tokenized treasuries, and onchain bonds allow banks, fintechs, and financial institutions to build new products and serve new customers. More importantly, they can do this without forcing these organizations to rewrite their legacy systems — systems that, while aging, have run reliably for decades. Stablecoins thus provide a new way for institutions to innovate.

– Sam Broner

On agents & AI

4. We’ll use AI for substantive research tasks

As a mathematical economist, it was difficult to get consumer AI models to even understand my work process back in January this year; yet by November, I could give models abstract instructions in the same way I would to a doctoral student… and they sometimes return novel and correctly executed answers. Beyond my experience here, we’re starting to see AIs used for research more broadly — especially in reasoning domains, where models are now directly aiding discovery and also autonomously solving Putnam problems (perhaps the world’s hardest university-level math exam).

It’s still an open question which fields this type of research assistance will help most, and how. But I’m expecting AI research to enable, and reward, a new type of polymath research style: one that favors an ability to conjecture relationships between ideas, and quickly extrapolate from even more conjectural answers. Those answers may not be accurate, but can still point in the right direction (at least under some topology). Ironically, it’s kind of like harnessing the power of model hallucinations: When the models get “smart” enough, giving them abstract space to bounce around can still produce nonsense — but can sometimes crack open a discovery, just like how people can be most creative when they’re not working in a linear, clearly stated direction.

Reasoning in this manner will require a new style of AI workflow — not just agent-to-agent, but more agent-wrapping-agent — where layers of models help the researcher evaluate the earlier models’ approaches and successively synthesize the wheat from the chaff. I’ve been using this approach to write papers, while others are conducting patent searches, inventing new forms of art, or (unfortunately) finding novel smart contract attacks.

However: Operating ensembles of wrapped reasoning agents for research will require better interoperability between models, along with a way to recognize and properly compensate each model’s contribution — both problems crypto can help solve.

– Scott Kominers, a16z crypto research team and professor, Harvard Business School

5. The invisible tax on the open web

The rise of AI agents is imposing an invisible tax on the open web, fundamentally disrupting its economic foundation. This disruption stems from a growing misalignment between the context and execution layers of the internet: Currently, AI agents extract data from ad-supported sites (the context layer), providing convenience to users while systematically bypassing the revenue streams (like ads and subscriptions) that fund the content.

To prevent the erosion of the open web (and preserve the diverse content that fuels AI itself), we need the mass deployment of technical and economic solutions. This could include models like next-generation sponsored content, micro-attribution systems, or other novel funding models. Existing AI licensing deals are also proving to be a financially unsustainable Band-Aid, often compensating content providers with a fraction of the revenue they’ve already lost to AI-cannibalized traffic.

The web needs a new techno-economic model where value flows automatically. The key transition for the coming year will be moving from static licensing to real-time, usage-based compensation. This means testing and scaling systems — potentially leveraging blockchain-enabled nanopayments and sophisticated attribution standards — to automatically reward every entity that contributes information to an agent’s successful task.

– Liz Harkavy, a16z crypto investment team

On privacy & security

6. Privacy will be the most important moat in crypto

Privacy is the one feature that’s critical for the world’s finance to move onchain. It’s also the one feature that almost every blockchain that exists today lacks. For most chains, privacy has been little more than an afterthought.

But now, privacy by itself is sufficiently compelling to differentiate a chain from all the rest. Privacy also does something more important: It creates chain lock-in; a privacy network effect, if you will. Especially in a world where competing on performance is no longer enough.

Thanks to bridging protocols, it’s trivial to move from one chain to another as long as everything is public. But, as soon as you make things private, that is no longer true: Bridging tokens is easy, bridging secrets is hard. There is always a risk when moving in or out of a private zone that people who are watching the chain, mempool, or network traffic could figure out who you are. Crossing the boundary between a private chain and a public one — or even between two private chains — leaks all kinds of metadata like transaction timing and size correlations that makes it easier to track someone.

Compared to the many undifferentiated new chains where fees will likely be driven down to zero by competition (blockspace has become fundamentally the same everywhere), blockchains with privacy can have much stronger network effects. The reality is that if a “general purpose” chain doesn’t already have a thriving ecosystem, a killer application, or an unfair distribution advantage, then there’s very little reason for anyone to use it or build on top of it — let alone be loyal to it.

When users are on public blockchains, it’s easy for them to transact with users on other chains — it doesn’t matter which chain they join. When users are on private blockchains, on the other hand, the chain they choose matters much more because, once they join one, they’re less likely to move and risk being exposed. This creates a winner-take-most dynamic. And because privacy is essential for most real-world use cases, a handful of privacy chains could own most of crypto.

– Ali Yahya, a16z crypto general partner

On other industries and applications

7. Prediction markets go bigger, broader, and smarter

Prediction markets have already gone mainstream, and this coming year, they’ll only become bigger, broader, and smarter as they intersect with crypto and AI — while also posing new and important challenges for builders to resolve.

First, many more contracts will be listed. This means we’ll be able to access real-time odds not just for major elections or geopolitical events, but for all kinds of in-the-weeds outcomes and complex, intersecting events. As these new contracts surface more information and become part of the news ecosystem (already happening), they’ll raise important societal questions about how we balance the value of this information and how to better design them so they are more transparent, auditable, and more — which is possible with crypto.

To handle the much larger volume of contracts, we’ll need new ways of aligning on truth to resolve the contracts. Centralized platform resolution (did a given event actually happen? how do we confirm it?) is important, but disputed cases like the Zelensky suit market and the Venezuelan election market show the limits. To address these edge cases and help prediction markets scale to more useful applications, new kinds of decentralized governance and LLM oracles can help determine truth for contested outcomes.

AI opens up further possibilities beyond LLMs for oracles. For instance, AI agents trading on these platforms can scour the world for signals that help provide short-term trading edge, helping surface new ways of thinking about the world and predicting what will happen next. (Projects like Prophet Arena already hint at the excitement in this space.) Besides serving as sophisticated political analysts that we can query for insight, these agents could also reveal new things about root predictors of complex societal events when we examine their emergent strategies.

Do prediction markets replace polling? No; they make polling better (and polling information can be fed into prediction markets). As a political scientist, I’m most excited by how prediction markets can function in concert with a rich and vibrant polling ecosystem — but we’ll need to lean on new technologies like AI, which can improve the survey-taking experience; and crypto, which can provide new ways to prove that poll and survey respondents are not bots but humans, among other things.

– Andy Hall, a16z crypto research advisor and professor of political economy, Stanford University

8. Crypto offers a new primitive for use beyond blockchains

For years, SNARKs — cryptographic proofs that let you verify computation without re-executing it — have been largely a blockchain-only technology. The overhead was simply too high: Proving a computation could take 1,000,000x more work than just running it. Worth it when you’re amortizing across many thousands of validators, but impractical anywhere else.

That’s about to change. In 2026, zkVM provers will hit roughly 10,000x overhead with memory footprints in the hundreds of megabytes — fast enough to run on phones, cheap enough to run everywhere. Here’s one reason 10,000x could be a magic number: High-end GPUs have ~10,000x more parallel throughput than a laptop CPU. By the end of 2026, a single GPU will be able to generate proofs of CPU execution in real time.

This could unlock a vision from old research papers: verifiable cloud computing. If you’re running CPU workloads in the cloud anyway — because your computation isn’t heavy enough to GPU-ize, or you lack the expertise, or legacy reasons — you’ll be able to get cryptographic proofs of correctness for a reasonable price overhead. The prover is already GPU-optimized; your code doesn’t need to be.

– Justin Thaler, a16z crypto research team, and associate professor of computer science, Georgetown University

Disclaimer:

1. This article is reprinted from [a16zcrypto]. All copyrights belong to the original author [a16zcrypto]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. Translations of the article into other languages are done by the Gate Learn team. Unless mentioned, copying, distributing, or plagiarizing the translated articles is prohibited.