Idan Ofrat (Fireblocks): “Institutions want TradFi on the front end, blockchain behind the scenes”

Custody was Fireblocks' original wedge. When did you realize it wasn't defensible on its own?

From day one, actually. Custody was always a means to an end. Our mission is to enable every enterprise to support digital assets, and when we started, custody was simply the biggest pain point — trading desks and exchanges were operating without the basic infrastructure to hold and transfer money securely. Once we had that layer, we started addressing everything else. I see three waves in digital assets: crypto trading, stablecoin payments, and now tokenized securities. Agentic payments, in my view, are just an extension of the stablecoin payments wave. Each wave has its own set of problems, and we've extended the platform to solve them as they emerge.

"Stablecoins are 65% of the volume we process — more than crypto”

Your Earn product tackles yield on stablecoins in DeFi. What was broken?

Two things: security and operational efficiency. And there are two types of customers. Crypto-native firms want to tap into DeFi but do it in a Wild West manner. They access protocols without any guardrails. Non-crypto-native firms find DeFi essentially unusable: the technology is too complex, and there's no equivalent to what they know from traditional treasury management. We wanted a solution that felt like TradFi on the front end (you lend, you borrow) with the blockchain complexity abstracted away, and cross-chain by default. This product matters more today than it did before, because stablecoins now account for around 65% of the volume we process on Fireblocks — more than crypto. Historically, on-chain yield really only existed for native tokens like ETH or SOL through staking. Earn is the first product that lets you generate yield on stablecoins in a secure, institutional way.

The original crypto thesis was disintermediation — direct access to protocols. Yet clients now trust you as an intermediary. Is disintermediation dead?

For institutions, largely yes. Once companies start operating meaningfully with digital assets, they want the same financial building blocks they have in TradFi. When you put money in a bank, the next thing you do is click a button to find a yield opportunity. The thesis behind Earn was: rebuild what exists in traditional finance, keep the front-end familiar, but change everything in the technology behind the scenes. Institutions need to trust both us and the protocols we connect to. They trust us because we have hundreds of engineers working on security, we have the certifications, and we vet DeFi protocols the way we'd onboard any vendor. Not every protocol makes it onto the platform.

Concretely, what do the guardrails look like?

We call it defence in depth. At the base is the key management layer — that's what we're best known for, our MPC infrastructure. On top of that, a governance layer: deterministic, rule-based policies that let an organisation define who can move which assets, in what amounts, under what conditions. Then more heuristic mechanisms — think of the difference between a firewall and an antivirus. Everything runs in parallel to create concentric circles of security. For DeFi specifically, we also curate which protocols can be accessed, by whom, and what actions are permitted. If a protocol gets breached, we can block transactions on the fly. Customers can opt out if they want to be more aggressive, but security is on by default.

"We shut down two years of engineering work”

What's the hardest product decision you've made?

In early 2022, we identified that embedded wallets would be a major trend — with consumer brands and B2B companies wanting to embed our technology into their own applications. MoneyGram or Western Union is giving its users stablecoin wallets. Exchanges offering non-custodial options alongside custody. There were two possible approaches: give customers a headless, bare-bones wallet infrastructure and let them build on top, or provide a full-fledged solution with UI, financial applications, and easy embedding. We bet on the first. Two years in, we realized the market was going the other way. So we shut down the wallet product we'd built and acquired Dynamic instead, which had the more complete solution. Business-wise, it was a fairly easy call. Emotionally, less so — you're killing two years of engineering work.

"Without banks, digital assets don't reach the mass market”

Serving the biggest banks means heavy compliance and long integration cycles. How do you keep that from slowing product velocity elsewhere?

Two points. First, we've been serving banks for around four years now, so the adaptation has been gradual — certifications, security requirements, enterprise agreements. It's a journey, not a switch. Second, this is a conscious bet. Banks are the entities that practically manage finance today. Without them, digital assets don't reach the mass market. We'd rather invest there than in niche players. On the org side, we're 900 people, 300 of them in engineering, and the platform is designed modularly, a shared core with dedicated building blocks per segment: banks, payments, and so on. A solutions team assembles those Lego bricks for specific customer needs. If a bank needs an on-premise deployment for a certain function, we move that brick from cloud to on-prem without touching the rest.

And internally — how are you using AI?

Across the board, from engineering to customer teams. The challenge was doing it without breaking security and privacy, so we built an internal AI governance layer — every time someone accesses data through AI, the request goes through it. Not everyone can access financial data or customer feedback; role-based permissions gate what AI can retrieve. It works well and is now core to how the company operates.

"The real threat isn't a competitor. It's the shift from humans to agents”

What's the biggest product threat to Fireblocks over the next few years?

Honestly, the product that would displace us probably doesn't exist yet. That's how the market works: every so often, a new technology disrupts everything. What we identified recently is that agentic AI will reshape how every SaaS company operates. Instead of human users on your platform, you'll have agents. The companies that don't make themselves usable by AI agents will fall behind. Every developer today is trying to build agent-driven workflows — including on Fireblocks. They don't read the docs themselves; they point an agent at our documentation and have it build the integration. So we've invested heavily in making the platform agentic: an MCP server so agents can connect to data and transact with it, LLM-friendly documentation, and a CLI that lets coding agents like Claude Code work directly with the platform. The threat isn't a new competitor — it's the shift from humans to agents.

IPO — still on the table?

An IPO is a means to an end. It's one way to raise funding and create liquidity for employees and investors, but it's not the only one. Stripe is a successful private company. All options are on the table, and the finance team evaluates market conditions on an ongoing basis. It's one path, not the only one.

"This isn't a zero-sum game”

Banks like JPMorgan and Morgan Stanley are moving deeper into digital assets. Do you see any of them trying to compete with Fireblocks directly?

I don't. No reasonable bank will try to build the same infrastructure we do. Every player has a skill set. Banks know how to build banking services; we know how to build digital asset technology. The few banks that did try to build in-house eventually came back to us — to Fireblocks or, occasionally, to one of our competitors. The same goes for payment companies: they don't want to divert their main line of business from building infrastructure. This is a 100% opportunity for us. And more broadly, I don't see this as a zero-sum game. The whole industry is re-platforming to digital assets. That's good for fintechs, for payments companies, for banks, and for us. The same logic applies to agentic payments — they won't just move existing payments from humans to agents, they'll create categories of payments that don't exist yet.

>> Read our latest Key Takeaways featuring Fireblocks (NYC, June 22nd)