Crypto has a security problem that's starting to look less like trivia and more like a core business risk. Lorenzo Valente, director of digital asset research at ARK Investment Management, says crypto hacks have caused $8.7 billion in losses so far in 2026. That's already higher than the $5.6 billion estimate for full-year onchain revenue.
Zoom out, and the picture doesn't soften much. Since 2017, the industry has reportedly lost about $26.9 billion to hacks while generating $47 billion in onchain revenue. Rough math puts losses at around 57 cents for every dollar earned.
That number raises a bigger question than "how many exploits happened": when applications get hit again and again, can crypto still produce lasting economic value - or is security just eating the upside?
The answer depends on what's included in those totals, how protocols manage risk, and whether users and institutions can tell the difference between systems that have matured and ones that still wobble.
What does 57 cents in hack losses per dollar earned mean?
Valente's comparison is a straightforward ratio between two things: reported hack losses and onchain revenue. The cited totals are about $26.9 billion in cumulative losses since 2017 versus $47 billion in cumulative onchain revenue over the same span. Divide losses by revenue, and you get roughly 57 cents lost per dollar earned.
But it's worth being careful about what that ratio does - and doesn't - mean. It doesn't say every protocol loses 57% of its revenue, and it doesn't mean each dollar of reported revenue is immediately available to cover losses. It's an industry-wide, multi-year comparison across many projects. In other words, it's a signal about the sector's economics, not a neat accounting result for any single chain, company, or investor.
Read the comparison as a risk signal, not a protocol score
The metric is helpful because it forces attention onto a cost that's easy to gloss over. A protocol might look strong on user growth, transaction volume, and fees, while the broader ecosystem still absorbs losses from compromised contracts, infrastructure, or other attack surfaces. If those losses are ignored, "health" can look better than it really is.
Still, the ratio has limits. "Onchain revenue" can vary by dataset, and loss estimates can shift as incidents are investigated or attributed. The source doesn't provide a full breakdown of what counts as revenue, which incidents are included, or whether losses are measured consistently across cases. So the headline comparison should trigger follow-up questions - not replace diligence.
The year-to-date comparison makes the point even sharper: $8.7 billion in reported hack losses already exceeds the $5.6 billion estimate for full-year 2026 onchain revenue. The source also suggests losses were on track to surpass revenue for the first time since 2020. As with any forecast, though, it's really a comparison of the cited estimates - not a guarantee about where the year ends.
The headline figure should prompt better questions
If you're trying to apply this to a protocol, the useful follow-ups aren't "is the industry average good or bad?" They're questions like:
- How much revenue does the protocol actually keep after incentives and operating costs?
- What assets and permissions could an attacker reach through the protocol's contracts and dependencies?
- If something breaks, how fast can the team pause, limit damage, or recover?
- Do users get clear disclosures about incident losses and security assumptions?
Those questions connect the big industry ratio to something you can actually evaluate. High-level stats can't tell you whether a specific app is safe. What they can tell you is that security costs and failure modes have to be part of the economic analysis, not an afterthought.
Why is security is a bottleneck for crypto?
Crypto puts software in front of assets that can often move continuously - and across borders. That changes the operating reality compared with many conventional products. A logic bug or a compromised key can turn into real money problems immediately, and reversals may be impossible or impractical once funds move.
Security, then, isn't only about "code quality." It covers the whole system: contract design, key management, upgrades, third-party dependencies, monitoring, incident response, and even the incentives that affect how quickly teams act on risk. And just because something passes an audit doesn't mean it's safe everywhere that matters - especially if weaknesses exist outside the audited code, or assumptions change after the review.
Attack surfaces grow with system complexity
The simplest contract has fewer moving parts than a protocol that connects to multiple chains, bridges, oracles, custody systems, and third-party applications. Each integration can add value, but it also adds assumptions. If one component fails, the impact can ripple through the system and put more assets at risk.
Bridges are the obvious example. They help move assets or messages across networks, but their security depends on how cross-chain activity is verified and who controls the relevant permissions. JPMorgan analysts have flagged bridge security as a long-standing industry challenge. That doesn't mean every bridge is doomed - it means teams and investors need to understand the specific design and trust model, not treat "cross-chain" as one generic category.
Operational controls matter too. Who can change an implementation? How many approvals are required to move funds? Are emergency powers protected, monitored, and limited? Even strong code can be undone by weak privileged-key controls.
Audits help, but they don't transfer responsibility
An audit is a review of a defined system at a particular moment. It can find real issues and improve confidence, but it can't promise the software will be invulnerable forever. Scope is everything: an audit may cover one contract while leaving out an integration, a later upgrade, or the operational process around key access.
A better way to think about it is security as ongoing operating discipline. That means independent review before launch, controlled changes after launch, monitoring for unusual activity, and a tested plan for responding to incidents. It also means making trade-offs explicit. More checks and slower upgrades can reduce some risks, but they can also add costs and delay features. The goal isn't "zero risk at any price." It's to keep the remaining risk understandable, proportionate, and manageable.
And if you want the real test: what happens under stress? If suspicious activity is detected, can the team respond without creating a second failure? If the plan depends on one person being online - or on a process nobody has rehearsed - the security plan isn't operationally ready.
How do major hacks affect DeFi and institutional adoption?
A major exploit can damage more than the project that got hit. Users can pull funds from related apps, liquidity providers may get more cautious, and counterparties can reassess their exposure to an entire category of products. In DeFi, where protocols build on each other, risk can spread via shared assets and integrations even when the original vulnerability is isolated.
The source highlights several high-value incidents in 2026: a $387 million Bitget hack, a $320 million Liquid Network exploit, and a $295 million Drift Protocol exploit. It also cites a $6.3 billion Binance hack. That last number is unusually large compared to the others, so it's worth checking against the underlying reporting and the exact definition of what counts as "stolen" before using it in a standalone analysis. The broader point doesn't hinge on that one figure - year-to-date losses are already large relative to the projected onchain revenue.
TVL moves can indicate caution, not prove the cause
DefiLlama figures cited in the source suggest total value locked in DeFi fell nearly 10% after the Binance exploit, from $85.9 billion on October 29 to $77.4 billion on the following Thursday. That's a meaningful move - but it doesn't automatically prove the hack caused the decline.
TVL can change for lots of reasons: asset prices, user withdrawals, incentives, and shifts between protocols. A hack may be part of the explanation, but a before-and-after comparison isn't the same thing as a controlled causal test. For teams and analysts, the practical questions are whether withdrawals concentrated in the exposed applications, whether liquidity returned after the incident was addressed, and whether broader activity changed in the same period.
Institutions need evidence about controls and recovery
Institutions rarely decide on participation based on one headline. Risk teams want clarity on custody, governance, liquidity, legal responsibilities, and the technical controls that protect assets. Repeated high-profile losses make these reviews harder. Even a firm interested in DeFi still has to factor in incident frequency, uncertain recovery paths, and dependencies it doesn't control.
That creates a threshold for adoption. Institutions usually need more than "this protocol was audited." They want a map of assets and dependencies, documented permissions, defined incident procedures, and a way to understand what happens if a key component fails. These controls don't guarantee capital will arrive. They just make the risk legible enough to evaluate.
Users face the same issue at a smaller scale. A high yield doesn't automatically make a product safe if the controls are opaque or disclosures are weak. If you can't tell what assumptions the system relies on, you can't do a meaningful comparison against a simpler alternative.
What should protocol teams, investors and users do?
The loss-to-revenue ratio isn't a reason to treat every crypto project as equally unsafe. It's a reason to treat security as part of the product's operating model and economics. The review will look different depending on your role, but it should focus on exposure, controls, and how the team responds when prevention fails.
For protocol teams: build a repeatable security process
Start with a clear inventory: what assets you control, which contracts and services can affect them, and who (or what systems) have privileged access. That inventory has to include external dependencies. If your protocol relies on a bridge, oracle, custodian, or relayer, that dependency belongs in your risk model - even if someone else operates it.
A practical process might include:
- Set boundaries: document what the protocol can access and which actions privileged roles can take.
- Review changes: require appropriate testing and independent review for material upgrades, not just for initial deployment.
- Limit authority: separate duties, protect keys, and use controls proportionate to the value at risk.
- Prepare response: define who can act during an incident, what can be paused, and how users will be informed.
- Learn from failures: track near misses and incidents, then update controls instead of treating each event as an isolated lesson.
How heavy the process should be depends on potential damage. A small experiment with limited funds shouldn't require the same governance overhead as infrastructure securing hundreds of millions. But a protocol shouldn't call itself "production-ready" while depending on undocumented emergency procedures.
For investors and users: assess exposure, not just reputation
An audit badge, a recognizable brand, or a big TVL figure won't answer all the security questions. Before you deposit or allocate capital, look at what the protocol actually does, what contracts hold assets, which external systems it depends on, and whether there are restrictions or emergency controls in place.
Then check the economics. Compare revenue with incentives, costs, and realistic loss scenarios - not just gross activity. Some protocols look busy while they pay heavily for capital, or while they rely on fragile infrastructure. If disclosures don't let you understand the basic risk, that lack of clarity is itself important.
A simple decision rule helps: if you can't explain how assets could be lost, who has authority over them, and what recovery might look like, reduce exposure or wait for better information. This isn't a prediction that a hack will happen. It's a way to avoid taking risk you didn't choose knowingly.
Conclusion
The reported 57-cent figure is a useful warning, not a final verdict on every crypto protocol. It highlights why security has to sit in the same conversation as revenue, growth, and adoption. A product that can generate fees but can't clearly explain how it protects assets - or how it will respond if protections fail - hasn't really accounted for its full operating cost.
For teams, the next step is to map dependencies and privileges, then stress-test the response plan against realistic failures. For investors and users, it's about understanding what can go wrong before deciding how much exposure is acceptable. In crypto, security isn't separate from economics. It's one of the conditions that makes those economics hold up.
