Blockchain vs database: there is no winner or loser — a 2026 re-reading

We published the original blockchain vs database comparison in 2022, when distributed-ledger technology was still expected to remake finance and the Internet of Things. We have rewritten it with four more years of evidence — keeping the original’s central claim, which held up, and retiring the predictions that didn’t.

The blockchain vs database question was never really “which is better?” Our 2022 answer — there is no winner or loser, the application determines the design — was right in form. What we, along with most of the industry, got wrong was the forecast: we expected financial incumbents to be forced onto distributed-ledger rails. In 2026 the honest scoreboard reads differently, and it is worth being precise about why, because the reasoning is a template for evaluating every over-hyped technology since.

What a blockchain technically is — the part that survives

Strip the ideology and a blockchain is a specific data structure with a specific guarantee: an append-only log whose blocks are hash-linked (Merkle-tree style), replicated across parties who don’t trust each other, with a consensus mechanism deciding what gets appended. That buys you two properties an ordinary database doesn’t natively give across organizations: tamper-evidence (rewriting history breaks the hashes) and shared write authority without a single owner.

A conventional database buys you nearly everything else: orders-of-magnitude better throughput and latency, mature query power, fine-grained access control, easy schema evolution, cheap operations, and — crucially — the ability to fix mistakes. Reversibility is a feature. Most of business runtime is error correction.

What the last decade actually taught

The trust problem was almost always smaller than advertised. Enterprise blockchain consortia assumed organizations couldn’t share a database without a neutral ledger. In practice they could — via APIs, contracts and auditors — and the flagship projects wound down with names and dates attached. we.trade, the trade-finance consortium backed by Deutsche Bank, HSBC, Santander, Société Générale and UBS, entered insolvency in June 2022 after failing to raise further funding, and proposed PwC as liquidator. TradeLens, the Maersk–IBM shipping platform, was discontinued in November 2022 and offline by Q1 2023, with Maersk stating plainly that it “has not reached the level of commercial viability necessary” — and competing carriers proving reluctant to join a network controlled by a dominant rival, which is a governance problem no consensus algorithm addresses. The hard part of inter-company data sharing was governance and data quality; the ledger solved neither and charged for both.

The “top dogs lose to the blockchain community” prediction failed. Banks were not disrupted by DLT; they absorbed the useful pieces. What actually modernized payments were conventional systems: UPI in India and SEPA Instant in Europe are centrally operated, database-backed rails delivering the real-time settlement blockchain promised — at national scale (our engineering guide to both).

Where the properties genuinely matter, the technology quietly works. The defensible niches: public cryptoasset networks themselves; regulated experiments in tokenized securities and central-bank digital currency pilots where multi-party finality is the point; cross-border settlement corridors with genuinely adversarial participants; and tamper-evident audit logging — though for that last one, signed hash chains and transparency logs (certificate-transparency style) deliver the guarantee without distributed consensus. Related: the identity world kept blockchain’s credential cryptography and dropped the mandatory ledger, as we cover in our Sovrin retrospective.

The decision test we actually use

When a client asks “should this be on a blockchain?”, we walk through four questions:

1. Are there multiple writers who genuinely cannot trust a common operator? Not “prefer not to” — cannot, structurally, including any regulator or industry utility that could play host.
2. Is append-only tamper-evidence a requirement, or would authorization plus audit logging do?
3. Can the business tolerate irreversibility — no correcting mis-entries, no GDPR erasure complications, no rollback?
4. Is the data volume and latency profile modest enough for consensus overhead?

Four yeses: a ledger design is worth engineering seriously. Any no — and question 1 alone has been a no in every system a client has brought us so far — a well-designed database with signed audit trails is the faster, cheaper, more maintainable answer. That isn’t cynicism; it’s respect for how good boring databases have become.

The transferable lesson

The 2022 post’s deepest sentence was accidentally its best: applications determine their design. The industry’s error was running that backwards — starting from a fascinating design and hunting for applications. The same test now applies to newer waves (we apply it weekly to AI architecture choices): name the property only this technology provides, prove your problem needs that property, and only then pay the complexity cost. Technology selection is our daily trade in product & software engineering — and “no winner or loser, only fit” remains the right frame. What changed since 2022 is that we now know where the fit mostly isn’t.

FAQ

Should I use a blockchain or a database for my application? A database, unless you can answer yes to all of: multiple writers who structurally cannot trust any common operator, a hard requirement for tamper-evidence, tolerance for irreversibility, and modest throughput needs. That combination is rare — most systems needing “trust” need signed audit logs, not consensus.

What happened to enterprise blockchain? The flagship consortia shut down, specifically and datably: we.trade entered insolvency in June 2022, TradeLens was discontinued in November 2022 and offline by Q1 2023. In both cases the ledger failed to remove the real work — governance, reconciliation, data quality — while adding operational cost and, in TradeLens’s case, a governance problem (carriers would not join a rival’s network) that decentralised consensus does not touch. The useful cryptography — hash-linked logs, verifiable credentials — was absorbed into conventional architectures and is doing fine.

When does a distributed ledger actually make sense? When multi-party finality without a trusted operator is the product itself: public cryptoasset networks, tokenized-asset and CBDC-style settlement experiments, genuinely adversarial cross-organization corridors. Outside those, tamper-evidence alone is served more cheaply by signed hash chains and transparency logs.

Rishi Jain leads business development at Vaagmodo and spends much of his time helping clients match technology decisions to actual business problems. Talk to us about yours: contact or info@vaagmodo.com.