What Is a Directed Acyclic Graph (DAG)?

A Directed Acyclic Graph (DAG) is a data structure from graph theory in which nodes are connected by directional edges that never loop back to a previous node. In cryptocurrency, DAG replaces the linear chain of blocks with an interconnected web of transactions, where each new transaction validates two or more prior ones.

DAG is not a type of blockchain. Both blockchain and DAG are types of distributed ledger technology (DLT), but they use different data structures to record and validate transactions. Both share the goal of decentralization: removing the need for a central authority to verify and record transfers of value.

If blockchain is a single railway track where trains must follow each other in sequence, DAG technology is more like a city road network where multiple vehicles travel different routes simultaneously, and the overall system moves faster as more participants join. If you encountered this term in a cryptocurrency whitepaper, this guide explains what it means, how it differs from blockchain architecturally, which projects use it, and what its real advantages and limitations are.

Clarification: DAG and blockchain are co-equal types of distributed ledger technology. DAG is not a subcategory or evolution of blockchain. They solve the same problem through structurally different designs.

DAG vs. Blockchain: Understanding the Difference

Blockchain and DAG solve the same fundamental problem in distributed ledger technology, but they use structurally different approaches. Blockchain records transactions in a linear chain of sequentially validated blocks, while DAG records them as an interconnected web of transaction nodes where each new transaction validates prior ones directly.

A blockchain groups individual transactions into blocks, then links those blocks chronologically. Validating a new block requires miners running Proof of Work (PoW), which means specialized computers expending significant computational energy to solve mathematical puzzles, or validators staking capital under Proof of Stake (PoS). Under heavy load, this sequential architecture creates a bottleneck: transactions queue for block inclusion, fees rise as users bid for priority, and throughput hits a ceiling determined by block size and timing. Ethereum's gas fees are the most widely encountered example of this fee market under congestion.

The blockchain trilemma, a widely-used framework popularized by Vitalik Buterin, describes the engineering challenge of achieving decentralization and security alongside scalability in a distributed network. Traditional blockchains like Bitcoin and pre-Merge Ethereum prioritize decentralization and security, accepting limited scalability as the tradeoff. DAG's architecture is designed to address that scalability constraint directly, though it introduces its own tradeoffs covered later in this article.

Bitcoin predates DAG-based cryptocurrency networks entirely. Bitcoin's 2008 design used a linear blockchain to solve the double-spending problem, and switching it to DAG would require a fundamental protocol redesign from scratch, not an incremental upgrade. Bitcoin's conservative design philosophy also prioritizes security and decentralization over throughput experimentation, which means DAG-based architectures were purpose-built as separate projects rather than evolved from Bitcoin.

Neither architecture is universally better. Blockchain prioritizes battle-tested security and a mature smart contract ecosystem; DAG theoretically prioritizes throughput and feeless transactions for high-frequency, low-value use cases. The right architecture depends on the application's specific requirements.

The Building Blocks: A Brief Graph Theory Primer

A graph, in mathematics and computer science, is a set of nodes (also called vertices) connected by edges. Before understanding what makes a DAG specific, the three components of its name need to stand on their own. Formally, a DAG is defined as G = (V, E) where V is a set of vertices and E is a set of directed edges such that no sequence of edges forms a cycle. In plain terms: a collection of data points, connected by one-way links, with no path that loops back to its starting point.

The word "directed" means each edge has a specific orientation. If an edge runs from node A to node B, that does not mean an edge runs from B back to A. The word "acyclic" means you can follow the directional edges forward as far as you like and never return to a node you already visited. A family tree illustrates both properties naturally: parents point to children (directed), and you cannot be your own ancestor (acyclic). Follow any path forward through the generations and you will never loop back to where you started.

The difference between a plain graph and a DAG comes down to two constraints. Any set of nodes connected by edges is a graph. A DAG is a specific type of graph where edges have directionality and where no sequence of directed edges forms a cycle. In a cryptocurrency DAG network, cycles would create a logical paradox: Transaction A validates Transaction B, which validates Transaction A, making it impossible to establish which came first or whether either is valid. The acyclic constraint preserves causal ordering and prevents that circular dependency.

One disambiguation is needed before continuing. The word "node" carries two meanings in cryptocurrency. In this article, node refers to a transaction in the graph structure. A network node, by contrast, is a computer participating in the protocol. When both senses appear in the same context, the distinction will be made explicit.

How Does a DAG Work?

A DAG-based cryptocurrency network processes transactions through a self-reinforcing web: every new transaction must validate prior ones before the network accepts it, which means each participant contributes directly to confirmation rather than waiting for a separate class of miners or validators to do it. Think of it like a new employee who must review and approve two pieces of prior work before their own submission is accepted into the system. The more employees join and review prior work, the faster the backlog clears.

Here is how a transaction moves through a DAG network:

1. A new transaction is created as a node. The sender generates a transaction, which enters the DAG as a new node in the graph structure, waiting to be connected to prior nodes.
2. The new transaction selects and validates two or more prior transactions. Using a selection algorithm, the new transaction references prior unconfirmed transactions and confirms their legitimacy before the network accepts the new one. In IOTA's Tangle, this process is called tip selection, and it uses a weighted random walk to choose which prior transactions to reference.
3. The selected prior transactions gain confirmation weight. Each prior transaction now has one more subsequent transaction referencing it. The more subsequent transactions reference a prior one, the higher its confirmation weight and the more confirmed it becomes.
4. The new transaction itself awaits confirmation. It sits as an unconfirmed tip in the DAG until subsequent transactions validate it in turn. Finality is probabilistic and deepens over time rather than occurring at a single instant.
5. As network activity grows, throughput increases. The more transactions enter the network, the more validators exist to confirm prior transactions. Congestion in a DAG works opposite to blockchain congestion: higher volume speeds confirmation rather than slowing it.

[DIAGRAM: DAG node-edge structure showing 7 transaction nodes (circles) connected by directional arrows, where each new transaction points back to two prior nodes. Structure extends outward in a web pattern from left to right, never looping. Contrast panel showing linear blockchain with sequential blocks in a chain. Alt text: "Diagram showing a Directed Acyclic Graph (DAG) cryptocurrency network with transaction nodes connected by directional edges with no cycles, contrasted against a linear blockchain chain structure."]

New transactions choose which prior transactions to validate using a selection algorithm. In IOTA's Tangle, the tip selection algorithm ensures that older unconfirmed transactions do not remain permanently unvalidated, distributing confirmation work across the full graph rather than concentrating it at the tip.

A transaction does not achieve instant finality in a DAG. It becomes progressively more confirmed as more subsequent transactions reference it, and high-confirmation transactions are treated as immutable in practice. This probabilistic finality differs from blockchain's block-based finality but achieves the same functional outcome under sufficient network activity.

So if there are no blocks and no miners, how does a DAG network reach agreement on which transactions are valid?

How DAG Achieves Consensus

Consensus in a distributed network means all participants agree on which transactions are valid and in what order, without relying on a central authority to arbitrate. DAG networks do have consensus mechanisms; they are structurally different from block-based consensus, not absent.

DAG consensus works through cumulative validation. Each new transaction validates prior ones, so the confirmed state of the ledger emerges from the accumulated weight of mutual references rather than from a single authoritative chain. Three major protocols implement this principle in distinct ways:

• IOTA Tangle: Tip selection combined with confirmation weight. New transactions use a weighted random walk to select prior unconfirmed transactions to validate, and a transaction's confirmation deepens as more subsequent transactions reference it.
• Hedera Hashgraph: Gossip-about-gossip protocol combined with virtual voting. Nodes share not just transactions but also records of what information they have shared, building a DAG of communication events. Virtual voting on this event DAG determines the consensus order without requiring explicit voting messages.
• Fantom Lachesis: Asynchronous Byzantine fault tolerance (aBFT) via DAG ordering. Lachesis builds a DAG of event blocks, orders them asynchronously, and finalizes transactions on the Opera mainnet without waiting for a fixed block time.

DAG Cryptocurrencies: Real-World Implementations

Several cryptocurrency projects use DAG as their core data structure or consensus layer. The five most prominent implementations demonstrate how different design teams have applied the same foundational concept in architecturally distinct ways.

IOTA and the Tangle

IOTA is a DAG-based distributed ledger designed for Internet of Things (IoT) use cases and machine-to-machine payments. Its DAG implementation, called the Tangle, requires each transaction to validate exactly two prior transactions before the network accepts it. This validation requirement, combined with the absence of miners to compensate, produces IOTA's feeless transaction model. New transactions choose which prior transactions to validate using the tip selection algorithm, a weighted random walk across the unconfirmed tips of the Tangle. IOTA originally used a centralized Coordinator node to prevent attacks during the network's early low-activity phase, a known security compromise that the IOTA Foundation has been working to remove as part of its IOTA 2.0 decentralization roadmap. The foundational architecture is described in the IOTA Tangle Whitepaper by Serguei Popov.

Hedera Hashgraph

Hedera Hashgraph takes a different architectural path from IOTA. Rather than building a DAG of transactions directly, Hedera uses hashgraph consensus to build a DAG of communication events. Nodes share transaction histories and records of what they have shared (the gossip-about-gossip protocol), and virtual voting on this event DAG determines transaction order without explicit vote messages. Hedera claims throughput exceeding 10,000 transactions per second (TPS) in testing conditions, with low fees and finality in seconds. The network operates under a governing council of enterprise members and is not open-source, representing a decentralization tradeoff made in exchange for governance stability. See the Hedera Hashgraph consensus documentation for the full protocol description.

Nano

Nano's approach is structurally distinct from both IOTA and Hedera. It uses a block-lattice structure, a DAG variant in which each account maintains its own chain of blocks and the accounts are linked together in a DAG structure. This design produces feeless, near-instant transactions without miners or validators. Nano was formerly known as RaiBlocks. The architecture is described in the Nano Whitepaper.

Fantom and Lachesis

Fantom uses its Lachesis consensus mechanism, a DAG-based protocol operating at the consensus layer rather than the transaction recording layer as in IOTA's Tangle. Lachesis orders events asynchronously via a DAG structure before finalizing them on the Opera mainnet, achieving near-instant finality and enabling EVM-compatible (Ethereum Virtual Machine-compatible) smart contracts. The architecture is described in the Fantom Lachesis consensus documentation.

Avalanche

Avalanche uses a DAG-based consensus protocol specifically on its X-Chain (exchange chain), achieving high throughput and sub-second finality for asset transfers. Avalanche uses DAG on its X-Chain only, not across all its chains. The C-Chain is EVM-compatible and supports smart contracts using a linear chain structure. See the Avalanche X-Chain DAG documentation for protocol details.

Smart Contract Support Across DAG Platforms

Some DAG-based platforms support smart contracts, but support varies significantly across implementations and generally lags behind Ethereum's ecosystem in maturity. Smart contracts are self-executing programs deployed on a distributed network that run automatically when predetermined conditions are met.

Fantom's Opera mainnet is EVM-compatible, meaning it runs the same smart contract code as Ethereum. Avalanche's C-Chain is also EVM-compatible and supports a full smart contract environment, though only the X-Chain uses DAG-based consensus. IOTA supports smart contracts via a separate IOTA Smart Contracts Protocol layer rather than natively within the base Tangle. Hedera supports smart contracts through its Hedera Smart Contract Service, which is EVM-compatible. Across all these implementations, the smart contract ecosystems are less mature than Ethereum's, with smaller developer communities, fewer audited protocols, and less decentralized finance (DeFi) infrastructure in production.

Advantages of DAG

DAG's architectural design produces several operational advantages over traditional blockchain, though these advantages come with real tradeoffs covered in the next section.

• Theoretical scalability. DAG throughput increases as network activity grows: more transactions mean more validators, which means faster confirmation. This is the inverse of blockchain congestion, where high demand slows the network and raises fees. Addressing the scalability vertex of the blockchain trilemma is the core architectural motivation for DAG's design.

• Feeless or near-feeless transactions. DAG networks have no transaction fee market because there are no miners or block validators to compensate. IOTA's Tangle is the primary production example of this model. Note that computational resources are still required by the submitting device; "feeless" refers to the absence of a network fee, not an absence of any cost.

• Parallel transaction processing. Multiple transactions are validated simultaneously across the DAG web, rather than waiting for sequential block inclusion. This parallel processing contributes to faster effective throughput under high-activity conditions compared to single-chain sequential validation.

• Energy efficiency. DAG networks require no Proof of Work mining, eliminating the energy-intensive computational puzzle-solving that Bitcoin-style blockchains depend on. The validation work is distributed across transaction submitters rather than concentrated in specialized mining hardware.

• Microtransaction viability. The feeless model makes high-frequency, low-value transactions economically viable. This is the architectural property that makes DAG a candidate for Internet of Things (IoT) and machine-to-machine (M2M) payment applications where per-transaction fees would exceed the transaction value itself.

Limitations and Risks of DAG

DAG's advantages are architecturally real, and so are its limitations. Understanding both is necessary to evaluate whether a DAG-based project's technical claims are credible.

Double spending is the attempt to spend the same cryptocurrency unit twice before the first transaction is confirmed. Blockchain prevents this by making the sequential chain mathematically expensive to rewrite; an attacker would need to control more than 50% of the network's mining power to execute a rewrite. DAG prevents double spending through the web of mutual validations: a transaction that conflicts with an already-validated prior transaction will not be referenced by subsequent transactions, effectively rejecting the invalid branch from the confirmed graph. This protection mechanism has a structural weakness, however, which connects directly to the first limitation below.

• Security vulnerability under low transaction volume. When a DAG network has few active users, confirmation weight across the graph is sparse. An attacker can more easily submit conflicting transactions that overwhelm the legitimate confirmation weight and get accepted into the graph. This is the structural inverse of DAG's scalability advantage: the same property that makes DAG faster under high activity makes it weaker under low activity.

• Smart contract immaturity. Most pure DAG implementations have less developed smart contract ecosystems than Ethereum. This limits decentralized application (dApp) development potential and DeFi infrastructure on native DAG platforms. The per-platform details are covered in the Smart Contract Support section above.

• Centralization compromises in early-stage networks. IOTA originally used a centralized Coordinator node to prevent attacks before its network activity was sufficient to provide adequate confirmation weight independently. Hedera Hashgraph operates under a governing council of enterprise members. Early-stage DAG networks have often required temporary or structural centralization measures to ensure security before full decentralization was achievable.

• Unproven at scale. DAG's throughput advantages are largely theoretical or demonstrated only in controlled testing conditions. Long-term production performance data is limited compared to Bitcoin's 15-plus year track record and Ethereum's multi-year stress testing under real DeFi load.

• Limited developer tooling and ecosystem maturity. Most DAG platforms lack the breadth of developer tools, audited libraries, and production-tested infrastructure that Ethereum's ecosystem has accumulated over years. Developers building on DAG platforms face fewer reference implementations, smaller talent pools, and less community-tested code to draw on.

DAG Security: The 34% Attack and Low-Activity Risks

DAG networks face a specific security risk that grows more serious when the network has few active users, and understanding this risk explains why some DAG implementations made centralization compromises that otherwise seem to contradict their decentralization goals.

The 34% attack is DAG's functional equivalent of blockchain's 51% attack, though the specific threshold varies by protocol and should not be treated as a universal constant. In a DAG network, an adversary controlling a sufficient share of network transaction activity can attempt to produce conflicting transactions that overwhelm the legitimate confirmation weight of the graph. Because confirmation in DAG depends on cumulative transaction references rather than computational work or staked capital, a well-resourced attacker on a low-activity network has a proportionally larger influence over which transaction branches accumulate weight. The attack threshold varies by protocol design and network activity level.

IOTA's original Coordinator node was a direct response to this vulnerability. The Coordinator acted as a centralized checkpoint that prevented invalid branches from being confirmed during the period when the Tangle had insufficient organic transaction volume to secure itself. IOTA has disclosed this as a known centralization compromise and has been working toward removing it in IOTA 2.0. Hedera's governing council represents a different form of the same tradeoff: governance stability and security assurance exchanged for reduced decentralization. DAG security models as a category are still maturing relative to Bitcoin's battle-tested track record spanning over 15 years of adversarial production conditions.

Real-World Use Cases for DAG Technology

DAG's feeless architecture and theoretically unlimited throughput make it structurally suited to applications where blockchain transaction fees would make the economics unworkable. Two categories stand out: IoT and machine-to-machine payment networks, and high-throughput settlement systems where fee-free micro-value transfers are required at scale.

DAG and the Internet of Things (IoT)

The Internet of Things (IoT) refers to the billions of networked physical devices that communicate and transact data and value autonomously, from smart home appliances to industrial sensors to connected vehicles. These devices increasingly need to make payments to each other or to service providers, at amounts so small and so frequent that any per-transaction fee renders the economics nonviable.

Imagine your autonomous vehicle approaching a highway toll point and settling the charge automatically, with no human authorization required and no gas fee consuming more value than the toll itself. DAG's feeless model makes this settlement economically viable at billions-of-transactions scale, where blockchain fees would make the entire use case impractical.

A smart electricity meter can settle energy usage payments between a household and an energy provider every 15 minutes, at amounts too small for any blockchain transaction fee to be proportionate. DAG's architecture removes that fee barrier, allowing granular real-time settlement that would otherwise require batching transactions to make the fees worthwhile.

In a manufacturing supply chain, industrial sensors purchase data readings from each other in a machine-to-machine (M2M) data marketplace, where each data packet carries micro-monetary value. A blockchain gas fee would make each packet purchase uneconomical; DAG's feeless architecture removes that constraint and enables a genuine market for machine-generated data at micro-scale.

DAG in Cryptocurrency Whitepapers: A Reader's Guide

A cryptocurrency whitepaper is a technical document published by a project's founding team that describes its architecture and consensus mechanism alongside its tokenomics and value proposition. The tradition was established by Satoshi Nakamoto's 2008 Bitcoin: A Peer-to-Peer Electronic Cash System, the pseudonymous document that introduced the blockchain data structure and the genre simultaneously. The term "whitepaper" is used loosely in cryptocurrency: some documents are mathematically rigorous technical papers; others are marketing materials that adopt the whitepaper label to project credibility. Calibrate your expectations accordingly when reading any DAG project's whitepaper.

DAG architecture typically appears in the "Architecture" or "Technical Design" section of a whitepaper. Projects that use DAG describe their structure using graph theory notation, often include a DAG topology diagram showing nodes and directional edges, and then explain the consensus mechanism layered on top of that structure. When you encounter DAG terminology in a whitepaper, the following glossary covers the terms you will most commonly see.

Common DAG whitepaper terminology:

• DAG topology - the structural arrangement of nodes and directional edges in the specific DAG implementation
• Acyclic transaction graph - an alternative term for DAG used in some whitepapers
• Tangle - IOTA's specific DAG implementation, where each transaction validates two prior transactions
• DAG-based consensus - consensus achieved through the DAG structure rather than through sequential blocks
• Directed graph - a graph where edges have directional orientation; a DAG is a directed graph with the additional acyclic constraint
• Vertex - the graph theory term for node; used interchangeably in formal whitepaper descriptions
• Tip selection - the algorithm for choosing which prior unconfirmed transactions a new transaction will validate

Three DAG whitepapers are worth knowing as reference points. The IOTA Tangle Whitepaper by Serguei Popov, formally titled "The Tangle," describes the mathematical properties of IOTA's DAG structure including the tip selection algorithm and probabilistic confirmation model. The Hedera Hashgraph consensus documentation covers the gossip-about-gossip protocol and virtual voting mechanism in technical detail. The Fantom Foundation's documentation describes the Lachesis aBFT consensus architecture and its relationship to DAG event ordering.

When you open a DAG project's whitepaper, check these five elements in the Architecture section:

1. DAG topology type. Does the whitepaper specify whether DAG is used at the transaction layer (like IOTA's Tangle, where each transaction is a node) or the consensus layer (like Fantom's Lachesis, where events are ordered via DAG before finalization)? These are architecturally different implementations with different security and scalability properties.
2. Consensus mechanism. How does the network reach agreement on the valid state of the ledger? Look for named mechanisms: tip selection, gossip-about-gossip, aBFT, virtual voting. A whitepaper that describes consensus only as "transactions validate transactions" without naming the mechanism is leaving out material technical detail.
3. Security model. How does the network prevent attacks when transaction volume is low? What is the attack threshold, and does the whitepaper acknowledge that the threshold varies by network activity level? Whitepapers that omit low-activity security discussion are leaving out a known structural vulnerability.
4. Decentralization design. Are there coordinator nodes, governing councils, or other centralization measures? If so, what is the stated path toward removing them? Both IOTA and Hedera have had to make centralization compromises; a credible whitepaper acknowledges this rather than claiming full decentralization.
5. Performance claims with methodology. Do TPS figures include testing conditions and network size? Unqualified TPS claims without context, such as "unlimited throughput" without a model or "10,000 TPS" without a test description, should be treated skeptically.

These five elements give you a framework to evaluate whether a DAG whitepaper's claims are credible and whether the architecture is a novel contribution or a standard implementation of known patterns.

Frequently Asked Questions About DAG

Is DAG a type of blockchain?

No. DAG and blockchain are both types of distributed ledger technology (DLT), but they use fundamentally different data structures. Blockchain records transactions in sequential blocks linked in a chain; DAG records them as an interconnected web of transaction nodes where each new transaction validates prior ones. Neither is a subcategory of the other.

How does DAG achieve consensus?

DAG networks achieve consensus through cumulative transaction validation rather than sequential block creation. Each new transaction validates prior ones, building a web of mutual confirmations; the more subsequent transactions reference a prior one, the more confirmed it becomes. The specific mechanism varies by protocol: IOTA's Tangle uses tip selection and confirmation weight, Hedera Hashgraph uses gossip-about-gossip combined with virtual voting, and Fantom's Lachesis uses asynchronous Byzantine fault tolerance (aBFT) via DAG event ordering.

Is DAG better than blockchain?

Neither is universally better; they represent different architectural tradeoffs suited to different use cases. DAG theoretically offers higher throughput, feeless transactions, and better scalability for high-frequency low-value applications such as IoT payments; blockchain offers more battle-tested security, a more mature smart contract ecosystem, and a longer track record in production environments. The right architecture depends on the specific requirements of the application.

What are the main disadvantages of DAG?

DAG's main limitations include security vulnerability under low transaction volume, where sparse confirmation weight makes the network more susceptible to coordinated attacks. DAG also has less mature smart contract support than Ethereum across most implementations, and several prominent DAG networks have made centralization compromises for security reasons, including IOTA's historical Coordinator node and Hedera's governing council model. DAG's throughput advantages remain largely theoretical, with limited empirical production data compared to decade-plus blockchain track records.

What cryptocurrencies use DAG?

The five most prominent DAG-based cryptocurrency implementations are IOTA (the Tangle, IoT-focused and feeless), Hedera Hashgraph (enterprise-grade hashgraph consensus), Nano (block-lattice structure, feeless instant payments, formerly RaiBlocks), Fantom (Lachesis DAG consensus layer, EVM-compatible smart contracts), and Avalanche (DAG-based X-Chain for high-throughput asset exchange).

What is a DAG in a cryptocurrency whitepaper?

When a cryptocurrency whitepaper describes its architecture as DAG-based, it means the project records or orders transactions using a web-like graph structure rather than a linear chain of blocks. DAG architecture is typically described in the "Architecture" or "Technical Design" section of a whitepaper, using terminology such as "DAG topology," "acyclic transaction graph," "Tangle," "tip selection," or "DAG-based consensus."

Key Takeaways

• A Directed Acyclic Graph (DAG) is a data structure from graph theory in which transaction nodes are connected by directional edges that never loop back, creating a web-like ledger rather than a linear chain.
• DAG is not a type of blockchain. Both are types of distributed ledger technology (DLT); they use different data structures to solve the same problem of decentralized transaction recording.
• DAG's primary architectural advantage is theoretically self-scaling throughput: higher transaction volume speeds up confirmation rather than slowing it, with no transaction fee market required to compensate miners.
• DAG's primary limitations are security vulnerability under low network activity, less mature smart contract support than Ethereum, centralization compromises in some early-stage implementations, and limited developer ecosystem maturity compared to established blockchain platforms.
• The five most prominent DAG-based cryptocurrency implementations are IOTA (the Tangle), Hedera Hashgraph, Nano, Fantom (Lachesis), and Avalanche's X-Chain.
• When evaluating a DAG project's whitepaper, check five things: the DAG topology type, the consensus mechanism, the security model at low activity, the decentralization design, and whether TPS performance claims include testing methodology.

The primary whitepapers for IOTA, Hedera Hashgraph, and Fantom are linked throughout this article for deeper technical study.

This article is for educational and informational purposes only. Nothing in this article constitutes investment advice, financial advice, or a recommendation to buy, sell, or hold any cryptocurrency. Cryptocurrency investments carry significant risk, including the risk of total loss. Conduct your own research and consult a qualified financial advisor before making investment decisions.