The Graph is a monolith. It bundles subgraph definition, indexing, querying, and curation into a single protocol, creating a powerful but rigid standard. This bundling creates a single point of failure and cost inefficiency for developers.
developer-ecosystem-tools-languages-and-grants
Blog
The Future of Indexing: Will The Graph Be Unbundled?
The one-size-fits-all indexing model is under siege. Specialized services for DeFi, NFTs, and social graphs are carving out market share, forcing a fundamental rethink of data infrastructure. This is the unbundling of The Graph.
introduction
THE UNBUNDLING
Introduction
The Graph's monolithic indexing stack is being disaggregated by specialized, cost-competitive alternatives.
Specialized protocols are unbundling it. Goldsky and SubQuery now offer faster, cheaper indexing by separating execution from settlement. This modular approach mirrors the L2 execution layer evolution, where specialized rollups like Arbitrum and Optimism unbundled execution from Ethereum's base layer.
The cost structure is the primary vector. The Graph's curation and indexing rewards are priced for a generalized, decentralized network. For many applications, a dedicated RPC node or a service like Goldsky provides the same data at a fraction of the cost and latency.
Evidence: SubQuery's integration with over 120 chains and Goldsky's sub-second streaming data pipelines demonstrate that developer demand is shifting towards performance and cost over pure decentralization for indexing.
key-trends
THE INDEXING BATTLEGROUND
Executive Summary
The Graph's subgraph model faces an existential challenge from specialized, application-specific indexing solutions that prioritize performance and sovereignty.
01
The Subgraph Bottleneck: Monolithic & Costly
The Graph's generalized model introduces latency and cost overhead for high-frequency dApps. Its decentralized curation is a governance win but a performance tax.\n- ~2-10s latency for complex queries\n- Indexer staking economics create fixed costs for all data\n- Subgraph syncing fails for real-time state (e.g., NFT bids)
2-10s
Query Latency
Fixed Cost
Pricing Model
02
Unbundling by Application: GoldRush, Reservoir, Moralis
Verticalized APIs are winning market share by owning the full stack for specific use cases like NFTs or DeFi. They bypass The Graph entirely.\n- GoldRush (NFTs): Sub-second NFT API, powers Coinbase NFT\n- Reservoir: ~500ms order book aggregation for NFT markets\n- Moralis: Streams for real-time wallet tracking
<1s
API Latency
Vertical
Specialization
03
The Sovereign Stack: Rollup-Centric Indexing
Appchains and rollups (Fuel, Eclipse) are building native indexing into their execution clients. Data availability layers like Celestia and EigenDA enable cheap, verifiable state histories.\n- Fuel's Sway language bakes indexing into the VM\n- EigenLayer AVSs could offer restaking for indexer security\n- Move-based chains (Aptos, Sui) have native indexing primitives
Native
Execution Layer
AVS
Security Model
04
The Graph's Defense: Substreams & Firehose
The Graph's response is Substreams, a high-performance data streaming framework. It enables massively parallel processing and direct integration into node clients.\n- Firehose provides raw blockchain data at ~100MB/s\n- Enables custom sinks (e.g., to PostgreSQL, Kafka)\n- Pinax (by StreamingFast) already serves $10B+ TVL protocols
100MB/s
Data Throughput
Parallel
Processing
05
The Verdict: Not a Monolith, But a Protocol
The Graph won't be "unbundled" into oblivion; it will be forced up the stack. Its role shifts from being the indexer to being the verifiable data availability layer for indexing.\n- Future: Substreams as the settlement layer for derived data\n- Indexers become specialized processors for specific sinks\n- Curators signal on data streams, not subgraphs
Settlement
New Role
Derived Data
Product
06
VC Takeaway: Bet on Specialization & Integration
Investment thesis shifts from monolithic indexing protocols to tightly integrated data pipelines. The winners will own a critical vertical or be baked into a dominant execution layer.\n- Bullish on: Application-specific APIs, rollup-native tooling, EigenLayer AVS for indexing\n- Bearish on: General-purpose query layers competing on speed alone
Vertical APIs
Bull Case
General Query
Bear Case
market-context
THE DATA
The Unbundling Thesis: Why Generalization Fails at Scale
The Graph's monolithic, chain-agnostic model is being dismantled by specialized, high-performance alternatives.
Generalized indexing is inefficient. The Graph's one-size-fits-all architecture forces every query through a generic pipeline, creating latency and cost overhead that specialized chains avoid. This is the same architectural flaw that doomed general-purpose L1s against rollups like Arbitrum and Optimism.
App-specific subgraphs are the future. High-frequency protocols like Uniswap and Aave require sub-second data freshness, which The Graph's decentralized network cannot guarantee. They will build proprietary indexing layers, mirroring the app-chain thesis of dYdX and Frax Finance.
New primitives enable unbundling. Tools like Ponder and Goldsky offer developer-friendly, self-hosted indexing with native multi-chain support, eliminating the need for a global marketplace of indexers. This shifts the value from network coordination to execution performance.
Evidence: The Graph processes ~1 billion queries daily, but its decentralized indexers introduce 2-10 second latencies. In contrast, a Ponder indexer on a dedicated RPC node achieves sub-100ms responses, making it viable for on-chain arbitrage and real-time dashboards.
THE GREAT UNBUNDLING
Indexer Archetype Comparison: Generalist vs. Specialist
A feature and economic comparison of indexing service models, analyzing the trade-offs between The Graph's curated marketplace and emerging direct-to-consumer specialists.
| Feature / Metric | Generalist (The Graph) | Specialist (Direct Indexer) | Super-App (GoldRush, Etherscan) |
|---|---|---|---|
Core Business Model | Query marketplace & delegation | Direct API sales & custom contracts | Ad-supported data + premium APIs |
Query Pricing Model | Dynamic, GRT-denominated | Fixed-rate, stablecoin-denominated | Freemium; $50-500/month for APIs |
Protocol Overhead Tax | ~28% (Indexer/Curator/Delegator cuts) | 0% (Direct to consumer) | 0% (Bypasses protocol) |
Time to Custom Subgraph | ~2-4 weeks (curation voting) | < 48 hours (direct deployment) | Not applicable (pre-built only) |
Supports Arbitrary RPC Calls | |||
Native Cross-Chain Indexing | Limited (via partnerships) | ||
Max Query Latency (p99) | < 2 sec | < 500 ms | < 1 sec |
Typical Client | Early-stage dApp, General query needs | High-frequency dApp (e.g., DEX, lender), Enterprise | Retail user, Analyst, Broad explorer use |
protocol-spotlight
THE UNBUNDLING
The New Indexing Stack: Who's Eating The Graph's Lunch?
The Graph's monolithic subgraph model is under siege by specialized protocols targeting its core weaknesses: cost, speed, and developer experience.
01
The Problem: Subgraphs Are Monolithic & Expensive
The Graph's subgraph model bundles indexing logic, data transformation, and query service into one inflexible, pay-per-query package. This creates vendor lock-in and unpredictable costs for high-throughput dApps.
- Cost Spikes: Query fees scale linearly with usage, creating unpredictable OpEx for protocols like Uniswap or Aave.
- Rigid Schema: Data transformations require full subgraph redeploys, slowing iteration.
- Centralized Curation: Indexer selection and tokenomics add friction for new chains.
100k+
Subgraphs
$0.50+
Avg Query Cost (1k)
02
The Solution: Envio's Hyperparallel Indexer
Envio unbundles the stack by separating indexing logic from execution. Developers write handlers in TypeScript, and Envio's engine processes blockchain data in parallel, delivering a performant, customizable GraphQL endpoint.
- Deterministic Pricing: Fixed-cost subscription model, not pay-per-query.
- ~10x Faster Sync: Parallel processing enables indexing speeds of ~1M blocks/hour.
- Multi-Chain Native: Index across Ethereum, Solana, Starknet in a single project.
10x
Faster Sync
-90%
Cost Model
03
The Solution: Goldsky's Real-Time Firehose
Goldsky attacks the latency problem by providing a real-time data firehose. It streams indexed blockchain data directly to applications via WebSockets or server-sent events, bypassing the request-response model.
- Sub-Second Latency: Delivers finality events in ~500ms, critical for on-chain gaming and trading.
- Managed Service: Fully hosted solution, removing devops burden from teams like LayerZero and Axiom.
- SQL & NoSQL Sinks: Pipe data directly to Postgres, Kafka, or Snowflake for analytics.
<1s
Event Latency
0 DevOps
Managed
04
The Solution: Subsquid's Decentralized Data Lake
Subsquid decouples data ingestion from querying. Its decentralized network of Archivists stores raw chain data in a common lake. Indexers (Processors) then transform this data on-demand into APIs.
- Cost-Efficient: Querying pre-ingested data is ~100x cheaper than RPC calls.
- Open Data: Raw data is a public good, eliminating vendor lock-in.
- EVM & WASM: Supports Ethereum, Polkadot, and any Substrate chain with a single SDK.
100x
Cheaper Reads
Decentralized
Architecture
05
The Meta-Solution: RPC Providers as Indexers
Infura, Alchemy, and QuickNode are bundling indexing into their core RPC offerings. They leverage their existing infrastructure and relationships to offer 'good enough' indexed data with minimal setup.
- One-Stop Shop: Developers get RPC, indexing, and analytics from a single dashboard.
- Enterprise SLAs: Guaranteed uptime and support, appealing to TradFi entrants.
- Network Effects: Direct integration with EIP-4337 bundlers and zk-rollup sequencers.
80%+
Dev Market Share
Bundled
Product Suite
06
The Endgame: Application-Specific Indexing
The final unbundling: protocols will run their own purpose-built indexers. Using frameworks like Apollo or Ponder, they index only the data they need, directly from an RPC or data availability layer like EigenDA.
- Total Control: No external API dependencies, maximizing reliability for core logic.
- Cost Arbitrage: Pay only for base-layer data storage and compute.
- Emerging Stack: Ponder + Ethereum + QuickNode RPC is the new The Graph for many teams.
$0
Middleman Tax
Purpose-Built
Architecture
deep-dive
THE UNBUNDLING
The Developer's Dilemma: Build vs. Buy in a Multi-Indexer World
The Graph's monolithic stack is being challenged by specialized, composable alternatives that force a fundamental architectural choice.
The Graph is a monolith. It bundles subgraph definition, indexing, query processing, and economic security into one protocol. This creates a single point of failure and a one-size-fits-all performance model that cannot be optimized for specific use cases like high-frequency DeFi or low-latency gaming.
Specialized indexers are unbundling the stack. Protocols like Goldsky (event streaming) and Subsquid (batch-processed datasets) decouple data ingestion from query serving. This allows developers to mix-and-match components, choosing a bespoke data pipeline optimized for cost, speed, or freshness.
The buy decision is now a composition decision. 'Buying' no longer means using The Graph's full stack. It means assembling a pipeline from Subsquid for historical data, Goldsky for real-time streams, and a decentralized RPC like POKT for base layer access. The core value shifts to the subgraph schema standard itself.
Evidence: Goldsky's sub-second latency for NFT mint tracking and Subsquid's 100x faster historical syncs for Arbitrum demonstrate that vertical specialization beats horizontal generalization. The Graph's network queries have plateaued as these alternatives gain traction.
counter-argument
THE MOAT
Counterpoint: The Graph's Network Effects and The Long Game
The Graph's defensibility stems from its entrenched developer ecosystem and the high cost of replicating its decentralized data layer.
The Graph's primary moat is developer inertia. Thousands of dApps, from Uniswap to Lido, have production queries hardcoded to subgraph endpoints. Migrating to a new indexing stack requires rewriting application logic, a non-trivial cost for established protocols.
Decentralized indexing is a coordination nightmare. Replicating The Graph's network of Indexers, Curators, and Delegators is a massive operational lift. Competitors like Subsquid or Goldsky offer managed services but lack the same credible neutrality and liveness guarantees.
The long game is query standardization. The Graph Protocol is pushing for subgraphs to become the SQL for Web3. If GRT becomes the settlement layer for verifiable queries, it creates a winner-take-most data marketplace that pure infra providers cannot match.
Evidence: Over 1,000 subgraphs serve data for protocols managing $10B+ in TVL. The network has paid out over $500M in query fees to Indexers, creating a powerful economic flywheel.
takeaways
THE GREAT UNBUNDLING
Takeaways: The Indexing Landscape in 2025
The Graph's monolithic model is under siege by specialized protocols targeting specific performance and cost vectors.
01
The Problem: The Graph's Monolithic Bottleneck
The Graph's general-purpose indexing forces all apps to pay for a full-stack they don't need. Its decentralized but slow consensus layer adds latency and cost for high-frequency data. The curation market is a capital efficiency trap for niche subgraphs.
- ~2-5 second indexing latency is unacceptable for DeFi
- Curation bonding locks capital in illiquid, speculative assets
- One-size-fits-all model fails for real-time or private data
2-5s
Indexing Latency
High
Capital Lockup
02
The Solution: Hyper-Purposeful Indexers (Goldsky, SubQuery, Envio)
New entrants are unbundling the stack by offering application-specific performance. They bypass The Graph's network layer for dedicated RPC endpoints and real-time streams. This is the infrastructure for on-chain order books and intent-based architectures like UniswapX and CowSwap.
- Sub-500ms finality-to-query latency
- Pay-as-you-go pricing vs. speculative token bonding
- Deterministic indexing enables complex event-driven logic
<500ms
Query Speed
OpEx
Pricing Model
03
The Solution: Decentralized Verifiability (TrueBlocks, KYVE)
Specialized protocols are solving for data provenance and integrity, not just availability. They provide cryptographic proofs that indexed data matches canonical chain history. This is critical for regulatory compliance, audits, and bridging where trust minimization is non-negotiable.
- Local-first indexing with client-side validation
- Arweave-backed immutable data provenance
- Enables trust-minimized bridges and oracles
Proofs
Data Integrity
Local
Execution
04
The Outcome: The Graph Becomes a Settlement Layer
The Graph's long-term moat shifts from being the only indexer to being the decentralized arbitrator. Its role evolves into a verification and dispute resolution layer for data attested by faster, specialized indexers. Think Ethereum to Polygon/Avalanche.
- GRT secures attestations, not raw queries
- Niche indexers handle performance, The Graph handles finality
- Modular design wins; monolithic stacks get disaggregated
Settlement
New Role
Modular
Architecture
ENQUIRY
Get In Touch
today.
Our experts will offer a free quote and a 30min call to discuss your project.
NDA Protected
Confidentiality24h Response
Time to ValueDirectly to Engineering Team
No Sales Fluff10+
Protocols Shipped
$20M+
TVL Overall