Skip to main content

Goldsky vs SQD

Last updated

A practical comparison of architecture, developer experience, streaming, RPC, and chain support between Goldsky and SQD.

Goldsky vs SQD: the verdict

SQD is a rebrand of Subsquid, dated July 2, 2024.⁠[1] It runs a decentralized data lake, SQD Network, behind a streaming Portal API, with a TypeScript Squid SDK and Pipes SDK on top.⁠[2]⁠[6] SQD Cloud hosts a customer’s Squid, and SQD 360 is a paid enterprise engagement arm.⁠[7]⁠[8]

Goldsky runs managed Subgraphs, backwards-compatible with graph-node, and Turbo Pipelines that stream onchain data into a customer’s own database.⁠[18]⁠[21]

SQD is genuinely stronger on three points: an open-source, self-hostable stack, a decentralized network instead of a single vendor, and zero egress fees.⁠[11] Goldsky answers with a managed platform that deploys existing subgraphs unchanged and runs streaming pipelines without customer-operated infrastructure.

SQD

Why SQD may be ideal for you:

  • Open-source, self-hostable Portal node and SDKs
  • Decentralized network of 2,000+ worker nodes
  • No egress fees on Portal or Cloud reads

Why SQD may not be ideal:

  • No drop-in subgraph deploy. Handlers need a TypeScript rewrite
  • Managed warehouse delivery only inside the $5,000+/month SQD 360 engagement
  • SOC 2 is in progress, Enterprise plan only

Choose SQD if open-source, decentralized infrastructure and zero egress fees outweigh a drop-in subgraph deploy.

Goldsky

Why Goldsky may be ideal for you:

  • Deploy an existing subgraph unchanged, from its IPFS hash
  • Vendor-managed Turbo pipelines to 11 sink types
  • Published SOC 2 Type II compliance

Why Goldsky may not be ideal:

  • Vendor-managed only. No self-hosted or open-source option
  • No Substrate parachain support
  • No published percentile latency figure

Choose Goldsky if a managed platform with drop-in subgraph support matters more than self-hosting.

Compare products, architecture, and chain support

SQD builds on a decentralized network with a Portal API, a Squid SDK, and a Pipes SDK.⁠[6] Goldsky answers with Subgraphs and Turbo Pipelines, plus Edge RPC.

Quick comparison of SQD and Goldsky across key dimensions
FeatureSQDGoldsky
Architecture
Deployment modelAdvantage: SQD Cloud, or a self-hosted, AGPL-3.0 Portal node⁠[2]Vendor-managed only. No self-hosted option
Network operatorsAdvantage: 2,000+ independent worker nodes, each staking 100,000 SQD tokens⁠[4]Single vendor, Goldsky-operated infrastructure
Developer experience
Build pathsPortal HTTP API, Pipes SDK, or Squid SDK, all TypeScript-first⁠[6]CLI-based subgraph deploy, or a Turbo pipeline config⁠[18]
Subgraph migrationRewrite AssemblyScript handlers into TypeScript for the Squid SDK⁠[10]Advantage: Deploy an existing subgraph unchanged, from its IPFS hash⁠[19]
Streaming and data delivery
Managed streaming pipelineNone at self-serve pricing. Pipes SDK runs on the customer’s own process⁠[11]Advantage: Turbo Pipelines, vendor-managed, 11 sink types⁠[21]
Egress feesAdvantage: None, stated on the Portal and Cloud pricing pages⁠[2]⁠[17]Goldsky publishes no egress fee and no egress guarantee
RPC services
Standalone RPC productNone. Portal says “Not an RPC replacement”⁠[2]Advantage: Edge RPC: JSON-RPC over HTTPS for EVM networks⁠[23]
RPC pricing2,000,000 requests a month free for Professional organizations, then $5.00 per million⁠[14]$5 per million requests, every method priced the same⁠[23]
Chain support
Chains and networks68 chains across 120 networks⁠[15]106 chains with subgraph support⁠[20]
Non-EVM breadthAdvantage: 10 Substrate chains, plus Hyperliquid⁠[15]No Substrate support. Turbo covers Hypercore⁠[20]

Architecture

SQD Portal is a streaming access layer in front of SQD Network, not the storage layer itself.⁠[3] Network data sits in immutable Parquet files, replicated 10 to 34 times across independent worker nodes.⁠[4] Each node runs on 4 vCPU, 16GB RAM, and a 1TB SSD, and stakes 100,000 SQD tokens.⁠[5]

The Portal node is AGPL-3.0 and self-hostable, but querying the Network at scale still needs locked SQD tokens.⁠[5] SQD is clearly stronger on open source and decentralization here. Goldsky runs Subgraphs and Turbo Pipelines as a vendor-managed service, with no self-hosted or open-source option.⁠[18]

SQD’s data model covers raw blocks, transactions, logs, traces, and state diffs for EVM. Solana gets instructions and token balances. Substrate gets calls and events.⁠[5] Aggregation runs client-side in the SDK, or inside a hosted Squid on SQD Cloud.⁠[7]

Developer experience

SQD offers three ways to build. The Portal HTTP API works from any language. The Pipes SDK is a TypeScript ETL library. The Squid SDK is a TypeScript framework with Postgres and auto-generated GraphQL built in.⁠[6] The sqd CLI drives local development.⁠[12] Local dev needs Node.js 20+, Git, and Docker Compose, with no RPC credentials needed, since the starter template streams from the public Portal.⁠[12]

Moving a subgraph to SQD has two paths. An alpha-quality Firehose adapter lets a real graph-node ingest from SQD Network.⁠[9] The documented path migrates the subgraph into a Squid, carrying schema.graphql over largely as-is, but it requires rewriting AssemblyScript handlers into TypeScript batch processors.⁠[10] SQD’s own comparison page states this. Goldsky hosts existing subgraphs drop-in. On SQD a subgraph becomes a Squid, so factor in a one-time migration to port the mappings to TypeScript.⁠[11] Goldsky deploys a subgraph unchanged from its existing IPFS deployment hash, with no manifest or schema change.⁠[19]

SQD claims its own Squid SDK batch processor indexes about 1,000 to 50,000 blocks a second. This is SQD’s own benchmark claim, not independently verified.⁠[13]

Streaming and data delivery

The Pipes SDK targets ClickHouse, Postgres, BigQuery, and Parquet, but the customer runs it. SQD’s own comparison page labels it “Pipes SDK (self-run)”.⁠[11] SQD Cloud has no managed streaming sink of its own.⁠[11] A managed, SQD-run delivery to ClickHouse, Snowflake, BigQuery, or S3 exists only inside the SQD 360 engagement, which starts at $5,000 a month.⁠[8]

Goldsky Turbo Pipelines are vendor-managed, with 11 sink types including Postgres, ClickHouse, Kafka, S3, and EventBridge.⁠[21] Transforms run in SQL through Apache DataFusion, or in TypeScript.⁠[22]

SQD charges no egress fees, on both the Portal and Cloud pricing pages. That is a real advantage over a managed platform with no published egress policy.⁠[2]⁠[17]

RPC services

SQD has no standalone RPC product. The Portal page says “Not an RPC replacement”.⁠[2] An RPC add-on inside SQD Cloud gives Professional organizations 2,000,000 requests a month free, then $5.00 per million.⁠[14]

Goldsky Edge RPC is a standalone product: JSON-RPC over HTTPS for EVM networks, at $5 per million requests, with every method priced the same.⁠[23]

Chain support

SQD covers 68 chains across 120 networks: 53 EVM, 10 Substrate, 2 Solana, 1 Bitcoin, 1 TVM, plus Hyperliquid.⁠[15] 38 of the 68 chains are marked real-time.⁠[15]

Goldsky’s coverage matrix lists 106 chains with mainnet or testnet subgraph support, and no Substrate entries.⁠[20] Turbo pipelines cover Hypercore. Substrate parachains are a genuine SQD advantage.

Goldsky’s current chain list is on the chains page.

Pricing

SQD’s Portal tiers are published ahead of launch and not yet billing.⁠[16] SQD Cloud Playground is free, and Professional compute runs $0.048 to $0.72 an hour.⁠[17]

SQD Cloud markets a “>99.9% Uptime SLA”. The mechanics apply only to dedicated Squids on Professional organizations, and deliver a bill reduction rather than a flat guarantee. The reduction is 10% from 95.00% to under 99.50%, 25% from 90.00% to under 95.00%, and 100% under 90%.⁠[17] SOC 2 is “in progress”, Enterprise only, not yet certified.⁠[16] Goldsky publishes SOC 2 Type II compliance today.⁠[25]

SQD publishes a query-response latency figure of 27ms P50 and 48.5ms P90. An unnamed third party ran the benchmark, and SQD published it on its own Portal page.⁠[2] Goldsky publishes no equivalent percentile latency figure.

Pricing comparison between SQD and Goldsky
FeatureSQDGoldsky
Entry Portal planStarter, $99/month, announced ahead of billing⁠[16]Starter plan, $100 one-time signup credit⁠[24]
Subgraph-style indexing computeNo subgraph product. Squid SDK runs on SQD Cloud compute⁠[17]First 3 subgraphs free, then $0.05 per worker-hour⁠[24]
Streaming computeCloud compute $0.048 to $0.72 an hour⁠[17]Turbo $0.10 per worker-hour, plus $1.00 per 100,000 events⁠[24]
Storage$0.60 per GB a month⁠[17]First 100,000 entities free, then about $3.89 per 100,000 a month⁠[24]
EgressNone⁠[2]No egress fee published, and no stated guarantee
Enterprise engagementSQD 360, from $5,000 a month⁠[8]Enterprise contract, custom terms

Goldsky publishes full pricing details, including volume discounts, on its pricing page.

When to choose Goldsky, and when to choose SQD

Choose SQD if:

  • You want an open-source, self-hostable indexing stack
  • Decentralized infrastructure matters more to you than a single managed vendor
  • You index Substrate parachains such as Polkadot, Kusama, or Bittensor
  • Zero egress fees matter for a large, read-heavy workload
  • Your team already writes TypeScript batch processors

Choose Goldsky if:

  • You run existing subgraphs and want them to deploy unchanged
  • You want a vendor-managed streaming pipeline, not a self-run library
  • You want a published SOC 2 Type II report today
  • You want a standalone RPC product with flat per-request pricing
  • You want warehouse delivery without a custom, five-figure engagement

Frequently asked questions

It depends on the workload. SQD fits teams that want an open-source, decentralized stack and are willing to run their own pipeline code.⁠[11] Goldsky fits teams that want a vendor-managed platform with drop-in subgraph support.⁠[18]

SQD runs a decentralized network of 2,000+ worker nodes behind a streaming Portal API.⁠[4] Goldsky runs managed Subgraphs and Turbo Pipelines on its own infrastructure.⁠[18]⁠[21] SQD’s own comparison page summarizes the gap: on SQD, a subgraph becomes a Squid, and factors in a one-time migration to port the mappings to TypeScript.⁠[11]

Not as a drop-in deploy. SQD’s documented path carriesschema.graphql over largely as-is, but requires rewriting AssemblyScript mapping handlers into TypeScript batch processors for the Squid SDK.⁠[10] An alpha-quality SQD Firehose adapter can run an unmodified subgraph against SQD Network instead.⁠[9] Goldsky, by contrast, deploys an existing subgraph unchanged from its IPFS hash.⁠[19]

SQD’s Portal tiers are published ahead of launch and not yet billing.⁠[16] SQD Cloud compute runs $0.048 to $0.72 an hour, with no egress fees.⁠[17] Goldsky’s Starter plan draws down a $100 signup credit, and Turbo bills $0.10 per worker-hour plus $1.00 per 100,000 events written.⁠[24]

SQD covers 68 chains across 120 networks, including 10 Substrate chains and Hyperliquid.⁠[15] Goldsky’s coverage matrix lists 106 chains with subgraph support, and no Substrate entries.⁠[20] Turbo pipelines cover Hypercore, so Substrate parachains are the clearer SQD advantage.

Sources

All sources were checked on October 28, 2026.

Show 25 sources

Ready to try Goldsky?

Start with the free tier to try subgraphs, data streaming, and multi-region RPC. No credit card needed.

© Endless Sky Inc. All rights reserved.

PrivacyTermsSecurity

System status