Best Ethereum RPC Providers: A Full Comparison

If you're shipping anything serious on Ethereum mainnet, the public RPC is not an option. Public endpoints aggressively rate-limit, drop connections under load, and don't expose archive or trace methods. For a chain settling roughly 1.2M transactions a day across DeFi, rollups, restaking protocols, and stablecoin rails, that's a non-starter from day one.

Choosing between paid providers is harder than it looks. Most providers advertise pricing in compute units or credits, and these aren't standardized. A single eth_call costs 1 credit on Dwellir, 20 on Quicknode, 26 on Alchemy, 80 on Infura, and 200 on Ankr. A provider advertising $0.10 per million credits can quietly cost $20 per million actual calls once you account for the multiplier. The only reliable way to compare is to normalize cost per real method call against the workload your application actually runs.

We looked at five providers offering production-grade Ethereum RPC: Quicknode, Alchemy, Infura, Ankr, and Chainstack. Here's how they compare as of May 2026.

Why Ethereum RPC requirements are different from other chains

Before getting into providers, a quick note on what makes Ethereum specifically demanding.

Ethereum's 10-year history makes archive access a real requirement for analytics, block explorers, tax tools, and anything querying historical state. Not every provider includes archive on lower tiers. Trace and debug methods (debug_traceTransaction, trace_block, the full suite) are infrastructure-expensive, and providers price them very differently — anywhere from a 2x multiplier on the standard credit rate to a 1000-credit flat charge per call. For trace-heavy workloads, that single decision can swing your monthly bill by an order of magnitude.

MEV protection is also a real consideration on L1 in a way it isn't on most L2s. Front-running and sandwich attacks have extracted billions from unprotected transactions. If your app submits swaps, liquidations, or any value-bearing transaction, routing through a private mempool or Flashbots-integrated endpoint isn't optional, it's a security requirement.

And finally, most Ethereum teams aren't only on Ethereum. They're on Arbitrum, Base, Optimism, and Polygon too. A provider that covers mainnet plus the L2 stack under one billing account materially reduces operational overhead.

Quicknode: lowest measured latency, full compliance posture, Streams for push delivery

Quicknode has been operating Ethereum infrastructure since 2017. The platform runs on a 30+ region edge network with automatic failover and full archive access on every plan, including the free tier.

The differentiator is the data layer on top. Streams handles real-time onchain delivery — filtered, push-based, with backfill and reorg handling — into PostgreSQL, Snowflake, S3, or a webhook. For most production workloads, replacing polling with push delivery is the single biggest cost lever available. A polling-based DeFi indexer running eth_getLogs on a tight interval can burn through credits fast. The same workload routed through Streams subscribes per market or event type and only consumes credits on actual data delivery.

Performance is measurable, not marketed. QuickLee is a public benchmark that runs eth_getBalance every 10 seconds across five regions on AWS, GCP, and Oracle Cloud:

• Average latency: ~29 ms
• Median latency: ~30 ms
• P95 latency: ~45 ms
• Failed calls: 0 across ~148K requests

Full historical archive is on every plan, free included. Trace and debug APIs are accessible on all paid plans at 2x the standard credit rate, which reflects the actual infrastructure cost of those operations rather than burying them behind enterprise tiers.

Overage runs $0.62/M credits on Build, dropping to $0.50/M on Business. Annual billing saves 15%.

In March 2026, Quicknode launched Flat Rate RPS for high-volume workloads: $799/month for 75 RPS on EVM chains, with no credit metering and no overage risk. It's purpose-built for trading bots, arbitrage workflows, and anything else doing continuous polling at sustained throughput. For workloads above roughly 50M calls/month, Flat Rate RPS is typically lower effective cost than any credit-based plan.

On compliance, Quicknode holds SOC 1 Type II, SOC 2 Type II, and ISO 27001, audited by Grant Thornton and recertified in Q1 2026. The 99.99% uptime SLA is contractual on enterprise tiers. No other provider in this comparison holds the full compliance trifecta.

Alchemy

Alchemy is one of the largest Ethereum RPC providers, well-known for its enhanced APIs (NFT, token, transaction simulation, wallet activity) and the most generous free tier in this comparison: 30M compute units per month, roughly 1.15M eth_call requests at 26 CU each.

Method weighting is where the cost surprises happen. eth_getLogs runs 60 CU and debug_traceTransaction runs 40 CU. For a log-heavy indexer or any application doing meaningful trace work, effective cost climbs faster than the per-million CU price suggests.

Alchemy holds SOC 2 Type II but does not have public SOC 1 or ISO 27001 certifications. There's no Streams-equivalent for push-based real-time data, so teams needing filtered delivery end up building indexer and webhook infrastructure themselves. There's also no self-serve dedicated node access for teams that need isolated performance guarantees.

For early-stage NFT or DeFi prototyping where the free tier matters and the compliance gap doesn't, Alchemy is a reasonable starting point. For production workloads that scale into trace, archive, or regulated environments, the gap widens.

Infura

Infura, by Consensys, is one of the original Ethereum RPC providers and remains deeply integrated with the ecosystem. It powers MetaMask, which alone routes a meaningful share of all Ethereum network traffic.

The unusual feature here is daily credit reset rather than monthly. A traffic spike exhausts the day's allocation, not the month's, which is structurally helpful for workloads with unpredictable bursts.

Method weights matter, especially for trace work. eth_call is 80 credits, eth_getLogs is 255 credits, debug_traceTransaction is 1000 credits. At 1000 credits per trace call, a Developer-tier daily quota (15M credits) gets exhausted in 15,000 trace requests. For analytics platforms, indexers, or any application that needs trace at production volume, this is a hard ceiling.

Infura does not publish SOC 2, SOC 1, or ISO 27001 documentation. There's no Streams-equivalent, no built-in indexing, and limited self-serve dedicated node options. The MEV protection through Flashbots integration is a real strength for DeFi applications, and the operational depth on Ethereum specifically is unmatched.

For MetaMask-adjacent tooling, small workloads with bursty patterns, or teams that need MEV-protected transaction submission, Infura is well-positioned. For trace-heavy applications, regulated environments, or production data pipelines, the structural gaps are material.

Ankr

Ankr offers Ethereum RPC through a globally distributed network of node operators. The simplest pricing model in this comparison, but with a structural quirk that significantly affects effective cost.

The catch: standard HTTP eth_call costs 200 credits. At $0.10 per million credits, that works out to roughly $20 per million eth_call requests, which is the highest effective rate in this comparison.

The gRPC option changes the math entirely. On Premium plans, gRPC method calls cost 10 credits each, a 20x reduction from HTTP. For high-volume workloads where gRPC integration is feasible, the unit economics shift dramatically. The trade-off is that teams have to handle gRPC integration themselves, which isn't free engineering time.

Ankr does not publish SOC 2, SOC 1, or ISO 27001 documentation. SLA commitments range from 99% on lower tiers to 99.99% only on Enterprise. There's no built-in streaming, indexing, or event-driven workflow support.

For non-critical or experimental workloads where budget is the primary constraint and the team can build the rest of the stack themselves, Ankr fits. For regulated environments or production data pipelines, it doesn't.

Chainstack

Chainstack focuses on giving teams direct ownership over nodes, regions, and scaling behavior. The pricing model is the cleanest of the credit-based providers: 1 RU per standard call, 2 RU per archive or debug call, and that's the whole pricing model.

Overage runs $20/M RU on Developer down to $5/M on Enterprise. Annual billing saves 16%.

Where Chainstack stands out is deployment control. Self-serve dedicated nodes start at $199/month, and multi-cloud placement (AWS, GCP, Azure) plus regional choice means teams can colocate nodes near their application infrastructure. For latency-sensitive workloads where the geographic relationship between application and node matters, this is real.

Archive access requires a paid plan, which differs from Quicknode's all-plans-free archive policy. There's no Streams-equivalent for push-based real-time delivery, so teams that need filtered delivery handle it themselves.

Chainstack holds SOC 2 Type II. It's the right fit for teams that want granular control over their deployment topology and have the in-house engineering capacity to manage multi-region, multi-cloud configuration.

True cost per 1M calls compared

Here's the normalized cost comparison across five of the most-used Ethereum RPC methods at each provider's standard paid entry tier. Lower numbers are better.

*Infura: daily quota model rather than per-call billing. Effective cost depends on quota consumption rate. debug_traceTransaction at 1000 credits per call exhausts a Developer tier (15M credits/day) in 15,000 calls.

*Ankr: HTTP pricing shown. On gRPC (Premium plans), per-call cost drops to 10 credits, roughly a 20x reduction.

Raw credit cost is only part of the picture. For most production Ethereum dApps, the workload is read-heavy, dominated by eth_call, eth_getLogs, and eth_getBalance. On those specific methods, Chainstack shows the lowest sticker cost on paper. But effective cost depends on three factors raw pricing tables don't capture:

1. What polling-replacement infrastructure you'd have to build. Streams replaces a polling-heavy eth_getLogs workflow with push delivery. The credit savings on polling alone often dwarf the per-call price difference between providers.
2. Whether trace work is part of your stack. A 2x multiplier for debug methods (Quicknode, Chainstack) is materially different from a 1000-credit flat charge per call (Infura). For trace-heavy applications, this single line item drives the bill.
3. What reliability and uptime cost you in production incidents. A provider that drops connections under load or fails over inconsistently costs real money in the form of failed transactions, missed liquidations, and engineering time spent on incident response.

Feature comparison at a glance

Which provider for which workload

Different workloads have different priorities. Here's how the providers actually map.

DeFi protocols and trading bots

Quicknode and Chainstack. Both offer low-latency WebSocket and dedicated node options for trading workloads. Quicknode leads on measured performance with ~29ms average and ~45ms p95 in QuickLee benchmarks. Method weighting favors eth_call-heavy patterns, and Flat Rate RPS at $799/month for 75 RPS removes overage risk for continuous polling and arbitrage workloads.

NFT platforms and wallets

Quicknode and Alchemy. Both have enhanced APIs and archive access relevant to NFT workloads. Quicknode pairs free-tier archive with production-grade indexing, metadata, and media delivery, plus Streams for real-time mint and transfer tracking. Alchemy's enhanced NFT APIs are the most polished out of the box.

Enterprise and regulated applications

Quicknode is the only provider in this comparison with the full SOC 1 + SOC 2 + ISO 27001 stack, audited by Grant Thornton and recertified in Q1 2026. For fintech, payments, and regulated deployments, the compliance posture is the deciding factor.

Trace-heavy analytics and indexers

Quicknode and Chainstack at 2x trace multiplier are materially cheaper than Infura's 1000-credit flat rate. Quicknode's /nanoreth-equivalent full archive plus Streams handles the polling-replacement piece. Chainstack's dedicated node option works for teams that want predictable trace pricing on isolated infrastructure.

High-volume data pipelines

Quicknode Streams and Ankr gRPC. Ankr's Premium gRPC endpoints offer a 20x per-call discount for teams comfortable building gRPC integration. Quicknode Streams takes a different approach: real-time filtered delivery to PostgreSQL, Snowflake, S3, or a webhook with backfill and reorg handling built in. For pipeline workloads, replacing polling with push delivery typically reduces effective cost more than any per-call discount.

Cross-chain and multi-chain applications

Quicknode covers 80+ chains under a single account. For teams building across Ethereum mainnet plus the L2 stack (Arbitrum, Base, Optimism, Polygon) and beyond, single-provider coverage means one platform for auth, billing, monitoring, and compliance. The operational overhead of managing multiple RPC relationships at production scale is real, and consolidating it under one provider has compounding benefits as the team scales.

Frequently asked questions

Which is the best Ethereum RPC provider in 2026?

For most production workloads, Quicknode offers the most complete stack: lowest measured latency in public benchmarks, the only full SOC 1 + SOC 2 + ISO 27001 compliance posture in this comparison, and built-in Streams for push-based real-time data. The "best" provider depends on workload: high-frequency trading, regulated environments, polling-heavy indexing, and prototyping all have different optimal answers.

How do compute units and credits actually affect pricing?

Different methods cost different amounts. A lightweight eth_blockNumber call and a heavy eth_getLogs call are both one request but consume very different credit amounts. Two providers advertising the same per-million credit price can produce a 4x difference in actual monthly spend because of method weighting. The only reliable comparison is normalized cost per real method call, calculated as: (CU weight × price per million CUs) ÷ 1,000,000, multiplied by monthly call volume.

Which Ethereum RPC providers are SOC 2 certified?

Quicknode (SOC 1 Type II, SOC 2 Type II, ISO 27001), Alchemy (SOC 2 Type II), and Chainstack (SOC 2 Type II) hold public certifications. Infura and Ankr do not publish SOC 2 documentation. For regulated applications, Quicknode is the only provider in this comparison with all three major certifications.

Do I need archive node access?

Archive is required when querying historical state from before the most recent ~128 blocks. Use cases that need it: analytics dashboards, block explorers, tax tools, historical balance queries, and any app computing state from before a given block. Full nodes are sufficient for DeFi swaps, wallet balance checks, and transaction submission. Quicknode includes archive on every plan including free. Chainstack requires a paid plan.

What's the cheapest Ethereum RPC provider?

Sticker price is rarely the cheapest total cost. Effective cost depends on method weighting against your actual workload, polling-replacement infrastructure you'd otherwise have to build, and uptime/reliability cost in production incidents. For high-throughput sustained workloads, Quicknode's Flat Rate RPS at $799/month for 75 RPS is typically lower effective cost than any credit-based plan once sustained throughput crosses roughly 50M calls/month.

How do I switch RPC providers without downtime?

Run both endpoints in parallel before cutting over. Most Ethereum clients (Ethers.js, Web3.js, Viem) support fallback providers or provider arrays. Set the new provider as primary and the old as fallback, monitor error rates and latency for at least 24 hours, then drop the old provider. Watch for method compatibility (custom enhanced APIs may not be portable), WebSocket reconnection behavior, and rate limit differences that could affect bursty workloads.