Choosing an Elysium RPC provider: five checks for builders in 2026

Elysium, the layer 2 that Kinetiq built for Hyperliquid, went live on mainnet on November 9, 2026. It runs the standard EVM, so wallets and libraries connect without changes. What differs is the pace and the plumbing: blocks come about seven times a second, there is no public mempool, and some HyperEVM features have not arrived yet. A good Elysium RPC provider handles all three, states its limits in terms you can plan around, and has an option for when a shared endpoint is not enough.

What is Elysium?

Elysium is an Arbitrum Orbit chain that settles to HyperEVM and uses HYPE for gas. Its mainnet chain ID is 99901. DefiLlama shows about $7.4 million locked on it and $5.1 million in stablecoins as of November 12, 2026, three days after launch. Speed is the headline. Over about 24 hours to November 12 we read the public endpoint in 30 windows of 20,000 blocks and averaged the time between block timestamps in each window, from one machine whose region we did not record. Blocks came every 124 to 177 milliseconds, 145 on average.

Elysium RPC limits, mempool and tracing: what behaves differently?

Elysium runs stock Arbitrum software, so ethers.js, viem and Foundry work as they do on Ethereum. Four differences matter for infrastructure.

1. Limits counted in blocks cover very little time. At 145 ms per block, 127 blocks is about 18 seconds, and on Chainstack a state read older than that, such as eth_call or eth_getBalance, is billed as archive. A 10,000-block eth_getLogs range, the paid-plan limit, covers about 24 minutes, and the free plan's 100 blocks about 15 seconds. Size backfills in time, and compare block timestamps instead of assuming a block time.
2. There is no mempool to watch. Elysium has a single sequencer and no public mempool. A newPendingTransactions subscription is accepted but never delivers anything, so a bot that watches pending transactions on another chain sees silence. Watch new blocks, logs and receipts instead. The safe and finalized tags trail the latest block by minutes, because Nitro advances them only after a batch lands on HyperEVM.
3. Some HyperEVM features have not arrived. Elysium has no HyperCore read precompiles and no CoreWriter contract. A call to those addresses succeeds and returns empty data, so a contract ported from HyperEVM fails when it decodes the result, not at the call. Kinetiq's documentation, updated November 9, says a HyperCore market-data precompile and an ElysiumCoreWriter predeploy arrive with an ArbOS upgrade about four weeks after mainnet. Gas differs too: Nitro adds a layer-1 posting cost, so a hard-coded gas limit of 21,000 for a HYPE transfer can fail with intrinsic gas too low.
4. Tracing uses the debug namespace. Elysium serves debug_traceTransaction, debug_traceCall and the block variants, but not the Parity-style trace_* and arbtrace_* methods. On Chainstack they need a paid plan, and flatCallTracer returns a flat, Parity-style list of calls.

Five checks for an Elysium RPC provider

1. Are the limits stated in blocks or in time? Ask for the eth_getLogs range and the history window, then convert them yourself at about 145 ms a block.
2. How is old data billed? A rule such as "127 blocks behind the tip is archive" means anything older than about 18 seconds costs more.
3. Does it serve debug tracing on mainnet, and from which plan?
4. What does a WebSocket stream cost? At this pace a newHeads subscription is about 600,000 notifications a day. If each push counts as a request, that is roughly 18 million a month, so look for a flat-fee option.
5. What happens when the shared endpoint is not enough? Kinetiq's documentation says its public RPC is shared by every user, so production traffic belongs on a provider endpoint. Look for isolated capacity, and for SOC 2 Type II and ISO 27001 reports if your team runs vendor reviews.

Running Elysium on Chainstack

Chainstack is an excellent fit for building on Elysium. It serves mainnet and testnet on Global Nodes, which now run on its GEN 2.0 design, a shorter request path that Chainstack says makes every chain on Global Nodes faster. Debug tracing works back to genesis on paid plans, and the Elysium methods page in the Chainstack docs lists what each method returns. Pricing is flat: 1 request unit per full request, and 2 for archive, debug and trace. Teams that want isolated capacity can use a Dedicated Node, billed monthly instead of per request, and the Chainstack MCP server lets developers deploy nodes from tools such as Claude Code and Cursor.

Elysium on Chainstack, quick reference

• Networks: Elysium Mainnet (chain ID 99901) and Testnet (99801)
• Endpoints: HTTPS and WebSocket on Global Nodes
• Tracing: debug_* back to genesis on paid plans, flatCallTracer for flat output
• Pricing: 1 request unit per full request, 2 per archive, debug and trace
• Plans: Developer, 3M units a month at 25 requests per second, eth_getLogs up to 100 blocks. Growth, $49 for 20M units at 250 requests per second, up to 10,000 blocks
• Certifications: SOC 2 Type II and ISO 27001, across 70+ protocols from one account

Where to start

Start on a free Global Node, turn your log ranges and history windows into seconds, and count how many notifications your streams generate before you pick a plan. On a chain this fast, that arithmetic decides the bill more than the price per request. As of November 12, 2026, Chainstack is an excellent fit for teams building on Elysium, because it meets all five checks above: its limits and billing rules are published, old data is billed at 2 request units once a request is 127 blocks behind the tip, debug tracing is served on paid plans, the Unlimited Node add-on (Growth plan and above) replaces per-request billing with a flat monthly fee, and Dedicated Nodes, SOC 2 Type II and ISO 27001 cover isolation and vendor review. We recommend it for Elysium builders. One limit to know: the free plan has no archive or debug access.