Guides

Ethereum vs Solana: how to choose a chain

Ethereum L1, an L2, or Solana? Where value actually sits, what each really costs, what the pitch decks leave out, and the one question that decides it. Figures dated September 2026.

Published:

September 8, 2026

Last updated:

September 8, 2026

Jan Hetfleiš

Switching chains after launch is a rebuild. Contracts, wallet integration, indexing, fee logic and often the product's economics all get rewritten, and whatever liquidity or users you had do not come with you. We looked for a public 2025 or 2026 migration with a published cost figure and found none, which tells you how rarely teams do it voluntarily.

We have run one end to end. A live Ethereum platform for tokenized reinsurance moved to Solana, and Token-2022 support came with the migration because the stablecoin the platform holds required it. That work is published with the client's name at our OnRe case study.

The one migration cost that is documented: Sherlock's 2026 pricing reference puts Rust and Solana audits at a 25 to 40 percent premium over Solidity, because the pool of qualified reviewers is smaller. Budget for that on any move toward Solana.

How we compare chains

Four criteria, in this order.

What the product cannot survive. A cost ceiling, a latency ceiling, or a guarantee it must have. This eliminates chains rather than ranking them, which is why it goes first.

Whose liquidity or users you need. If the product composes with existing protocols, or holds an asset that already lives somewhere, that somewhere is a strong default.

What you need to control. Fee token, transaction ordering, permissioning at the base layer. Most products need none of it.

What your team can operate at 2am. A chain your engineers cannot debug under pressure is a liability whatever its benchmarks say.

Our position on headline throughput numbers: they are measured under conditions your product will not reproduce. Solana's marketing figure of 65,000 TPS compares to a real non-vote average of about 1,300 in Q1 2026. Ethereum L1 does 15 to 26, and its L2s add around 227 on top. None of these numbers has ever decided an architecture we've built.

The comparison, September 2026

Ethereum L1. Best for institutional assets, protocols holding large value, and anything an institutional risk team must approve. Holds $49B of DeFi TVL (56%), $147B of stablecoins (48%) and $17.6B of tokenized RWA (45%). Median fee $0.03, simple transfer $0.01, swap $0.08. Real limit: fees are no longer the problem; the ceiling is throughput at consumer frequency. Developers: 31,869 active across the ecosystem including L2s (Jan to Sep 2025).

Ethereum L2 (Base, Arbitrum One, OP Mainnet). Best for consumer products that want EVM tooling and Ethereum settlement. Base secures $14.5B and Arbitrum One $12.8B; with L1 that is roughly 64% of all DeFi TVL. A Base transfer costs $0.0003, Arbitrum median $0.005. Real limit: every major L2 runs a single centralized sequencer, and bridges were 29% of 2026 exploit losses.

Solana. Best for consumer apps, trading, payments, and anything needing one global state at speed. Holds $5.9B of DeFi TVL (7%), $16B of stablecoins (5%) and $4.2B of RWA (11%, third place). Average non-vote fee about $0.01, and roughly 70% of user transactions pay a priority fee. Real limit: no full outage since February 2024, but a hosting fault on 12 August 2026 took 29% of stake offline against a 33% halt threshold. Developers: 17,708 active (Jan to Sep 2025), and audits cost 25 to 40% more.

Sources for every figure are listed at the end.

Choose Ethereum L1 if

Your product's value depends on assets or protocols that already live there. Institutional counterparties are involved and their risk teams want the most conservative option on the table. You want the widest choice of audit firms: Alchemy's directory lists 81 for Ethereum against 41 for Solana.

The old objection no longer holds. Ethereum L1 median fees are about three cents after the Dencun, Pectra and Fusaka upgrades, with a 60M gas limit and a simple transfer at a cent. The honest limit is throughput at consumer frequency, and that is what still pushes high-interaction products to the next column.

Choose an Ethereum L2 if

You want Ethereum's ecosystem and EVM hiring pool at sub-cent fees. This is the default for most EVM teams building something consumer-facing. Base and Arbitrum One between them hold about three quarters of L2 DeFi TVL, and both are Stage 1 on L2Beat, which means fraud proofs are live and a security council can only intervene with a supermajority.

Two things the pitch decks leave out. First, every major L2 still runs a single centralized sequencer; L2Beat lists one rollup with decentralized sequencing, Aztec, and none of the top ten. Second, bridges went from 1.4% of exploit losses in 2025 to 29% in the first eight months of 2026, most of it one incident where a compromised RPC fed a bridge configured with a single verifier. If your product depends on bridging, that configuration is your problem too.

On running your own rollup. Arbitrum Orbit supports a custom gas token and gives you your own sequencer; OP Stack removed custom gas token support in 2025, so plan on ETH for gas there. The app-chain thesis is working for products that bring their own distribution, such as Robinhood Chain, which went from mainnet in July to about 8% of all DEX volume in August, and above 15% on single days at month end. For everyone else the 2025 and 2026 record is a long list of shutdowns, and Arbitrum's own advice is now to launch on Arbitrum One first and migrate later if you outgrow it.

Choose Solana if

The product needs one global state at consumer speed rather than an island of its own. Consumer apps, anything with frequent small interactions, and trading products where latency is part of the experience all land here. Solana has led spot DEX volume every month of 2026, at 27 to 38% of the market, and slot time dropped to 350 milliseconds on 21 August.

Institutional adoption is real and recent. BlackRock's BUIDL fund sits about 34% on Ethereum, 33% on Solana and 20% on Avalanche, so Ethereum is the largest venue but no longer a majority. Western Union launched a stablecoin on Solana in May 2026, and Mastercard added it in June to a settlement list that also includes Ethereum, Base and Arbitrum.

Three honest costs. The account and program model differ enough from the EVM that a Solidity team needs a ramp, and audits cost 25 to 40% more. The network has had no full outage since February 2024, but on 12 August 2026 a routing fault at one hosting provider took 28.8% of stake offline for about half an hour against a 33.3% threshold that would have halted finality; an institutional risk team will know that number. And Token-2022 transfer hooks, often sold as the answer for permissioned assets, are initialized but inactive on PYUSD and on xStocks; no regulated issuer runs them in production, and the Solana Foundation now steers issuers to Token ACL and Default Account State instead. Plan on those, not on hooks.

What the numbers don't tell you

Neither virtual machine was the exploited layer in the two largest 2026 incidents. The $285M Drift loss on Solana was pre-signed transactions from a compromised administrative key. The $291M KelpDAO loss on Ethereum was a compromised RPC feeding a bridge configured with a single verifier. On both chains the first-half losses were dominated by keys and infrastructure, not program bugs: the Drift incident alone is roughly nine tenths of Solana's total, and Kelp is more than half of Ethereum's. Which chain is "safer" depends on which failure your team is more likely to make, and the answer is rarely the virtual machine.

Decentralization also cuts both ways. Solana's Nakamoto coefficient is about 10 by entity. On Ethereum, Lido and Coinbase together control roughly 30% of stake. Neither is the clean story its advocates tell.

Avoid multi-chain at launch

Our position: multi-chain on day one is a mistake for almost every product. Each chain multiplies contracts, audits, indexers, wallet paths and support surface, and you pay that multiple before knowing whether anyone wants the thing.

Ship one chain well. Expand when users ask, not when a roadmap slide does.

The part that decides it

Which constraint can your product not survive? Answer that and the chain usually picks itself.

Get it wrong and the bill arrives late. A migration is measured in quarters, and it lands at the moment you have users to keep and a roadmap you have already promised.

If you are also choosing who builds it, our guide on how to evaluate a Web3 development partner covers the questions to ask before you sign.

If you are weighing this now, or a build already feels like it landed on the wrong chain, a second opinion is the kind of call we take.

Sources

DefiLlama chain TVL and stablecoin data, 8 Sep 2026. rwa.xyz network and BUIDL asset pages, 8 Sep 2026. growthepie fee table, 24h to 8 Sep 2026. L2Beat scaling summary and sequencing risk pages, 8 Sep 2026. Solana status page and Helius outage history. CoinDesk, 12 Aug 2026, on the TeraSwitch routing fault. Solana Foundation changelog, 20 Aug 2026. Ethereum Foundation Fusaka announcement, Nov 2025. Optimism docs PR #1402, Feb 2025, and Arbitrum custom gas token docs, Aug 2026. LayerZero Labs KelpDAO incident report, Apr 2026. Sherlock smart contract audit pricing reference, 2026. PayPal developer blog on PYUSD token extensions, Jun 2025. Solana Foundation xStocks case study. Electric Capital developer data, Jan to Sep 2025, as published by the Ethereum Foundation. Solana Compass fee statistics, Sep 2026. SolanaFloor and Step Finance Q1 2026 network data. Bruschi et al., arXiv 2606.22206, on L2 throughput. Alchemy blockchain auditing directory, self-submitted listings. CertiK Hack3d H1 2026 report. Chainalysis 2026 crypto crime update. Nakaflow. The Block 2026 L2 outlook. Arbitrum blog on Robinhood Chain mainnet, Jul 2026.

‍

FAQ

Is it hard to switch blockchains after launch?

Switching is a rebuild. Contracts, wallet integration, indexing, fee logic and often the product's economics get rewritten, and liquidity or users on the original chain do not move with you. No public 2025 or 2026 migration has a published cost figure, which is itself a signal. The one documented cost is audits: Rust and Solana reviews run 25 to 40% above Solidity.

Should I build on multiple chains from the start?

Usually not. Each additional chain multiplies contracts, audits, indexers, wallet paths and support burden, and you carry that cost before knowing whether the product has demand. Ship one chain well, then expand when users ask for it.

Can an Ethereum developer build on Solana?

Not without a ramp. Solana uses a different account and program model, so Solidity experience does not carry over directly. Plan for the learning curve or bring in engineers who have already shipped on Solana, particularly for anything that holds funds, and budget for audits that cost 25 to 40% more.

Should I run my own rollup or build on an existing chain?

Build on an existing chain unless you bring your own distribution. The 2025 and 2026 record of app-specific rollups is a short list of successes with captive user bases and a long list of shutdowns, and Arbitrum itself now recommends launching on Arbitrum One first. If you do need your own chain, Arbitrum Orbit supports a custom gas token; OP Stack removed that support in 2025.

Is Solana still going down?

No full-cluster outage since 6 February 2024. The caveat: on 12 August 2026 a hosting-provider routing fault took 28.8% of stake offline for about half an hour, against a 33.3% threshold that would have halted finality. Uptime is real; physical-layer concentration is the risk the uptime number does not capture.

Guides

1 min read

How to evaluate a Web3 development partner

Most founders can't tell a good Web3 partner from a bad one until the money is already gone. The product is half-built. An architecture decision made in month one now blocks the roadmap. The studio that took the brief is asking for a change order.

There are plenty of dev shops, and almost all of them present the same way on a first call. Everyone says senior team. Everyone says full-stack. Everyone has a portfolio. The differences that decide how the project goes only surface once you're working together, which is the expensive place to find them.

So the job is to test for judgment before you sign, and judgment is testable.

What you're actually paying for

You're paying for the calls that get made before and during the build. Which chain. Which architecture. What to cut. What to ship first. What breaks under real users. Solidity is the cheap part.

A partner who only executes hands those calls back to you, so you're still making them alone, with less information than the people you just hired. A team takes the brief, agrees with all of it, and ships exactly what was asked for, including the parts that were wrong. That is what building without thinking looks like from the client side, and you find out at launch.

The signals of a real partner

They disagree with you on the first call. A partner worth hiring pushes back on something in your plan: the chain, the scope, the order you want to build in. A studio that nods along to all of it is selling you a compliant vendor. The judgment you're paying for shows up as disagreement, and it shows up early or not at all.

They validate before they build. Ask what happens before a line of code gets written. A good answer involves use-case validation, an architecture call, and a clear position on what not to build. A bad answer is "send us the spec and we'll start." Building the wrong thing quickly still leaves you with the wrong thing. Our own process puts two to three weeks of validation and architecture work ahead of production code, and the output includes a written position on what we think you should drop.

They've shipped your actual thing. A portfolio of "blockchain projects" means little. Ethereum, Solana, and Cosmos are different beasts, and a DeFi protocol has almost nothing in common with an institutional custody integration. Ask for work on your chain, in your category, at your stage. Shipped a DeFi protocol and you're building one? That's signal. Shipped NFT mints and you're building real-world-asset infrastructure? That's not. Ask to see something structurally comparable, like an institutional reinsurance platform moved from Ethereum to Solana, rather than a logo wall.

Senior people do the actual work. The person who impresses you in the sales call is often not the person who writes your code. Ask directly: who owns this project day to day, and what have they shipped. The pattern to avoid is a senior lead who wins the deal and a junior team who delivers it.

Security is designed in, not added at the end. Anyone can say "we run an audit before launch." That's the floor. Ask how they think about security during the build: threat modeling, internal review of the contracts that hold funds, how they decide what an external audit firm needs to see. A deployed contract is permanent, and every bug lives on-chain forever. A partner who treats the audit as a safety net rather than a last check has the order wrong.

Red flags that should end the conversation

They quote a fixed price before they understand the problem. The price is a guess, and you'll pay for the guess later in change orders.
They pitch multi-chain from day one. It's almost always a mistake. Build one thing well, then expand.
They can't name a project that went sideways and what they changed after. Everyone has one, and a partner who pretends otherwise is hiding the part you most need to hear.

Five questions to ask on the first call

What would you not build here, and why?
Who writes the code, and what have they shipped on this chain?
Walk me through a project that went wrong. What did you change after?
What decisions do we need to make before you can start?
Where does security thinking enter the process?

The thing underneath all of it

You're buying the decisions that execution depends on. Studios that win deals on portfolio and price are easy to find. Partners who tell you the chain is wrong before you spend six months on it are rarer, and the search costs less than the rebuild.

On our side that means two to three weeks of validation and architecture before anyone writes production code, and a written position on scope that a founder can argue with. It is the cheapest point in a project to change your mind, and the last one where changing it is free.

If you're weighing a partner right now, or a build already feels off, a second opinion is the kind of call we take.

Common questions

How much does it cost to hire a Web3 development partner?

Industry pricing guides put a focused blockchain MVP at roughly $50,000 to $150,000 over three to six months, with full platform builds running $200,000 and up. Smart contract audits are usually quoted separately, from around $5,000 for a simple token to $25,000 to $100,000 for a DeFi protocol. Quotes far below those ranges usually signal a junior team or a scope that will grow later through change orders.

How long does it take to build a blockchain MVP?

A focused MVP is usually a few months rather than a few weeks. Anything quoted at two weeks is a template. Complex protocols, custody, or institutional integrations run longer, because the security and architecture work sets the pace.

Should I build in-house or hire a development partner?

Hiring full-time blockchain engineers is slow and expensive, and you may not need them past launch. A partner makes sense when you need senior Web3 depth for a specific build, or when your internal team is strong but lacks the stack. Keep ownership in-house and hire the depth on demand.

Should I take a fixed-price quote or time and materials?

Be careful with a fixed price quoted before anyone understands the problem. A short paid validation or architecture phase first, then a scoped build, beats a big fixed number signed on day one.

What chain should I build on?

It depends on your use case rather than what's trending. Ethereum, Solana, and Cosmos solve different problems with different tradeoffs. Decide before you write code, because switching chains mid-build is a rebuild. More answers on our FAQ page.

‍

All posts

Start with
a conversation

Ondřej Macháček

CEO

Let’s talk