Solana writes a new block every few hundred milliseconds. Bitcoin takes ten minutes on average. Why doesn't Bitcoin just build faster?
Because speed costs something else. The bill is called the blockchain trilemma, and it explains why Bitcoin, Ethereum and Solana are built so differently. Holding one of the three also means holding a decision about where it pays that price.
The trilemma is a price list, not a law of nature
The best-known version sits in the Sharding FAQ of the Ethereum wiki, traceable to March 2017: blockchains can have at most two of three properties. Vitalik Buterin defined the three in 2021:
- Scalability: the chain processes more transactions than a single regular computer, such as a laptop, can verify.
- Decentralization: the chain runs without depending on a small group of large actors.
- Security: the chain holds up when a large share of participants attacks it, ideally up to half.
The catch is in the first line. If anyone should be able to verify the chain on a normal computer, the chain cannot process more than such a computer handles. More throughput then means bigger computers, so fewer people checking, or work moved off the main chain.
Buterin writes that with "simple" techniques you only get two of the three. The trilemma does not say scaling is impossible. It says someone pays for it.
The question is where. Bitcoin, Ethereum and Solana have given three different answers.
Bitcoin pays with capacity
Bitcoin caps every block at a weight of four million units (BIP 141, in force since SegWit in 2017) and writes one every ten minutes on average. That keeps the data small enough for an ordinary computer to verify the whole chain. Bitnodes counted 25,343 reachable Bitcoin nodes on 25 September 2026.
Volume is meant to move one level up, mainly to the Lightning Network: payment channels that run off the chain and settle on it at the end. That level is not growing right now. Publicly visible Lightning capacity stood at 5,851 BTC on 28 December 2025 and at 3,795 BTC on 30 August 2026, per mempool.space. That is a good third less.
Bitcoin keeps verification cheap and sends the volume upstairs. Whether anyone arrives upstairs is the open point of that choice.
Ethereum took the same road in 2020, only further. And it is now walking part of it back.
Ethereum pays with fragmentation and is now pulling load back
In October 2020 Buterin wrote that Ethereum would be "all-in on rollups" for the foreseeable future: separate chains that bundle transactions and post only data and proofs to Ethereum. Since then Ethereum has mostly made room for that data, in so-called blobs:
| Upgrade | Date | Blob target per block |
|---|---|---|
| Dencun | 13 March 2024 | 3 |
| Pectra | 7 May 2025 | 6 |
| Fusaka, BPO1 | 9 December 2025 | 10 |
| BPO2 | 7 January 2026 | 14 |
The target grew almost fivefold in under two years. Activity moved accordingly: in the 30 days to 23 September 2026, L2BEAT counted 98.9 % of all operations on rollups and 1.1 % on Ethereum itself. Most of that, though, comes from a single application, the derivatives exchange Lighter.
The price of this choice is fragmentation: users, money and liquidity spread across many chains, each with its own rules and its own operators. Buterin himself wrote in February 2026 that rollups had decentralized "far slower" than expected and that the original vision of their role "no longer makes sense".
Since then the main chain has been growing again. The gas limit, the amount of computation per block, rose from 36 million in February 2025 to 45 million in July and 60 million in November 2025. The Ethereum Foundation names the goal as "towards and beyond 100M".
Ethereum is pulling load back onto its main chain, and that moves it towards the corner where Solana already stands.
What that corner costs, Solana shows.
Solana pays with the number of people who can check
Solana runs a single chain with nothing moved off it, and bets on fast blocks and heavy computing power. The slot time, the beat at which blocks are produced, was 400 milliseconds for years and has been cut in steps since August 2026, to 300 on 25 August per solana.com. The computation allowed per block rose from 60 to 100 million units on 29 July 2026.
That takes hardware. The Anza documentation lists at least 12 CPU cores, 256 GB of memory, several terabytes of NVMe storage and a symmetric 2 Gbit/s line, with 10 recommended, for a validator. That is not a laptop. It is a server in a data centre.
The number of validators has fallen accordingly. Solana Compass counted about 2,560 in March 2023 and 795 at the end of January 2026. A query of the Solana network on 25 September 2026 returned 685 validators with stake; 18 of them together hold more than a third of the stake, enough to halt the network.
The choice has also delivered: per Solana Compass, there has been no network-wide outage since 6 February 2024. Solana is fast and by now stable. It pays with the number of people who can check along.
The obvious objection
"Ethereum has almost 900,000 validators. Isn't decentralization settled there?"
The number is right: validatorqueue.com counted 895,222 active validators holding 43.2 million ETH on 25 September 2026. But it counts accounts, not operators. Since Pectra a validator may hold up to 2,048 ETH instead of 32, so large providers consolidate into fewer accounts. A validator count measures accounts. How many independent operators stand behind them is in none of the three numbers.
The number to watch
| Chain | Where it pays | Number that would have to move | Status |
|---|---|---|---|
| Bitcoin | main-chain capacity | public Lightning capacity | 3,795 BTC (30 Aug 2026), 5,851 in December 2025 |
| Ethereum | fragmentation across rollups | main-chain gas limit | 60 million (25 Sep 2026), goal above 100 million |
| Solana | circle of verifiers | validators with stake | 685 (25 Sep 2026), 2,560 in March 2023 |
The table answers a question you can ask of any announcement: which number would have to move if the chain really pushes the trilemma further out? If Solana gets faster, the validator count would have to hold. If Ethereum raises the gas limit, running a node would have to stay affordable at the same time. If Bitcoin is to grow through Lightning, capacity there would have to rise.
The next checkpoints are set. Ethereum's Glamsterdam upgrade is due on the Sepolia testnet in early October 2026, with no mainnet date yet. Solana's new consensus design, Alpenglow, was approved in September 2025 with 98 % of the votes cast and has been activating on the testnet since 22 September 2026; it has no mainnet date either. Both upgrades promise more speed. For both, it is open where they will pay for it.
FAQ
Is the trilemma proven? No. The Sharding FAQ calls it a claim ("The trilemma claims"), and Buterin limits it to "simple" techniques. Sharding, rollups and data availability sampling are attempts to get around it with techniques that are not simple. Whether one of them breaks the trilemma or only moves the price shows in the number that moves afterwards.
Why no transactions per second? Because the figure depends on what counts as a transaction. A transfer, a swap on an exchange and a validator's vote cost different amounts of computation, and each chain counts differently. The per-block limits (weight, gas, compute units) are more comparable because they are fixed.
Are 25,343 Bitcoin nodes a lot? It is a count of nodes reachable from outside; nodes behind firewalls are missing, and a count can be inflated with many fake nodes. What matters for the trilemma is less the number than whether an ordinary computer can verify the chain. For Bitcoin, it can.
Not investment advice, not a recommendation, not a forecast. Historical patterns are not a promise.
Sources: Ethereum Wiki, Sharding FAQ (archived 11 March 2017) · Vitalik Buterin, "Why sharding is great", vitalik.eth.limo, 7 April 2021 · Vitalik Buterin, "A rollup-centric ethereum roadmap", Ethereum Magicians, 2 October 2020 · BIP 141 (Segregated Witness) · Bitnodes API, 25 September 2026 · mempool.space Lightning statistics (28 December 2025, 30 August 2026) · Ethereum Foundation Blog: Dencun (27 February 2024), Fusaka (6 November 2025), Protocol Priorities 2026 (18 February 2026) · ethereum.org, Pectra · L2BEAT Activity, 25 August to 23 September 2026 · The Block, 45M gas limit (22 July 2025), 60M gas limit (27 November 2025), Buterin on layer 2 (3 February 2026) · validatorqueue.com, 25 September 2026 · solana.com, Reduced Slot Times and 100M CU Blocks · Anza Docs, Validator Requirements · Solana Compass via Cointelegraph (January 2026) and Solana Compass (10 August 2026) · Solana RPC getVoteAccounts, 25 September 2026 · Solana Status, Alpenglow vote (2 September 2025) · ForkLog, Alpenglow testnet (23 September 2026) · The Cryptonomist, Glamsterdam (18 September 2026)
Study the Past — Improve your Future 🥋