Solana: building a high-speed censorship-resistant financial network

Executive overview

Ethereum's proof-of-work roots cap it at 11–65 transactions per second and drive gas fees to $50+. Existing blockchains treat speed as a secondary concern; Solana treats it as the primary one.

Anatoly Yakovenko, former Qualcomm engineer, invented proof of history — a cryptographic arrow of time that pre-schedules validator turns rather than forcing consensus after every block. The result is a network benchmarked at 50,000+ TPS with a theoretical hardware ceiling near 700,000 TPS at 1 gigabit bandwidth.

The core insight: censorship resistance at scale requires a system fast enough that the minimum set of validators needed to corrupt it can be made very large — and that requires raw throughput, not just clever consensus design.

The problem with existing blockchains

• Bitcoin's thermodynamic security (proof of work) is real but caps throughput at ~7 TPS; by design, the chain cannot grow quickly without undermining verifiability
• Ethereum adds smart contracts and tokens but still runs proof of work in ETH1, capping useful throughput at 11–65 TPS
• High gas fees ($40–60+ per transaction at peak) price out ordinary users and small transactions
• Most DeFi activity (e.g., Uniswap: ~100–200k transactions/day) consumes a large fraction of Ethereum's entire global capacity
• Centralized exchanges exist because on-chain execution is too slow for real markets; DeFi works around this with AMMs (automatic market makers), which are elegant but information-inefficient

How proof of history works

• Classic proof of work: block producers solve a puzzle in parallel, race to publish; simultaneous blocks create forks that require resolution — a fundamental delay
• Proof of history forces a slow, single-threaded, recursive hash (SHA-256) that proves time has passed without parallel brute-force — a cryptographic clock
• This creates an arrow of time: validators can prove the sequence of events without coordinating in real time
• Solana pre-schedules validator turns (like TDMA radio — time-division multiple access); each validator gets a slot and if unused, it passes automatically
• Compared to classical BFT systems (e.g., Tendermint): those require explicit acknowledgment after every turn; Solana's schedule eliminates that overhead
• Origin: Yakovenko had the eureka moment at 4am with two coffees and a beer, working from Qualcomm's radio communication background

Solana's architecture and performance

• Current benchmark: ~50,000 TPS with peaks over 65,000; theoretical hardware limit at 1 Gbps bandwidth is ~700,000 TPS
• Requires ~1 Gbps input/output per validator — higher bar than running a Bitcoin node, by design
• As hardware improves (more CPU cores, GPU lanes, faster PCI buses, cheaper bandwidth), throughput ceiling rises automatically
• Parallelizable transactions (e.g., 700,000 independent markets each doing 1 event/second) can all be scheduled concurrently over a single shared state table
• Serum, a central limit order book DEX on Solana, does ~20 million transactions/day vs. Uniswap's ~100–200k
• For comparison: Visa handles ~5,000 TPS at steady state; NASDAQ does ~200–300 trades/second (with 30x more orders and cancels from bots)

Settlement vs. execution layers

• Settlement layer (Bitcoin): slow, final, thermodynamically secure; like an international wire — limited throughput is acceptable at that level
• Execution layer (exchanges): price discovery, order books, clearing — physics-limited by how fast information can be gathered, compressed into a price, and propagated
• Ethereum sits in the middle: proof-of-work security with execution capability, but too slow for real markets
• Solana targets the execution layer: a Byzantine fault-tolerant messaging network guaranteeing fair, censorship-resistant access — the same guarantee NASDAQ gives co-located traders via equal-length ethernet cables
• Proof of stake replaces thermodynamic energy with token-weighted keys; Solana adopts this and optimises purely for speed, accepting "weak subjectivity" (trust on first use, like TLS certificates) in exchange for performance

Censorship resistance and the SOL token

• Goal: maximise the Nakamoto coefficient — the minimum number of independent validators that would need to collude to corrupt the network
• To grow this set to tens of thousands, the network must handle large message volumes, more cryptography, more bandwidth — requiring better hardware
• SOL token represents voting weight in the network; freely transferable so no central authority can control validator issuance
• Solana's "religion": maximise censorship resistance; not competing for the sovereign-money or store-of-value narrative (Bitcoin's domain)
• The network is framed as a guaranteed peer-to-peer messaging layer — like a neutrino laser that can't be jammed — where the token primarily prevents spam

Building the ecosystem: validators first

• Project started late 2017/early 2018, immediately hit the crypto winter; committed funds from crypto-native VCs evaporated
• Survived with early backers: Foundation Capital, Slow Ventures, MultiCoin; validators (node operators) came in each round without hesitation
• Early validator community (~40 people) ran through repeated mainnet crashes (first launch crashed in 20 minutes, second in two hours) and kept returning
• Validators found critical security bugs and some went on to found security firms — deep early engagement created compounding value
• Team recruited from Qualcomm: principal engineers and senior directors with 10+ years working on constrained hardware, drawn by the technical challenge
• FTX partnership originated from a live demo ("Break Solana") where Yakovenko showed Sam Bankman-Fried real smart contract transactions confirming on screen — SBF immediately mobilised his engineering team

Use cases enabled by high throughput

• Serum: central limit order book DEX, NASDAQ-style, running on Solana — 20M transactions/day, tighter price discovery than AMMs
• Audius: decentralised music platform, 4M users; artists connect directly to fans via cryptographic keys, paid directly per stream, no intermediary taking a cut
• General case: eliminates rent-extraction at every layer — Visa's ~5% cut, NASDAQ's ~$10B/year, ad-network fraud (~60% of spend)
• Tokens on Solana can represent anything: equities, stablecoins (USDC), any asset — the open permissionless network is the execution layer regardless of what is being traded
• High-frequency machines (bots that synthesise world information into prices) and human-facing products can coexist on the same network with no intermediaries between them

Ethereum vs. Solana

• Neither can "kill" the other — open source communities can't be shut down; Linux vs. Windows analogy applies
• Overlapping features mean healthy competition; teams socialise at conferences, share ideas
• ETH2 (sharding + proof of stake) is a theoretical roadmap; Solana is live and shipping
• Ethereum's constraints drove real innovation (AMMs, DeFi); Solana's speed enables a different class of applications (full order books, high-frequency settlement)
• Developer pitch: build on Solana if the vision is a billion directly connected users with no shards, no layer-twos, no intermediaries — "we'll eat glass to make it happen"

Long-term vision

• Self-custody is the gateway to the truly connected peer-to-peer world; estimated ~30M people self-custody today
• Scale that to 30–40M active participants and DeFi Summer-style coordination events become global phenomena happening in hours
• The layer-one technical debate (Solana vs. ETH2 vs. others) is like Intel vs. AMD — important to engineers, irrelevant to end users
• Success looks like thousands of independent engineers and validators building grassroots, not a Google-style centralised organisation
• Ultimate framing: a super-connected world where any two parties can always synchronise, always have fair access, no one can get ahead — a financial system without the gunk