The cryptocurrency industry has long treated Visa's 24,000 transactions per second (TPS) as the holy grail of scalability. New blockchains boast of 'beating Visa' speed, and Layer 2 solutions use it as a benchmark. However, this figure is largely a myth. Visa's actual network processes roughly 1,700 TPS on average, with peaks occasionally reaching 2,000+. The 24,000 TPS figure originates from internal stress tests under ideal conditions—far from the messy reality of global payment networks.
Where Did the 24k TPS Myth Come From?
When Bitcoin launched, early adopters realized that 7 TPS would never suffice for global adoption. Visa's reported '24,000 TPS' was adopted as a target, but the number stuck without critical examination. In reality, Visa's own documentation states its network handles about 1,700 TPS. The inflated figure serves marketing purposes, much like the exaggerated TPS claims from emerging blockchains.
Blockchain TPS Overpromises Are Common
New projects regularly claim tens of thousands of TPS, but these metrics usually come from controlled tests on a handful of cloud servers. Once deployed on a decentralized global mainnet, performance plummets. Blockchains like EOS and NEO achieve high throughput by limiting validator nodes—sacrificing decentralization for speed. Such systems cannot offer the censorship resistance that makes Bitcoin valuable.
The Inevitable Trade-off: Speed vs. Decentralization
There is a fundamental trade-off in blockchain design: increasing throughput typically requires reducing node count or raising hardware requirements, making the network more susceptible to attacks or regulatory capture. Bitcoin's Lightning Network and Ethereum's sharding are attempts to scale without centralization. Any project claiming to surpass Visa's 24,000 TPS should be scrutinized for its architectural compromises—do they rely on a coordinator (like IOTA), legal threats (like Hashgraph), or a small set of validators?
Conclusion: Be Skeptical, Read the Fine Print
When evaluating blockchain performance, always ask: Are the test results from a public mainnet? What is the node distribution? Does the system depend on a coordinator or permissioned validators? The real benchmark for global payment systems should be Visa's actual ~1,700 TPS, not the mythical 24,000. Only by grounding claims in reality can the industry build truly scalable and decentralized solutions.