Grigore Roșu, Founder & CEO of Pi Squared, the team building Web3 infrastructure beyond blockchain. Grigore is a professor of computer science at the University of Illinois and a former NASA engineer. He also founded Runtime Verification, now the leading formal verification company in Web3. In 2023, he founded Pi Squared with a vision to solve fragmentation and trust limitations in today’s blockchain infrastructure.
In today’s rapidly evolving blockchain space, TPS, or transactions per second, is often dismissed as somewhat of a shallow metric, but it’s still one of the clearest signs a blockchain is ready for prime time. Across academia and industry, Grigore heard all the counterarguments framing TPS as a vanity metric that doesn’t actually prove decentralization. However, try running a payments app, a high-frequency trading desk, or an AI-powered gaming world on a network that tops out at 20 TPS and then come back to him and say it’s irrelevant.
The thing is, the “TPS doesn’t matter anymore” narrative is missing the bigger picture. If blockchains are ever going to handle programmatic payments, real-time stablecoin settlement, AI-driven automation, tokenized assets moving between chains, they need raw throughput. Not just in bursts, but sustained, reliable, predictable throughput. Without it, none of this stuff works at scale.
Recent moves in the market strengthen my point. When PayPal teamed up with Solana for PYUSD, the buzz wasn’t about the stablecoin’s logo or whitepaper, it was more about speed. Same story with Visa testing USDC settlements on high-throughput chains. Big institutions and consumer-facing apps want performance that feels like traditional infrastructure. For them, sub-second latency isn’t a “nice to have,” it’s an absolute pre-requisite for success.
Why TPS Still Works As a Signal
Grigore will admit, there are some metric limitations in that TPS can be misleading if you strip it from context, ie: you can inflate it with internal transfers, spam transactions, or simplified workloads. And while it says nothing about finality times, censorship resistance, or validator diversity, it’s still a simple, relatable proxy for whether a chain or protocol can handle scale.
Some people will still argue that focusing on TPS encourages centralization because you end up with validator setups that look like enterprise data centers. Others say modular systems with fancy data availability layers can scale just fine without pushing raw transaction limits. While those are fair takes, they don’t change the reality in production today.
Users aren’t waiting for the perfect balance between theory and practice; they’re just picking the chain that works for what they need. But what if you have high TPS without compromising decentralization and verifiability, but only because you use a better, next-generation protocol?
Real-world Pressure on Throughput
The minute you step into real applications, the need for throughput becomes painfully obvious. Think AI agents running automated trades across three chains at once, or stablecoins being streamed in micro-payments by the second, or tokenized assets moving back and forth between institutions without settlement delays. None of that is happy running on a network that can only breathe when it’s quiet.
We’re already seeing this pressure show up in the market. Base and Blast are building infrastructure for social apps that could onboard millions overnight, while zkSync and Scroll are aiming to process zk-proofs for entire financial systems. And perhaps not every chain needs that kind of speed, but the ones that don’t will eventually find themselves boxed out of the most valuable markets.
FastSet: A Different Take on Scaling
This is where Grigore thinks the FastSet protocol changes the conversation. The approach is actor-based, meaning they can process stuff in parallel across the network instead of forcing every single node to touch every single transaction in the same order. That’s how they just blew past 100,000 TPS with sub-100ms finality, on regular, $1,000-or-less hardware, not enterprise-grade, liquid-cooled monster rigs.
Just 24-core CPUs with 256 GB of memory, which is basically a decent workstation. Compare that to the specialized validator setups you see in some high-throughput L1s and L2s, and you start to see why this matters. If you lower the cost of participation, you get broader decentralization, not just speed.
Why the Hardware Thing Matters More Than You Think
A lot of people get caught up in the numbers: 100,000 TPS now, aiming for 1 million TPS at mainnet, but the real trick here is accessibility. If a protocol can hit those speeds without pricing out small validators, it keeps the door open for a healthier, more distributed network.
It also shifts how you think about scaling. In the FastSet model, if you need more TPS, you just add validators. Horizontal scaling, but without the messy trust trade-offs that usually come with bridging or central coordination. That’s a lot closer to how you want this stuff to work when AI agents and institutions are all hammering the same infrastructure at once.
The AI + Web3 Collision Is Coming Fast
AI is going to eat a huge chunk of blockchain traffic in the next few years. Autonomous agents making trades, negotiating smart contracts, even buying services in micro-transactions, it’s all going to need low-latency, high-throughput settlement. If you can’t offer that, you’re not even in the conversation.
Looking ahead, DeFi apps that rely on rapid settlement will want the speed, tokenization platforms will want the verifiability, and AI systems will want the parallel execution model. And the fact that it’s affordable to run a validator means you don’t have to beg the same five players for permission to join.
For enterprises, it’s even simpler. If you’re a payments provider or an RWA trading desk, you want predictable, instant settlement that doesn’t melt down in a traffic spike. That’s why, despite all the criticism, TPS will remain one of the first metrics decision-makers look at. It’s not perfect, but in the real world, where apps need to scale, it’s still one of the quickest ways to tell if a network might survive serious demand.
The post Yes, TPS Still Matters. Debunking Myths Around the Metric’s Supposed Downfall appeared first on Metaverse Post.
