I Built a 30,000-TPS Chain, So I’ll Say It: TPS Is for Investors, Finality Is for Payees

If you work in payments, acquiring, or cross-border clearing and you’re choosing which chain to build on—don’t ask “how many TPS does this chain do.” That’s a number the marketing department carefully operates. What actually decides whether it can carry your business is two things nobody likes to talk about: the real transaction volume after you strip the water out, and whether a payment counts.

I write this chapter from the chain-builder’s chair. I built a software-hardware integrated public chain. To push performance up, the team built a dedicated hardware acceleration engine—moving the heaviest work, signature verification, off software and onto hardware in parallel, hitting roughly 30,000 TPS, with a team of over a hundred. And then? We invested enormously, and eventually faded out.

Precisely because I pushed that number up with my own hands, I know better than most: TPS is an overrated number. This isn’t sour grapes—it’s that I personally hit the lab peak, so I know how many of the steps I tripped over sit between a lab peak and actually clearing an invoice with it. Back then we could say “~30,000 TPS signature verification,” but that was the peak of the signature-verification segment—not end-to-end throughput including state writes, network broadcast, and final confirmation. Marketing wanted the big number; the engineers knew where the denominator was.

Most people think whether a chain can do payments comes down to its headline TPS. In fact, headline TPS is a number for the marketing department, not the payments department.

The Framework: Adjusted Transaction Volume

Not every transaction that ran on-chain is real economic activity. Subtract bot wash trades, exchange-internal consolidation, market-maker hand-to-hand churn, and airdrop sybil addresses from the total (unadjusted), and what’s left is adjusted volume—the transactions someone actually made for a real purpose, with real money.

One picture says it:

A big pool (on-chain total) → three filters (bots / internal transfers / wash trades) → the small puddle that drips out the bottom is the water your business can use.

TPS measures how big the pool is; payments care how much drips out. Two different things, merged by marketing into one.

The TPS Myth: Four Switched Concepts

Pull “performance” apart and it hides at least four muddled accounts, merged at marketing time into one scary number:

1. Peak vs. sustainable. The instantaneous peak in an ideal lab, vs. sustainable throughput under 7×24 load, state bloat, and uneven nodes—two different things. One goes on the slide; the other carries the business.
2. Signature verification vs. end-to-end. I know this one best—fast signature verification doesn’t mean the whole path is fast, from submission to state write to network broadcast to finality. We accelerated the heaviest segment with hardware, but the other staves of the barrel were still there.
3. Throughput vs. finality. What payments truly fear isn’t “slow,” it’s “does this actually count.” Reversibility matters far more than transactions-per-second. A slow authorization, nobody returns the goods; a payment reversed by a chain reorg is an incident.
4. Capacity vs. economic activity. A chain can process N transactions; that doesn’t mean N units of real demand exist. Capacity is not usage.

A common comparison, punctured: the oft-cited “Visa, 24,000-plus TPS” is itself a network capacity ceiling, not a daily average—by annual transaction count, Visa’s daily throughput is actually on the order of under ten thousand per second. Pitting a chain’s lab peak against Visa’s peak capacity is two watered-down numbers propping each other up.

Drain the Pool and Look

This is the framework’s main stage. On-chain “total volume” holds three kinds of water: bot/sybil wash trades (for airdrops, rankings, manufacturing “ecosystem vibrancy”); exchange and project internal transfers (consolidation, hot/cold wallet moves, bridge internal bookkeeping—money didn’t change owners); and market-maker wash trades (hand to hand, to inflate “volume” for valuation or leaderboards).

Lay those three filters, and the total often shrinks to a fraction—that’s the real economic activity. For payment people, the damage is direct: when due diligence shows you “this chain does hundreds of millions of transactions a day, far past Visa,” and you don’t adjust, you’re doing technical selection on a number that was wash-traded into existence.

So the real due-diligence question isn’t “how many TPS”—it’s three: After stripping internal transfers, how many real payments? Is finality probabilistic or deterministic? Under congestion, what’s the tail (P99) of settlement fees and confirmation time?

So Can Stablecoin Payments Hold Up? A Layered Answer

Don’t answer in the abstract. By scenario, the answer differs:

• Small, instant, probabilistic-finality-tolerant retail payments: parts of high-throughput chains plus L2s are already usable. The pain isn’t peak TPS, it’s the fee tail and experience consistency under congestion.
• Large B2B / cross-border clearing: what’s needed is deterministic finality and predictable settlement time, not transactions-per-second. This is exactly where stablecoins have a structural edge over the correspondent-banking system (multi-day, multi-intermediary)—the value isn’t “fast as tens of thousands per second,” it’s “7×24, irreversibly settled in minutes.”
• Full replacement of Visa-scale retail: short-term unrealistic, and unnecessary.

The judgment collapses to one line: a stablecoin’s payment value comes not from “faster than Visa,” but from “replacing the slow, expensive, fragmented cross-border clearing pipes behind Visa.” The race it needs to win isn’t the TPS arena—it’s the infrastructure track.

If you work in payments/acquiring/cross-border, landing it is four points: set metrics by scenario, not by marketing; write “finality” into risk control, not “TPS” into the deck; demand adjusted volume in due diligence, and if they can’t give it, treat it as nonexistent; remember performance is the entry ticket, distribution is the moat.

TPS is a number for investors; finality is a number for payees. Payments never ask how fast you run, only whether this one counts.

The chain you’re selecting—can they give you the real payment count after the water’s drained out?

—— From The Stablecoin Operating Manual: Distribution, Reserves, and Compliance, Chapter 14

Sources (verified):

• Author’s firsthand account: HPB ~30,000 TPS signature verification, in-house hardware acceleration, software-hardware integrated chain (around 2018).
• Visa: “24,000-plus TPS” is a network capacity ceiling, not a daily average; by annual transaction count, daily throughput is on the order of under 10,000/sec.
• Adjusted volume methodology: Visa Onchain Analytics, Allium et al. provide adjusted volume/count de-noising (de-watering ratio varies by chain/coin/time). https://visaonchainanalytics.com/