Zcash NU7: 25-Second Blocks, a Five-Year Fee Lock, and the Last Call for Sprout

A merchant in Mexico City quotes a price, then waits for three confirmations while the peso drifts. For most of Zcash's decade-long life, that wait has been roughly four minutes — an eternity at a counter, a rounding error on a settlement desk. NU7 proposes to cut it to about seventy-five seconds. That is the headline, and it is the least interesting thing in the upgrade. The same package converts sixty percent of every block's transaction fees into a claim that does not mature until roughly February 2031, and switches off Sprout — the network's original shielded pool — on a deadline, whether or not anyone is still standing inside it. One change is about speed. The other two are about who carries the cost of privacy, and for how long. Follow the money, not the noise. The money here says Zcash is being restructured as a long-duration asset while it is still priced as a short-duration narrative. Some background, because the three changes only make sense together. Zcash launched in 2016 as the first production system to ship zk-SNARKs at the protocol level, letting users prove a transaction valid without revealing sender, receiver, or amount. Its shielded architecture has been rebuilt twice: Sprout was the original pool, Sapling replaced it in 2018 with far cheaper proving, and Orchard followed in 2022 with a new curve and unified addresses. Each generation left the previous pool alive but increasingly vestigial. Block interval has been 75 seconds since genesis — a deliberate compromise. Shielded transactions are heavy: a spend carrying a zk-SNARK proof is orders of magnitude larger than a transparent input, and verification costs real CPU. Seventy-five seconds gave the network room to propagate and validate without routinely forking. NU7 compresses that to 25 seconds. Three confirmations — the informal merchant standard — fall from roughly 225 seconds to about 75. The upgrade has been running on testnet and is awaiting final integration into node software, so the parameters are settled even if the release engineering is not. Then the two riders. Sixty percent of the fees collected in each block leaves circulation at the protocol level, with those tokens scheduled to flow back to miners starting around February 2031. And Sprout is shut down outright: any funds still in that pool must be moved before activation or become permanently inaccessible. Speed, deferred compensation, forced migration. Read those as a set and NU7 stops looking like a performance patch. Start with latency, because it is the easiest to misread. Cutting block time by a factor of three does not merely make confirmations faster; it changes the shape of the wait. At 75 seconds, arrival times cluster loosely and a merchant quoting three confirmations builds a wide margin into the price. At 25 seconds the distribution tightens. For anyone running a point-of-sale terminal, that is the difference between accepting crypto as a marketing line and accepting it as an operational fact. I spent much of 2020 inside this problem. My fifty-page report that year traced how unstable stablecoin pegs propagated into cross-border remittance corridors in Latin America, and the finding that survived contact with reality was unglamorous: merchants do not price crypto risk as volatility, they price it as settlement uncertainty. A corridor settling in four minutes carries a premium that one settling in seventy-five seconds does not. Latency is a tax line. It just never appears on the receipt. So the speed change is real. But it collides with a constraint the announcement does not dwell on. Zcash is not a transparent chain with a privacy layer bolted on. Shielded throughput is the product. Tripling block frequency triples the rate at which the network must propagate and verify proof-carrying transactions. Propagation windows comfortable at 75 seconds get tight at 25. The question is not whether nodes can produce blocks that fast — they can — but whether the slowest meaningful node can receive and validate one before the next arrives. Get that wrong and you do not get faster settlement; you get more frequent shallow reorgs, which is a worse outcome for exactly the merchants the upgrade is meant to win. Now the fee mechanism, the part I would put in front of any allocator. Sixty percent of per-block fees, removed from circulation, returned to miners around February 2031. Strip the language and you have a five-year zero-coupon claim issued by the protocol and held by the people who secure it. Miners are asked to forgo most of their fee revenue today for a claim maturing in 2031. Follow the money and ask who is underwriting whom. In a bull market that trade is easy to sell: fee revenue is small relative to the block subsidy, so the sacrifice looks like rounding. That is precisely what makes it interesting. It is immaterial today and a template tomorrow. If the subsidy keeps halving and fees become the security budget, a deferred-fee structure becomes the difference between a miner who stays online through a drawdown and one who does not. Volatility is the tax on impatience — and a five-year lock is a machine for manufacturing patience, whether or not the holders consented to it. The Sprout shutdown is where the upgrade stops being technical. Sprout is the 2016 pool, deprecated in spirit for years, its proving costs absurd by modern standards. But deprecation in spirit and deprecation in code are different events, and only one is reversible. On activation, any balance still in Sprout is not frozen and not recoverable — from the perspective of whoever holds the key, it is simply gone. Who gets left behind in a forced migration? Almost never the sophisticated user, who reads upgrade notices and runs a node. It is the person who set up a shielded address in 2017, wrote the seed phrase on paper, and put it somewhere safe. The people privacy infrastructure claims to serve — dissidents, journalists, remittance recipients in capital-controlled economies — are structurally the least likely to be watching a network upgrade calendar. I audited contracts for seven utility tokens during the 2017 ICO cycle, and the pattern has held for nine years: governance failures almost never announce themselves as governance failures. They surface later, as unclaimed balances. A pool declared dead by a small group of protocol engineers is a governance decision wearing an engineering costume. There was no vote that mattered, because there never is; turnout was never going to clear five percent, and the outcome was never going to depend on it. The consensus read writes itself: Zcash gets faster, therefore usable, therefore adopted. I do not buy the middle term. Privacy is not a latency problem. The person who most needs a shielded transaction is not choosing between 75 seconds and 25; they are choosing between custodial and non-custodial, and the friction that decides it lives at the on-ramp — liquidity depth, delisting risk, the willingness of an over-the-counter desk to touch a shielded deposit at all. Shave fifty seconds off confirmation and you have improved a metric that was never the binding constraint. The decoupling thesis in one line: NU7 is a supply-side improvement to a demand-constrained asset. It makes the chain better at something the market has not yet decided it wants to pay for. That is not a criticism — infrastructure work usually precedes demand — but it means the upgrade tells you little about ZEC's price over two quarters and a great deal about its architecture over five years. The fee lock cuts the other way. Deferring sixty percent of fees to 2031 shrinks the float today and concentrates a claim in miners' hands tomorrow — a supply story dressed as a security story, which is the most durable kind of supply story. Three numbers to watch. The Sprout migration count as activation approaches — the only honest measure of whether privacy infrastructure respects the people it was built for. The February 2031 unlock, the first real test of whether deferred miner compensation behaves like a bond or like a promise. And whether 25-second blocks survive contact with shielded transaction bandwidth once mainnet traffic, not testnet traffic, sets the pace. Speed is a promise. Finality is a contract. NU7 delivers the first and defers the second — and the gap between them is where users get hurt.