Solana's slot time fell to 350 milliseconds on August 21, the first cut since the network launched in 2020. Validators approved SIMD-0525 at epoch 1020, and most of the coverage framed it the same way: the chain got faster, with a roadmap toward 200 milliseconds. That reading is half right. The slot time dropped 12.5 percent, and so did the amount of work each block is allowed to carry.
The per-block compute budget went from 100 million compute units to 87.5 million. Both numbers move together by design. A slot that arrives 12.5 percent sooner is handed 12.5 percent less compute, which leaves the network's per-second ceiling exactly where it sat before the vote.
Do the arithmetic. One hundred million units every 400 milliseconds works out to 250 million units per second. Eighty-seven and a half million every 350 milliseconds is also 250 million per second. The confirmation clock ticks faster while the capacity behind it stays put.
| Change | Slot time | Per-block budget | Per-second budget |
|---|---|---|---|
| Before SIMD-0286 | 400 ms | 60M CU | 150M CU/s |
| After SIMD-0286 (July) | 400 ms | 100M CU | 250M CU/s |
| After SIMD-0525 (August) | 350 ms | 87.5M CU | 250M CU/s |
The capacity jump already happened, back in July
Solana did add real capacity recently, just not this week. In late July, SIMD-0286 raised the block limit from 60 million units to 100 million at epoch 1009, a 66 percent increase that lifted the per-second budget from 150 million to 250 million. That was the throughput upgrade. The 350 millisecond slot time is a latency upgrade. Treating the two as one overstates what August actually delivered.
Faster confirmation still counts for something. A swap or a trade gets its inclusion opportunity sooner, and to a user that latency is exactly what the word fast means. It is not the same thing as room for more transactions per second, and that gap is where the headlines blurred. Other chains have chased the same latency prize before, including TON, which dropped its block times to 400 milliseconds with Catchain 2.0.
More slots means more work per validator
Shorter slot times carry a cost the throughput debate tends to skip. Every 350 milliseconds a validator now has to vote, gossip, and process, where before it had a comfortable 400. The wall-clock demand climbs even though the per-second work does not. Anza's own framing concedes the strain. Smaller operators could find the economics stop penciling out and drift toward the larger ones, and that risk lands hard on a network where one provider already held 27 percent of staked SOL against a self-imposed 25 percent ceiling.
This is the second Solana parameter in a season to reach mainnet through the SIMD process, after an earlier proposal aimed at the network's SOL burn. The roadmap runs 350 to 300 to 250 to 200 milliseconds, each step needing its own validator vote. Jacob Creech of the Solana Foundation put it plainly: "We're in a new era of 350ms. Next stop, 300ms." Every stop on that path is set to trim the per-block budget again, and hold the per-second number right where it is now.
