Solana's First 48 Hours With 100M CU Blocks: P90 Fees Fall 30%, Congestion Collapses

Two days after Solana raised its per-block compute ceiling to 100 million compute units, on-chain data shows the upgrade has materially changed how blocks fill. Congestion at the old limit has nearly disappeared, and the competitive bidding that pushed transaction fees into the tail has eased measurably.

SIMD-0286 activated at Epoch 1009 on July 29, raising Solana's maximum block compute budget from 60M to 100M CUs, a 66% increase in per-block capacity. The first two days of on-chain data now offer a specific read on what that headroom change has actually produced.

SIMD-0286 48-Hour Data: Block Cap Congestion Falls from 12.5% to 0.5%

The headline shift is in how often blocks were hitting the limit. Before the upgrade, 12.5% of Solana blocks were crowding the effective 60M CU cap, sitting within roughly 5% of the ceiling. As of July 31, just 0.5% of blocks approach the new 100M CU limit.

The other side of that figure: 10 to 23% of blocks now exceed what was previously Solana's hard cap in any given hour, per the Pine Analytics data. That shows the additional headroom is already being used across normal network activity.

P90 Transaction Fees Down 30%

The fee impact is where users feel the difference most directly. At the 90th percentile, transaction fees fell from 29,800 lamports to 20,800 lamports since activation, a 30% reduction, per Pine Analytics.

The median transaction still pays around 5,600 lamports, the base fee, unchanged from before the upgrade. Priority fee competition concentrated in the tail, where users bidding for execution in the most congested blocks were pushing each other's costs higher. With more compute headroom per block, that pressure has loosened.

Solana Compass fee data tracks the same trajectory across the week of activation. P90 fees ran in the high 20,000s lamports from July 25 through July 28, then stepped down to the low 20,000s on July 29 and held there through July 30:

Median Block CU Load Stays at 30M: Where Pre-Upgrade Congestion Actually Lived

The median block sitting flat at roughly 30M CUs, and the median fee staying at base level, points to something specific about where the pre-upgrade congestion was located. The old 60M CU cap was creating a bottleneck for the minority of blocks that were genuinely full, at 12.5% of total blocks. That was enough to force priority-fee competition at busy times without making the average block look saturated.

SIMD-0286 addressed exactly that constraint: the 60M cap created predictable ceiling encounters at demand peaks, not uniform network saturation. Widening the ceiling cleared those moments without changing the median block load, which is a cleaner outcome than an upgrade that simply spreads the same transactions across more space.

Developers should note what SIMD-0286 did not change: the per-account compute limit of 12M CUs stays in place. Programs bumping into account-level execution limits won't see relief from this upgrade. Block-level expansion and the per-account cap are independent constraints.

What the Utilization Figures Mean for Validators and Capacity

The 10–23% of blocks now exceeding the old 60M CU cap in any given hour is the key figure for gauging how quickly the new headroom has been put to use. Actual transaction activity is already filling the space; demand was not waiting on capacity.

For validators, the transition has been smooth. No consensus stalls were reported through activation, confirming the validator network handled the larger blocks without sacrificing block timing.

The current picture leaves substantial distance between current utilization and the new ceiling. The 0.5% cap-crowding figure at 100M CUs means the network has absorbed the expanded headroom without immediately pressing against it. Whether that distance narrows depends on how transaction volume grows relative to the available space over the coming weeks. Two days of data is a signal, not a trend, but on the evidence from the first 48 hours, the upgrade has done what it set out to do.