Solana increased the maximum computation per block from 60 million to 100 million compute units on July 29, 2025 — a 66% jump, according to the Solana Foundation. The upgrade, known as SIMD-0286, only touched the overall block limit. It left the per-account compute unit cap at 12 million CUs and the maximum block accounts-data size delta unchanged at 100 MB. That means hot accounts — those with shared writable state — still face the same bottleneck they did before.
What the upgrade actually changes
Before the increase, 11.2% of blocks produced between July 22, 2025 (when the 60M-CU limit was activated) and the latest increase used at least 56 million CUs. The new ceiling gives validators more headroom to pack transactions into a single block, at least in theory. One block on August 2 recorded 25,373,012 consumed CUs, 1,328 processed transactions, and 1,069 successful transactions. But the Solana Foundation hasn't provided any before-and-after measurements for transaction inclusion, fees, block propagation, or replay in official upgrade materials. A single block can't establish network-wide outcomes.
What stays the same — and why it matters
The per-account compute unit cap remains 12 million CUs. The maximum block accounts-data size delta stays at 100 MB. So any transaction that competes for the same hot state — a popular NFT marketplace, a DeFi pool, a memecoin launch — still hits the same per-account ceiling. The higher block ceiling doesn't reduce priority fees for those transactions. If you're trying to write to a hot account, you're still bidding against everyone else for a slice of that 12 million CU pie.
Risks the technical proposal flagged
The technical proposal for SIMD-0286 warns that larger blocks can take longer to execute, slow network progress and validator catch-up, and may cause unforeseen issues for infrastructure beyond validators. No post-activation measurements have been published to show whether those risks have materialized. The Solana Foundation hasn't said when or if it plans to release such data.
The question now is whether the 66% increase in block compute will actually improve throughput — or just shift the bottleneck from the block level to the account level. Without real-world metrics, the answer remains unclear.
