Solana 2027 Roadmap Targets Double Compute Capacity
Anza's 2027 Solana roadmap targets double compute capacity, onchain vector processing, Rotor block propagation and clearer transaction ordering.

Solana core developer Anza has outlined a 2027 technical roadmap centered on higher compute capacity, more expressive transactions, changes to block propagation and protocol simplification. Anza CEO Brennan Watt presented the plan at Breakpoint 2026, with the official Solana roadmap summary grouping the work into performance, expressivity, market structure and resilience. The roadmap describes engineering targets for 2027 rather than features already active across Solana.
Performance is one of the largest components. Anza wants to double network compute capacity from roughly 250 million to about 500 million compute units per second while reducing the compute consumed by token-account operations by around 80%. Doubling compute capacity would expand the amount of computational work the network can process, but it should not be interpreted as an automatic doubling of transaction throughput.
After Alpenglow, @bw_solana laid out @anza_xyz roadmap for Solana in 2027
Performance: network compute capacity doubles, transactions cut 80% CU units for token accounts
Expressivity: vector processing for models running fully onchain, 64 → 96 accounts per transaction, read…
— Solana (@solana) November 19, 2026
Transactions Gain More Compute and Account Access
The expressivity roadmap includes increasing the number of accounts available to a v1 transaction from 64 to 96. That change is already tracked by Solana as an in-development proposal, while the network previously activated Transaction V1 with a 4,096-byte maximum size. The higher account limit would let a single atomic transaction interact with more programs, liquidity pools and other accounts without changing the 64-account limit for legacy and v0 transactions.
Another planned optimization would let programs read account data directly instead of requiring validators to copy it through several existing paths. Solana currently lists Account Data Direct Mapping as pending feature activation rather than live functionality. The roadmap therefore combines changes at different development stages, and individual features will still require their own releases and activation processes.
Anza is also targeting native vector processing for models designed to run fully onchain. The capability could support computational workloads that rely heavily on vector operations, but the published roadmap does not establish performance benchmarks for large AI inference models or specify which architectures could realistically execute entirely inside Solana’s runtime. Vector-processing support expands the potential computational scope of the network, but it does not mean general-purpose AI inference has already moved onchain.
That distinction differs from existing decentralized compute systems such as io.net, where AI inference runs on distributed GPU infrastructure while Solana handles coordination and economic functions. Blockchain execution and externally coordinated GPU computation remain technically different models for running AI workloads.
Rotor and Constellation Target Network Market Structure
The market-structure component builds on Alpenglow, whose Votor consensus system replaced TowerBFT on Solana mainnet in October and targets roughly 150-millisecond finality. Rotor, the second major Alpenglow component, remains planned for a later release and would replace Turbine’s tree-based block propagation with a flatter relay architecture intended to reduce network latency. Alpenglow’s voting and finality system is live, while its Rotor propagation layer is not.
Constellation addresses a different part of block production. Anza describes it as a mechanism intended to provide clearer transaction-ordering rules and greater competition among validators participating in transaction inclusion. Rotor targets how blocks move through the network, while Constellation targets how transactions are ordered and proposed.
The resilience pillar is broader, calling for legacy abstractions, special cases and older interfaces to be reduced so future protocol changes can be introduced with less complexity. The roadmap follows Solana’s recent transition from the mainnet-beta label to Mainnet 1.0, a protocol-versioning milestone separate from Alpenglow’s October activation. None of the 2027 roadmap targets should be treated as automatically deployed by Mainnet 1.0, because each technical change retains its own development and activation path.
This article is for information only and is not investment advice. We report under our Editorial Policy; to flag an error, see our Corrections Policy.


