Protocols & Policy · Technology preview
SDF's Oct. 28 Foundation News post frames a new L1 execution model that removes the single-global-order bottleneck so capacity can grow by adding machines, with early single-validator benchmarks of 1.18 million payments per second on one machine to 7.02 million on eight (nearly 6×), and says the preview is live on FutureNet, not as a mainnet capacity claim.
The Stellar Development Foundation on October 28 published a technology preview of Project Thruput, a new Layer 1 execution model it says removes the single-global-order bottleneck so capacity can grow by adding machines [1]. In early company benchmarks labeled as single-validator execution, capacity rose from 1.18 million payments per second on one machine to 7.02 million payments per second on eight machines, nearly 6× the original capacity [1]. The preview is live on FutureNet today; those figures are not presented as mainnet production capacity or as end-to-end network throughput [1].
In plain terms, Project Thruput is an SDF technology preview of how a validator can execute many payments in parallel against shared state, including when those payments compete for the same accounts [1]. The mainstream business read is payment-rail infrastructure: whether more hardware can raise execution capacity without sharding the chain. Same-day Meridian conference chrome (Oct. 28–29, Lisbon) is label only [1][2]. This piece is distinct from the same-day Securitize×Stellar tokenization KEEP and from soft Archax Meridian color.
Most blockchains process transactions in one strict global order [1]. Some designs run unrelated work in parallel, but when two transactions touch the same state they fall back into line, and extra machines sit idle [1]. Contention remains the bottleneck when many payments compete to update one balance or pool, because designs often treat touching the same state as depending on it [1].
SDF says Project Thruput changes that model. Transactions in a block start from the same ledger snapshot and can run side by side even when they update the same state; compatible changes combine when the block commits [1]. For a balance, a transaction can record a delta such as “subtract 5” rather than a final absolute, so many payments can update the same account in one block while still respecting the balance’s limits [1]. Each validator can spread that execution across many machines that work as one against shared state, adding capacity without splitting the blockchain into shards [1].
Under heavy contention, with every payment moving among just 100 accounts, SDF reports the 1.18 million to 7.02 million payments-per-second path as machines were added [1]. Those benchmarks measure real RV32 ERC-20 transfers and include execution, signature checks, and block open-to-commit work in Thruput’s core execution engine [1]. They do not measure end-to-end network throughput [1]. Each figure is the median of three runs on AWS r8a.16xl instances; the blog points exact results to the GitHub repository thruputchain/whitney and does not invite invented intermediate machine counts or extra pps figures into this story [1].
@StellarOrg posted around 11:30 a.m. ET the same day calling Project Thruput a technology preview of a new blockchain execution model and quoting SDF CEO and Executive Director Denelle Dixon; that company post links outward to The Defiant coverage and is treated here as X lead only, not a second numeric primary [2]. The numeric claims stay on the Foundation News blog [1].
SDF also describes a fee posture for the preview: no per-transaction fees; gas still measures and caps each transaction’s execution; when demand exceeds capacity, an account’s scheduling share is set by its unlocked native-token balance without spending those tokens or competing through priority fees [1]. That is company framing for Project Thruput as bloged, not a mainnet fee-roadmap claim.
The objection is mechanical. Early single-validator execution numbers on a technology preview venue are not a claim about today’s public Stellar mainnet capacity [1]. The contention setup is deliberately narrow (100 accounts), and the company itself says the figures do not measure end-to-end network throughput [1]. Aspirational mission language about serving ten billion people and making free smart-contract execution practical at scale is company framing in the same post; it is not a verified outcome [1].
Le Coin’s view is that the ops-relevant tell is whether capacity scales by adding machines under contended shared state without sharding [1]. What would change the view is reproducible third-party verification of the whitney results under the same caveats, or a later company filing that labels mainnet or end-to-end network throughput with numbers this Oct. 28 post does not state [1]. Until then, the checkable public record is the October 28 Foundation News post and the FutureNet technology preview as labeled [1].
Holding the Corner.
Not financial advice.