Patrick WoodheadWhen Filecoin first launched, the network introduced a breakthrough cryptographic proof for storage: the Proof-of-Replication (PoRep). This mechanism allowed a storage provider (SP) to verifiably show they were storing a client’s data on a unique hard drive replica. It was one of the pioneering successes in the emerging “DePIN” (decentralised physical infrastructure network) model, long before the term became widely used.
Thanks to that design, Filecoin quickly ramped up storage capacity at a scale that was once unimaginable. The network became the largest decentralised storage system in the world, supporting exbibytes of capacity.
However - and this is critical - while Filecoin enabled proof of storage, it did not provide proof of retrieval. In other words: there is (and always has been) no on-chain mechanism by which a storage provider can submit a proof that “I served this data to a client” and thereby earn a block reward for that retrieval.
The Retrieval Gap
The reason for this isn’t a lack of desire. It’s a fundamental cryptographic impossibility. There is a well-known result in the literature often framed as the “impossibility of fair exchange” between two untrusted parties without a trusted third party. Simply put: if SPs could earn on-chain rewards for serving data, they could cheat (for example, by inventing thousands of clients or by not actually serving the data but still submitting the proof, both of which are very hard to distinguish from real data transfer between two parties). Consequently, there will never be a robust trustless proof of data retrieval in the protocol.
This absence of retrieval incentives meant that, although data was stored under the PoRep regime, many storage providers simply did not bother to ensure it was readily retrievable. Data thus gravitated toward being archival in nature, with limited guarantees around service and egress. The PoRep design itself also required data to be cryptographically sealed, meaning it would take hours to “unseal” before retrieval, so providers were asked to keep a “hot” unsealed copy to serve requests. However, the compliance to do so was very low, as there was no incentive to do so.
Early Retrieval-Incentive Attempts: Filecoin Spark
To address this challenge, we developed the protocol known as Filecoin Spark: a system that randomly selected files stored on Filecoin and asked a committee of “Spark Checkers” to fetch them from storage providers. Based on those checks, we could derive a retrievability metric for the Filecoin network, and for individual providers.
Spark yielded meaningful progress: it drove the network retrieval success rate from ~1% up to a peak around 26%. It also helped identify the subset of 20-30 storage providers delivering strong retrieval performance. However, it also revealed limitations:
- Providers learned to distinguish Spark-checker traffic from organic retrieval traffic and thus could game the system (for example by servicing Spark check requests but not client requests).
- Spark couldn’t measure how much data a provider was serving (i.e., egress) over time. Two providers with identical Spark scores could have very different actual retrieval volumes, but Spark couldn’t tell the difference.
Enter the Filecoin Onchain Cloud & Filecoin Beam Era
When we launched Filecoin Onchain Cloud, a new possibility emerged in the network: the ability to validate hot (unsealed) data via a lighter “Proof of Data Possession (PDP)”. With PDP, we could verify that some copy of the data was accessible (though not exactly which SP held it).
In parallel, we built Filecoin Beam: an access / delivery layer between clients and storage providers. Beam is designed to fetch data from SPs and relay it to clients, acting as a trusted measurement point for real client traffic (rather than synthetic probes as was the case with Spark). With Beam we can measure actual real egress from each SP, how many bytes they actually serve to clients via Beam as the delivery layer.
Note: Yes, this architecture does require trust in Beam’s reporting, but recall that a trustless proof of retrieval is impossible in the protocol. Please note that the Beam team does not operate any mainnet storage providers nor partner with SPs, which strengthens its impartiality and credibility.
“Pay Per Byte” — The Incentive Engine
Here’s where it all comes together: We’re proud to introduce Pay Per Byte — the first-ever scheme in Filecoin’s history that pays storage providers for serving data (i.e., retrieval/egress) and charges clients for the bytes they actually consume.
- Storage providers receive 7 USDFC per TiB served to Beam.
- Beam takes 7 USDFC per TiB it serves to clients.
- This means for the client, it costs up to ~0.014 USDFC per TiB of egress, across both cache-hits that get served back straight from Beam and cache-misses that fall back to the SP for delivery.
This is a landmark shift. After years of retrieval incentive research and protocols (Spark, PDP, etc), Pay Per Byte is the first time we’re getting real, measurable economic alignment: SPs are rewarded for data served, and clients pay for actual usage. The result? A more economically sustainable, higher-quality service layer, which in turn opens up a broader range of users and use-cases for Filecoin.
Why It Matters. The Big Picture
Pay Per Byte doesn’t just solve a long-standing incentive gap. It also sets up the Filecoin Onchain Cloud as a core infrastructure component in the next-gen web stack: call it the x402 layer (resource-payments) meeting the infra layer (egress payments).
Think of it in terms of this example: you have an academic journal that charges “2 USDFC for access”. Behind the scenes, the journal has to pay their cloud egress bills covered by the upfront costs to the clients. Historically, that happened in centralised clouds. With Pay Per Byte + x402 we now have in-protocol mechanisms for both resource payments AND infrastructure payments — all in a wallet-centric, verifiable, pay-as-you-go model.
In short: subscriptions are passé. User-owned credentials, wallets, minimal friction, and true pay-as-you-go are the future. Pay Per Byte puts Filecoin Beam (and Filecoin OC) squarely in that paradigm.
Next Steps
- Rolling out Pay Per Byte to initial clients and SPs, and measuring early real-world usage.
- Monitoring SP performance, egress volumes, cache hit-rates and refining the scheme.
- Exploring integrations: x402 payments, ERC8004 and more.
- Communicating to clients: “You pay only for what you use. No monthly subscription fees.”
Conclusion
For years, the storage side of Filecoin has worked: providers store data, submit proofs, and the network scaled. But retrieval remained the missing piece: no on-chain proof, limited incentives, unclear service guarantees. With Filecoin Spark, we made strong progress. Now with Filecoin Beam’s Pay Per Byte we finally align incentives, measurement and usage in a meaningful way.
This isn’t just a feature. It’s a foundational shift in how decentralised storage networks enable active data delivery, not just passive archiving. It positions Filecoin, through the Filecoin Onchain Cloud, to emerge not simply as a “store once” network, but as a “serve many” network, with proper economics, open access, and next-generation infrastructure architecture.
Here’s to the next chapter of decentralized data, served and paid for byte-by-byte.