Filecoin's Reward Actor has never asked whether anyone wanted the bytes.
It pays against proof of capacity — PoRep sectors sealed, PoSt proofs submitted on schedule — and that is the whole of its epistemology. Roughly seventy percent of FIL's supply is earmarked for storage provider rewards across a six-year halving schedule layered onto a baseline-minting curve. A provider can seal a terabyte of cryptographic noise, keep the proofs valid, and collect the identical emission as a provider hosting a customer's production archive. The network has known this since mainnet in October 2020. It has mostly shrugged.
Now the team wants to change the payout function. The plan as stated: redirect block rewards toward paid storage. This is not a technical upgrade. It is a rewrite of the emission oracle, and the hard problem is metrology, not cryptography.
One factual claim reached me in the announcement; three others were unattributed opinion. I will keep them separate, because a network-upgrade proposal and a press release share almost no properties — but the direction of the claim names the correct disease, so it is worth examining on its own terms.
Filecoin's structural problem is that raw byte-power and real data are decoupled. Providers are rewarded per sector, and the marginal cost of sealing worthless data is nearly the marginal cost of sealing valuable data: a few tenths of a FIL in ProveCommit gas, plus pledge collateral and hardware. Rational providers therefore optimize for bytes, not customers. Filecoin Plus was the first patch — a notary registry, verified clients, and a ten-times quality multiplier. It shifted where power got sealed. It did not create paying customers, because the multiplier rewards verification, not payment.
The resulting economics are unflattering. Storage fee revenue has historically been a small fraction of provider income; my working figure from protocol-level tracking is under twenty percent, and I would not defend the second digit. The rest is emission. That is a subsidy flywheel — new issuance buys capacity, capacity buys the appearance of adoption, the appearance buys the next narrative. It is not a Ponzi: the capital comes from inflation, not later depositors, and providers post real hardware and real collateral. But it is a flywheel with a false bottom, and it stops when issuance decays. The standard governance path for a fix is a FIP — off-chain discussion, then a community vote — and reward-distribution parameters are precisely the class of change that draws the loudest objections, because storage providers are both the electorate and the affected party.
The Reward Actor's state carries a KPI. UpdateNetworkKPI takes a single scalar — currRealizedPower — and the reward function blends it with a baseline through BaselineExponent, set to ln(2), meaning the baseline power target doubles annually until the curve saturates. Simple minting and baseline minting are combined. One epoch, one number in, one number out.
That single-scalar design is the whole story. There is no dimension in the reward computation for "paid." Adding one is not a parameter tweak; it is a consensus-critical change to a state transition. New actor code, state migration, coordinated upgrade — not a soft fork, not a config flag.
So the real proposal is: add a second KPI. Then define it. Then defend it. Storage payments are denominated per epoch per byte, and deal terms vary in duration, replication, and retrieval path. Collapsing them into a scalar comparable to currRealizedPower requires a normalization convention. Every convention is a governance decision, and every governance decision is a lobbying surface.
Here is where it breaks, and I want the arithmetic rather than the hand-wave. Suppose a verified paid byte earns multiplier m against a raw byte. An adversarial provider becomes its own client: wallet A pays wallet B, and the flow books as revenue. True marginal cost is ProveCommit gas plus the time value of the pledge — and since gas is recycled into the network, the attacker's net cost is small relative to a block reward share. Self-dealing is profitable for essentially any m that meaningfully exceeds 1.
Therefore m has to sit near 1. Therefore the mechanism cannot work through pricing. It has to work through detection and slashing — graph analysis over client-provider payment flows, wallet clustering, temporal correlation between a deal payment and the provider's own treasury movement. That is an adjudication problem dressed as an economic one, and every adjudication layer is a governance chokepoint.
There is a cheaper path, and I expect convergence on it: reuse Filecoin Plus. The notary registry and verified-client allowlist already exist and already carry a multiplier. Add a price floor to the verified-deal definition, weight by payment, and you have paid storage without new metering infrastructure. It also imports Filecoin Plus's weaknesses wholesale — notary discretion, geographic concentration, and a review process never stress-tested against a coordinated set of self-dealing applicants.
Note also that any second KPI must be written by some actor holding write permission. Whoever holds that permission holds the emission dial, and that is the single most valuable privilege in the system.
One more note the announcement skipped: the oracle input is endogenous. In a price feed, latency is the failure mode — the feed reports a stale truth. Here the measured party produces the measurement. No faster feed fixes that. You fix it with cost of forgery, and the cost of forging a self-dealt deal is approximately gas.
Everyone reading this sees the upside: emission tied to revenue, supply effectively tightening, a cleaner adoption metric. The contrarian reading is that the reform assumes the demand it is meant to create, and if that demand is absent, tying emission to revenue converts a subsidy into a cliff.
Run it. Paid storage is under twenty percent of provider income. The proposal reweights payout toward that twenty percent. A provider with heavy raw-capacity exposure sees effective yield fall — not gradually, but at whichever epoch the multiplier activates. Some exit. Capacity contracts. PoSt security is a function of total power, so contraction is not merely a growth problem; it is a security parameter moving the wrong way, and it moves fastest where power is already concentrated.
I have watched this film. After Bitcoin's fourth halving, miner revenue collapsed and hash power drifted toward a handful of pools — the decentralization was never in the hardware, it was in the margin. Filecoin is running the same script in a different unit of account.
And "enhances network transparency" is the weakest item in the package. The causal chain from reward weighting to transparency is not obvious, and I suspect it is filler — the sentence you append when an announcement has four bullets and three ideas.
Watch three things: the FIP number, if one gets filed — that separates engineering from messaging. The multiplier m in the final text — anything above 1.5 behind a notary gate is a self-dealing invitation. And the on-chain ratio of paid to raw bytes, measured quarterly, not weekly.
My forecast: the proposal lands diluted. A multiplier under 1.2, a Filecoin Plus-shaped gate, a phased activation that lets large providers reposition. The supply-tightening narrative trades first and validates last.
The uncomfortable question is not whether Filecoin can measure paid storage. Roughly, it can. The question is whether a network that spent five years paying for capacity can survive being paid only for demand — and whether anyone checked the order of operations.
