Pyth Network (PYTH): the oracle after its 2026 overhaul and the 100% Rule

Pyth Network is a price oracle: trading firms and exchanges publish their own prices to it, and Pyth turns them into one feed, with a confidence interval, that lending protocols and perpetual exchanges read on more than 100 blockchains. Since 30 September 2026 the Pyth DAO has spent every dollar it receives from Pyth products on buying PYTH, its own token, on the open market. Pyth calls this “the 100% Rule”.

The rule comes from OP-PIP-136, a proposal posted on the Pyth DAO forum on 24 September 2026 and now marked as passed. Pyth announced it on its blog on 8 October. It closes a year of deep changes for the network: its own appchain is gone, data access now needs an API key, staking rewards have stopped and most income comes from paid subscriptions.

What is Pyth Network

A blockchain cannot look up the price of bitcoin, gold or Apple stock by itself. An oracle writes that data on-chain, and a lending protocol fed a wrong price can liquidate healthy loans.

Pyth’s answer, since the start, has been “first-party” data. Instead of paying node operators to scrape public price sites, it asks the firms that make the prices to send them directly. Its April 2021 announcement listed Jump Trading, Virtu, Hudson River Trading, Cumberland and CTC among the first publishers. The project was incubated by Jump, and its mainnet went live on Solana on 26 August 2021 with more than 30 assets, quotes every 400 milliseconds and up to 32 providers per asset.

Three bodies share the work today. The Pyth Data Association, a Swiss non-profit, commissioned most of the published audits. Douro Labs, whose chief executive Mike Cahill is described on Pyth’s blog as a core contributor to Pyth Network, operates the Pyth Pro data stack on behalf of the DAO. The Pyth DAO, which votes on Solana through Realms, decides how the network’s share of revenue is used. Its Pythian Council carries out what the votes approve.

How the Pyth oracle works: pull updates and confidence intervals

Most oracles push prices on-chain on a schedule, or when the price moves past a threshold. Pyth works the other way round. In its “pull” model, the docs explain, anyone can submit a price update to the Pyth contract, and an app must update the on-chain price before reading it. The trader brings the latest signed price along with the transaction. Feeds update every 400 ms, and the gas for writing on-chain is spent only when someone needs the number.

The second idea is the confidence interval. “At every point in time, Pyth publishes both a price and a confidence interval for each product,” reads the best-practice guide. The band widens when publishers disagree. Integrators are told to value collateral at the lower edge of the band and debts at the upper edge, and to halt new activity on an asset when the band grows too wide against the price. Its SDKs refuse stale prices by default, and equity feeds follow market hours.

The Pyth Pro pipeline

Since 2026 every Pyth price comes out of one stack, first called Lazer and now sold as Pyth Pro. Publishers send updates over authenticated WebSocket connections. Relayers run by Douro Labs check them and feed an ordered message queue. Aggregators built into the routers read that queue and compute the median price, a confidence interval from the spread of the middle half of the quotes, best bid and ask and, for perpetual markets, funding rates.

Clients subscribe to fixed-rate channels of 50, 200 or 1,000 ms, or to a real-time channel that Pyth’s 23 April 2026 blog put at 1 ms.

From Pythnet and Wormhole to five signing routers

Until August 2026 the path was different. Publishers sent prices to Pythnet, an application-specific blockchain that aggregated the prices. The 19 guardians of the Wormhole bridge then signed the result, with 13 signatures needed, and a service called Hermes handed the signed updates to apps on more than 100 chains.

The switch came late. A March 2026 forum plan aimed at 1 July, a May blog post at 31 July; the docs now say the Pyth Core upgrade completed on 26 August 2026 at 16:00 UTC. Under the new design, “five independently operated routers” sign Merkle roots of the aggregated prices, and three signatures out of five are enough. Contracts on each chain check those signatures against the quorum and the price against the Merkle root. They sit at new addresses but keep the same ABI, so integrators did not have to rewrite code.

On the EVM side, OP-PIP-132 (25 August 2026) pointed 33 deployments, 19 of them on mainnets, to the new data source and to a “Pyth Pro guardian set” of five keys. Old Pythnet updates no longer verify. The docs do not name the operators of the five routers, and we found no audit of the cutover contracts or of the five-router set-up in Pyth’s public audit repository.

Paid access from $500 a month, and products cut

The economic case was written down plainly. A March 2026 forum post by Douro Labs put Pyth Core income at about $9,000 in February against about $108,000 gross for Pyth Pro, while running Pythnet and two parallel systems cost more than $1 million a year. OP-PIP-100 passed in April 2026. “Pythnet is being retired,” the blog said on 23 April.

Users pay for that choice. “Hermes now requires an API Key,” the upgrade docs say; Hermes is the service apps call for signed prices, and a blog post of 26 May 2026 had set the requirement for after 31 July. The free plan gives the Pyth Terminal only, with no API access. Starter costs $500 a month for crypto prices at up to one update a second. Pro starts at $2,500 a month for every asset class at up to 1 ms. Keys come with a free trial. On-chain verification is supported by default on the major EVM chains, Solana and Sui.

We could not confirm whether any free on-chain path is left for small DeFi apps.

Two products went the same year. Express Relay, an auction product for liquidations and order flow, was wound down by OP-PIP-124 on 22 July 2026: it brought the DAO under $2,000 a month, cost more than that to run, and a planned sale to another Solana protocol fell through.

Oracle Integrity Staking (OIS) was the other. Launched in late 2024 with a 100 million PYTH reward pool from the Pyth Data Association, it drew about 1 billion PYTH across some 120 publisher pools, and no slashing proposal ever reached the DAO. OP-PIP-103 cut its rewards to zero in April 2026.

Then, on 7 October, CO-PIP-139 retired the Price Feed Council. The proposal, now marked passed, says Pythnet “has now been sunset” and keeps OIS only as a “legacy, unstake-only” reference. We did not find the exact day Pythnet stopped.

The April blog post defended the end of OIS rewards in one line: “Enforcement does not require ongoing emissions”, only stake that can be lost. Governance staking still works.

Pyth Pro revenue: what the Douro Labs reports show

Pyth Pro launched on 23 September 2025 as a subscription for banks, trading firms and other data buyers. Douro Labs reports its income on the DAO forum every month. Under CO-PIP-9 the DAO gets 60% of Pyth Pro and Pyth Indices revenue and 90% of Listing as a Service (LaaS) revenue; Douro keeps the rest.

The report posted on 5 October 2026 gives these figures:

Product	Split DAO / Douro	Gross	DAO share
Pyth Pro	60 / 40	$768,084	$460,850
Listing as a Service	90 / 10	$12,700	$11,430
Pyth Indices	60 / 40	$135,688	$81,413
Total		$916,471	$553,693

The report labels the period in two ways and indices are billed in arrears, so this is best called the distribution of 5 October 2026. Since September 2025 the total is $4,135,517 gross, of which $2,506,630 went to the DAO.

The first report, on 2 December 2025, counted $262,221 collected since launch and committed annual recurring revenue (ARR) above $1 million. Pyth gave more than $10.4 million of ARR for August 2026 and $11.5 million for September, up 86% on the quarter: Pyth’s own “overall ARR” figure, not an audited one. Indices account for $1.81 million of fixed ARR.

DefiLlama’s fee pages, read on 16 October 2026, show the other side. Pyth Core revenue was $311,603 over a year but close to zero since late August, after OP-PIP-128 and OP-PIP-130 set EVM protocol fees to zero. Entropy, the on-chain randomness service, made $638 in 30 days. Nearly all of Pyth’s income is now subscriptions.

The 100% Rule: what the PYTH buyback really covers

The DAO started buying its own token on 12 December 2025, when Pyth launched the PYTH Reserve. Under that first rule, OP-PIP-87, it spent one third of its non-PYTH treasury each month, with a vote each time. The amounts shrank, from about 2.75 million PYTH in March 2026 to 669,662 in August, by the forum’s figures. The last round under the old rule, funded on 15 September, turned 161,714.28 USDC and 45.30 SOL into 2,582,433.55 PYTH, according to the September purchases report.

OP-PIP-136, “Pyth Strategic Reserve V2”, posted by the forum user KemarTiti, replaces the one-third rule and drops the monthly votes. The Realms vote page did not load for us, so we cannot give the vote count or closing day. The proposal is marked passed, and Pyth says the first purchases under it took place on 30 September. The first transfer sent 323,428.56 USDC and 90 SOL to the Pythian Council’s operations multisig, to be swapped for PYTH.

100% of the DAO’s share, not 100% of revenue

Some coverage summed it up as Pyth spending all of its revenue on PYTH. That is not what the text says. The rule covers the money the DAO itself receives: its approved share of Pyth Pro, LaaS, the Data Marketplace and Pyth Indices, plus Pyth Core and Entropy fees and other assets sent back to the DAO. The forum discussion that led to the proposal was direct about it: “This does not mean 100% of gross revenue.” Douro Labs keeps its contractual 40% of Pyth Pro and Indices income and 10% of LaaS income.

The proposal adds that it “applies only to the DAO’s existing approved share and does not change any product’s commercial terms.” On the 5 October distribution, the buyback budget was $553,693, not the $916,471 billed. Pyth’s 8 October post is worded the same way: “The Pyth DAO has committed 100% of the funds it receives from Pyth products to its PYTH Reserve.”

The guardrails on each purchase

The trading rules carry over from OP-PIP-87:

• at most 5% slippage per transaction and at most $25,000 per transaction;
• orders go through an aggregator first, with Jupiter named as the example;
• limit orders sit at least 0.1% below the displayed price;
• every transaction is published as proof, and the monthly purchase reports continue.

The Council may not sell the PYTH it buys, borrow against it, use leverage or derivatives, or distribute it without a separate DAO proposal. PYTH handed over directly, rather than bought, is not to be presented as an open-market purchase.

Accumulation or buyback: the forum debate

That last line answers an objection raised in the forum thread on 9 September. A member posting as scp wrote that “the core issue here is the distinction between accumulation and buybacks”: if Douro Labs paid the DAO in PYTH from its own holdings, no token would be bought from the market. Douro’s reply, from the account zenyas, was that “Douro Labs will continue transferring the DAO share of the revenue in USDC on a monthly basis.” The USDC is then swapped for PYTH on-chain.

OP-PIP-140 and the size of the Reserve

A follow-up, OP-PIP-140, posted on 7 October and marked passed, fixes the plumbing. It recovers 34,375.87 USDC stuck in a wrongly created token account and moves the DAO’s full USDC balance, 748,076.17 USDC, and about 880.9 SOL to the Council’s multisig. That sum includes the latest Douro Labs distribution, which landed in the treasury instead of going straight to the multisig. Whether that balance has been swapped yet is not known: the October purchases report is not out.

How big is the Reserve? The 8 October post gives 41 million PYTH in one paragraph and 42 million in another, so we write 41 to 42 million until the next report. Mike Cahill called the commitment “the strongest alignment this network has had”.

How big the PYTH buyback is next to the supply

A rough Coindaro calculation gives the scale. The 5 October distribution to the DAO was $553,693. With PYTH trading near $0.088 at 13:54 UTC on 16 October 2026, that buys roughly 6.3 million PYTH, about 0.08% of the 7.875 billion PYTH in circulation according to CoinGecko.

Against that sits the last scheduled token unlock. By our subtraction (10 billion maximum minus 7.875 billion circulating), about 2.125 billion PYTH remain locked, due in May 2027; the exact day and split are not verified. The buyback grows with revenue, but at current figures it is small next to that release.

Readers searching for a Pyth Network price prediction will not find one here. Coindaro does not forecast prices. The supply facts are above: monthly purchases, the May 2027 release and a Reserve that cannot be sold without a vote. Nothing here is investment advice.

PYTH token: supply, allocation and unlocks

PYTH is an SPL token on Solana with a maximum supply of 10 billion. It went live on 20 November 2023 with 1.5 billion tokens (15%) in circulation and an airdrop to more than 90,000 wallets on 27 chains. The distribution page says locked tokens unlock 6, 18, 30 and 42 months after launch.

Category	Share	Amount	Unlocked at launch
Ecosystem growth	52%	5.2B PYTH	700M
Publisher rewards	22%	2.2B PYTH	50M
Protocol development	10%	1B PYTH	150M
Private sales	10%	1B PYTH	none
Community and launch	6%	600M PYTH	600M (all)

The first unlock, on 20 May 2024, released about 2.13 billion PYTH, which The Block valued at $1.25 billion at late-April prices. The third, in May 2026 (sources give 19 or 20 May), released about 2.13 billion again, 1.13 billion of it for ecosystem growth. A community idea to delay it by six months was not binding and did not happen. The private sales share covers two historical rounds whose amounts were never published.

PYTH can still be staked for governance; the yield programme is over. DefiLlama counted 1,116,601,382 PYTH staked on 16 October 2026 at 10:10 UTC across all programmes. Votes run on Realms, as Operational (OP-PIP) or Constitutional (CO-PIP) proposals.

Who uses Pyth: adoption figures as Pyth reports them

Most adoption numbers come from Pyth itself. Its homepage lists more than 138 data publishers, 720 data consumers and 3,600 live price feeds, with figures “as of September 2026”. The KPI dashboard is still dated April 2026. It then showed $8.5 billion of total value secured, down 29.8% in a month, and at least 114 connected blockchains.

Pyth’s strongest claim is in tokenised stocks, commodities and FX traded as perpetual futures. The 8 October post says more than 94% of tracked volume in these “RWA perps” ran on its data over the past quarter, more than $2 trillion. Its own September report gives a lower share for that month alone, about 90%. Pyth names Trade[XYZ], Lighter, Kraken, Coinbase and Phoenix among perp venues using its data, and Kalshi and Polymarket among prediction markets. It says Kalshi and Coinbase named it in CFTC and SEC filings; we did not open those filings.

On the data-vendor side, the September Douro report says Pyth was approved as a distributor of Nasdaq Basic, and Pyth Pro went live on Stellar. The Data Marketplace resells data from Euronext FX, Fidelity Investments, Tradeweb and others. In August 2025 the US Department of Commerce chose Pyth to publish GDP data on-chain, with Chainlink handling other Bureau of Economic Analysis feeds.

Pyth Network vs Chainlink

The two oracles are built differently. Chainlink pays independent node operators to fetch and push prices on-chain on a heartbeat or deviation rule. Pyth gathers prices from the firms that trade and lets each app pull an update when it needs one. Chainlink still secures far more value, going by DefiLlama figures cited by the Italian site Criptovaluta.it on 16 October. We could not open DefiLlama’s oracle dashboard ourselves, so we give no market share.

Where Pyth leads, by its own count, is speed and non-crypto markets such as RWA perps. Since 2026 the difference also runs through security design. Chainlink keeps a decentralized node network; Pyth signs its prices with five routers and runs its pipeline through one company.

Pyth incidents and risks

The failures are public. On 20 September 2021, between 12:21 and 12:23 UTC, the BTC/USD aggregate dropped to $5,402 while exchanges traded near $43,500. Pyth said two sources had published a near-zero price, CoinDesk reported, and apps that used the price without its very wide confidence band allowed liquidations. On 4 December 2021, faulty mSOL prices with wide bands led to $124,000 of liquidations on three Mango Markets accounts, according to Mango’s post-mortem; Mango then filtered out prices with wide bands.

On 22 May 2026 Pythnet stopped producing blocks and Pyth Core feeds went dark until a coordinated validator restart, about four hours by the count of the French site CryptoActu. Pyth said its Pyth Pro infrastructure ran normally throughout. We found no post-mortem.

Coindaro rates Pyth Network 3 out of 5 for risk. On the plus side: five years of operation, revenue paid in USDC and reported monthly, first-party data and audits by Zellic, OtterSec, Trail of Bits, CertiK and Sec3. We found no case of Pyth itself losing user funds. Against it:

• prices are now signed by three of five routers instead of 13 of 19 Wormhole guardians, and Douro Labs runs the off-chain stack;
• OIS, the staking layer meant to make publishers accountable, is legacy, and the Price Feed Council is gone;
• about 2.125 billion PYTH unlock in May 2027, against purchases of roughly 6 million PYTH a month at current figures;
• paid API access may push small DeFi integrators to other oracles.

What to watch next at Pyth Network

The October purchases report will show the first full month under the 100% Rule and whether the 748,076 USDC moved by OP-PIP-140 has been converted. Two open questions matter for security: who operates the five signing routers, and whether an audit of the new Core contracts is published. The September Douro Labs report also lists corporate-action data for US equities (splits, dividends, mergers) as coming soon, with no date. The next fixed date is May 2027, when the last 2.125 billion PYTH are scheduled to unlock.