Author: Chloe, ChainCatcher
Over the past two years, prediction markets have become the most prominent narrative in the crypto industry. The entire sector saw a total transaction volume of nearly $10 billion by the end of last year, and monthly growth momentum is expected to accelerate significantly in the second half of 2025.
But on the other side of this celebration, there is a role that always stands out from the spotlight and is repeatedly criticized by users: the oracle.
UMA's double-edged sword
The numerous major controversies surrounding Polymarket over the past year—including whether Ukrainian President Zelenskyy wore a suit (with a cumulative trading volume of $237 million), the Ukraine mining agreement (involving $7 million, with some large investors manipulating votes using approximately 5 million UMA), and whether the Trump administration will declassify UFO documents in 2025 (a $16 million market, publicly referred to by users as a "whale proof" scam)—all point directly to the same source: UMA's Optimistic Oracle and its token governance structure.
UMA's Optimistic Oracle design logic is as follows: anyone who proposes a result puts up a deposit; if no one raises a dispute during the challenge period (usually 2 hours), the result is considered true by default; if there is a dispute, the UMA token holders will vote through the Data Verification Mechanism (DVM) to decide.
The advantages of this mechanism are obvious: it is cheap, can handle long-tail events, and can handle "subjective problems," such as "whether Zelensky's shirt counts as a suit," which traditional price oracles cannot handle at all.
However, several controversies surrounding Polymarket have exposed the flaws in this design. For example, the Ukraine mining deal controversy last March, which saw a total trading volume of approximately $7 million, focused on whether Trump would reach a rare earth mining agreement with Ukraine before April.
Despite no agreement being reached, the market settled with a "Yes". According to The Defiant and Cryptopolitan, the main reason was that a large UMA holder held approximately 5 million UMA through three accounts, representing about 25% of the voting weight in that round, and pushed the vote to "Yes". Polymarket subsequently clarified in a Discord announcement: "This was not a system malfunction, but a result of the governance mechanism's operation, therefore refunds are refused."
It can be argued that Polymarket's reliance on UMA is facing systemic risks. Originally designed as a "neutral truth-setting layer," the oracle's centralized distribution of governance tokens has now turned it into a tool for a few to manipulate market outcomes.
According to RootData, a crypto asset data platform, until September of last year when Polymarket began to promote cryptocurrency events, it urgently needed to introduce a more certain data source, so it began to outsource some settlement work to Chainlink, an oracle with a completely different system.
Chainlink: Another Dilemma for Industry Leaders
CoinDesk reports that Polymarket is introducing Chainlink to improve its prediction accuracy. The two companies announced that Polymarket will use Chainlink to automatically settle asset price-related markets, reducing the risks of delays and tampering. Initially focused on crypto asset price markets, they are simultaneously exploring its application in more subjective markets.
The significance of this collaboration is that Polymarket has moved from relying on UMA's "subjective consensus-based decision-making through group game" approach to having Chainlink directly read market prices and make automated judgments.
From a market perspective, Chainlink is the undisputed leader in the oracle sector, accounting for over 87% of the market capitalization and 61.58% of TVS (approximately $62.9 billion), significantly ahead of the second-ranked Chronicle (10.15%) and the third-ranked RedStone (7.94%).
In other words, its penetration in DeFi is almost saturated. Mainstream protocols, from Aave, GMX, and Synthetix for liquidation and pricing, to Curve for security and Lido for cross-chain standards, almost all use different services provided by Chainlink.
Market share is reflected in its deployment. Chainlink provides 2,000 price feeds (on-chain persistent price feed services) on approximately 27 chains and has deployed Data Streams (low-latency, on-demand verified high-frequency price feed services) on 37 networks; the CCIP (Chainlink cross-chain communication protocol) mainnet has covered 70 public chains and L2, and approximately 200 cross-chain tokens registered as CCIP standards are available.
This scale is equivalent to Chainlink expanding itself from a "price feed intermediary on a single chain" to an "information and asset exchange layer between multiple chains".
However, saturation also means that DeFi is no longer on its growth curve. According to Galaxy's in-depth report, about 97% of Chainlink's cumulative revenue (about $399 million) comes from Price Feeds, while VRF (verifiable random number generator, used for NFT minting and on-chain games), Automation (automated execution), and CCIP account for only about 1.5%, 0.6%, and 0.5% respectively.
In other words, Chainlink's cash flow is highly concentrated on its most mature and commoditized price feed business, which is already saturated and has extremely limited room for marginal growth.
Chainlink's response is to bet on its three incremental growth curves.
The first item is RWA and institutional finance.
Chainlink's partnership matrix shows that it has previously collaborated with Swift and several other institutions to complete cross-chain trials of tokenized assets; last year, it further promoted the plan to put corporate actions data on-chain with 24 major financial institutions, and the DTCC Smart NAV pilot distributed mutual fund NAV data on-chain.
In the same year, Chainlink partnered with Mastercard to enable on-chain crypto purchases for over 3 billion cardholders; the U.S. Department of Commerce (BEA) has also put core macroeconomic data such as GDP and PCE on-chain through Chainlink Data Feeds, initially covering 10 public chains.
The second is CCIP cross-chain communication.
CCIP has become one of the cross-chain standards. JPMorgan's Kinexys, in collaboration with Chainlink and Ondo, completed a cross-chain DvP settlement experiment for tokenized US Treasury bonds; Aave used it to promote GHO cross-chain, and Lido adopted it as the official cross-chain standard for wstETH; in the same year, CCIP was also launched on Aptos, extending its reach to the Move ecosystem.
As of October 2025, CCIP's cumulative token transfer volume was nearly $2 billion.
The third point is prediction markets and "financialization of event settlement".
The integration with Polymarket marks the beginning of this curve. It represents Chainlink's expansion from its original focus on "asset prices" to the broader field of "event settlement." As the demand for automated settlement of asset classes such as US stocks, commodities, ETFs, and macroeconomic indicators from prediction markets exploded, Chainlink found a natural extension to its original price-related business.
Overall, while Chainlink holds a leading position in the market, the growth of traditional DeFi price oracles has peaked; it must rely on RWA, institutional finance, CCIP, and the financialization of prediction markets to rebuild the next growth curve.
The potential on these curves is considerable. According to BCG estimates, the tokenization scale of RWA could reach $16 trillion by 2030, and the SWIFT track processes $150 trillion in settlements annually, but the realization cycle is measured in "years," while the patience of token holders is usually measured in "days."
This mismatch may be the core pressure that Chainlink, as the industry leader, will still face in 2026.
Multiple oracles are encroaching on the BTC of the prediction market.
In early April of this year, Polymarket announced a partnership with Pyth Network.
The platform offers short-term price fluctuation predictions for commodities such as gold, silver, WTI crude oil, and natural gas, as well as more than ten US stocks including NVDA, AAPL, TSLA, COIN, and PLTR, and major stock indices and ETFs. Settlement data will be provided in real time by Pyth via WebSocket, with Polymarket sampling once per second.
As a first-party data provider (directly published by market makers and institutions such as Jump Trading, Jane Street, Blue Ocean, and LMAX), Pyth adopts an on-demand pull model, enabling data to be delivered to the application layer with low latency.
This division of labor isn't just Polymarket's choice. Kalshi, regulated by the US CFTC, has also integrated Pyth as its settlement data source for its newly launched commodity center, covering commodities such as gold, silver, Brent crude oil, natural gas, copper, corn, soybeans, and wheat. Pyth Pro also provides direct market data access to Kalshi's market makers, and will be expanded to indices, stocks, and forex in the future.
When both Polymarket and Kalshi choose Pyth as the settlement layer for traditional financial assets, this is not just an engineering decision of individual platforms, but reflects the convergent demand of the entire prediction market sector for an "institutional-grade high-frequency data settlement layer".
Pyth has captured a portion of the market in this area, but this position is a subset of "traditional financial asset events," sharing a niche with Chainlink's crypto-related events and UMA's subjective events.
This three-tiered division of labor structure allows us to observe the reality of the oracle market as revealed by the prediction market.
First, no single oracle can fully serve a mature prediction market.
UMA's community adjudication mechanism cannot handle high-frequency prices; Chainlink's on-chain feed model is not optimal for millisecond-level event settlement; Pyth, while having a clear advantage in low-latency prices, is completely unable to handle text-based issues.
Second, every time Polymarket introduces a new oracle, it expands its reach beyond the realm of "tradable events".
From UMA's non-standard events to Chainlink's crypto assets, and then to Pyth's traditional financial assets, each step incorporates more real-world uncertainties into the scope of on-chain betting. Following this logic, future macroeconomic indicators (GDP, CPI, interest rate decisions), central bank interest rate decisions, listed company profits, and even the release of AI models could all become market categories on Polymarket.
As long as a verifiable data source exists, a corresponding market can be built.
Conversely, for oracle projects, this also means that the rapid expansion of the prediction market will not allow any single oracle to monopolize the benefits. Each new market will be allocated to the oracle "most suitable for handling that type of data structure," with multiple oracles sharing the spoils without overlap.
Conclusion
By 2026, the oracle track has essentially evolved from an early "data pipeline" to a "verifiable fact layer" that supports the entire on-chain economy.
Its services are no longer limited to DeFi liquidation and collateral valuation, but also include compliance verification of RWA on-chain, trusted transmission of cross-chain information, and settlement of real-world uncertainties in prediction markets.
The prediction market serves as a magnifying glass for observing the competition in this red ocean market.
Polymarket's three-track division of labor, coupled with Kalshi's simultaneous choice on traditional financial assets, reveals a reality: no single oracle can fully serve a mature on-chain application. Each issue on the platform is assigned to the oracle best suited to handle that type of data structure.
Infrastructure fragmentation is a reality. But when no single project can exclusively reap the benefits, who can truly be irreplaceable?
