Oracle: The Second Battlefield Behind the Prediction Market War

Author: Chloe, ChainCatcher

In the past two years, prediction markets have become the brightest narrative in the crypto industry. The total trading volume across the entire sector approached $10 billion by the end of last year, with monthly growth momentum significantly accelerating in the second half of 2025.

However, on the other side of this celebration, there is a role that has always stood outside the spotlight and has repeatedly been criticized by users: oracles.

The Double-Edged Sword of UMA

In the past year, several major controversies surrounding Polymarket have emerged, including whether Ukrainian President Zelensky "wore a suit" (with a total trading volume of $237 million), the Ukrainian mineral agreement (involving $7 million, with a major player using about 5 million UMA to manipulate votes), and whether the Trump administration will declassify UFO files in 2025 (a $16 million market, publicly labeled by users as a "whale proof" scam). The source of these controversies points directly to the same issue: UMA's Optimistic Oracle and its token governance structure.

The design logic of UMA's Optimistic Oracle is as follows: anyone can propose a result and stake a deposit; if no one disputes it during the challenge period (usually 2 hours), the result is assumed to be true by default; if there is a dispute, UMA token holders vote to decide through the Data Verification Mechanism (DVM).

The advantages of this mechanism are clear: it is cheap, can handle long-tail events, and can address "subjective issues," such as whether "Zelensky's outfit counts as a suit," which traditional price oracles cannot handle.

However, the controversies surrounding Polymarket have revealed flaws in this design. For example, in the Ukrainian mineral agreement incident last March, the prediction event had a total trading volume of about $7 million, focusing on whether Trump would reach a rare earth mineral agreement with Ukraine before April.

Despite no agreement being reached, the market was still settled as "Yes." According to reports from The Defiant and Cryptopolitan, the main reason was that a major UMA holder controlled about 5 million UMA through three accounts, accounting for about 25% of the voting weight, and pushed the vote towards Yes. Subsequently, Polymarket clearly stated in a Discord announcement: "This is not a system failure, but the result of the governance mechanism's operation, therefore we refuse to refund."

It can be said that Polymarket's reliance on UMA is facing systemic risks. Originally designed as a "neutral truth adjudication layer," the oracle's centralized distribution of governance tokens has instead become a tool for a few individuals to influence market outcomes.

According to the crypto asset data platform RootData, until September of last year, when Polymarket began to promote cryptocurrency events, it urgently needed to introduce a more reliable data source, so it started to delegate part of the settlement work to another completely different oracle system, Chainlink.

Chainlink: Another Dilemma for the Leader

CoinDesk reported that Polymarket began to introduce Chainlink to improve its method of determining prediction results. The two parties announced that Polymarket would use Chainlink to automatically settle markets related to asset prices, reducing latency and the risk of manipulation. Initially focusing on cryptocurrency price markets, they also explored applications in more subjective markets.

The significance of this collaboration lies in Polymarket shifting from relying on UMA's "crowd-gaming style subjective consensus adjudication" to a path where Chainlink directly reads market prices and automates determinations.

From a market perspective, Chainlink is the undisputed leader in the oracle space, with a market capitalization share exceeding 87% and a TVS share of 61.58% (approximately $62.9 billion), creating a significant gap with the second place Chronicle (10.15%) and third place RedStone (7.94%).

It can also be said that its penetration in DeFi is almost saturated. Mainstream protocols, from Aave, GMX, and Synthetix's liquidation and pricing to Curve's security references and Lido's cross-chain standards, almost all utilize different services provided by Chainlink.

The market share is reflected in its layout. Chainlink provides 2,000 price feeds across about 27 chains (on-chain resident price feeding services) and has deployed Data Streams (low-latency, on-demand verification high-frequency feeding services) on 37 networks; the CCIP (Chainlink Cross-Chain Interoperability Protocol) mainnet has covered 70 public chains and L2, with about 200 tokens registered as CCIP standards available for use.

This scale means Chainlink has expanded from being a "single-chain price feeding intermediary" to an "information and asset exchange layer between multiple chains."

But saturation also means that DeFi is no longer its growth curve. According to a deep report from Galaxy, about 97% of Chainlink's cumulative revenue (approximately $399 million) comes from Price Feeds, while VRF (Verifiable Random Function, used for NFT minting and on-chain gaming), Automation (automated execution), and CCIP together account for only about 1.5%, 0.6%, and 0.5%.

In other words, Chainlink's cash flow is highly concentrated in the most mature and commoditized price feeding business, and this segment's market is already saturated, with extremely limited marginal growth potential.

In response, Chainlink has placed its bets on three incremental curves.

The first is RWA and institutional finance.

From Chainlink's partnership matrix, it can be seen that it has previously collaborated with Swift and several institutions to complete cross-chain experiments for tokenized assets; last year, it further advanced on-chain plans for corporate actions with 24 major financial institutions, while 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 purchase processes for over 3 billion cardholders; the U.S. Department of Commerce (BEA) has also put core macro 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 choices for cross-chain standards. JPMorgan's Kinexys collaborated with Chainlink and Ondo to complete cross-chain DvP settlement experiments for tokenized U.S. Treasury bonds; Aave uses it to promote GHO cross-chain, and Lido has adopted it as the official cross-chain standard for wstETH; in the same year, CCIP also launched on Aptos, extending its reach into the Move ecosystem.

As of October 2025, CCIP has accumulated nearly $2 billion in token transfers.

The third is prediction markets and "event settlement financialization."

The integration with Polymarket marks the beginning of this curve. It represents Chainlink's expansion from serving only "asset prices" to the broader field of "event settlement." As the demand for assets that can be automatically settled in prediction markets explodes, including U.S. stocks, commodities, ETFs, and macro indicators, Chainlink finds a natural extension of its original price business here.

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 its next growth curve.

The potential on these curves is significant. According to BCG estimates, the scale of RWA tokenization 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 token holders' patience is usually measured in "days."

This mismatch may be the core pressure Chainlink, as a leader, will still face in 2026.

Multiple Oracles Eating Away at the Prediction Market Pie

In early April this year, Polymarket announced a partnership with Pyth Network.

On this platform, there are prediction markets for commodities such as gold, silver, WTI crude oil, and natural gas, as well as over a dozen U.S. stocks including NVDA, AAPL, TSLA, COIN, PLTR, and major indices and ETFs for "short-term fluctuations." Settlement data will be provided in real-time by Pyth via WebSocket, with Polymarket sampling once per second.

Pyth, as a first-party data provider (with market makers and institutions like Jump Trading, Jane Street, Blue Ocean, LMAX directly publishing), uses a pull model for on-demand data delivery, allowing low-latency delivery to the application layer.

This division of labor structure is not just a choice for Polymarket alone. Kalshi, regulated by the U.S. 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, with plans to expand to indices, stocks, and foreign exchange categories.

When both Polymarket and Kalshi choose Pyth as the settlement layer for traditional financial assets, it reflects a broader trend in the prediction market sector towards a "high-frequency data settlement layer for institutional use."

Pyth thus captures a portion of the market in this area, but this position is a subset of "traditional financial asset events," with Chainlink focusing on crypto and UMA on subjective categories, each occupying its own space.

From this three-tiered division of labor structure, we can observe the realities of the oracle sector revealed by prediction markets.

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 feeding model is not optimal for millisecond-level event settlements; while Pyth has clear advantages in low-latency pricing, it cannot fully handle text-type issues.

Second, every time Polymarket introduces a new oracle, it expands the map of "tradable events."

From UMA's non-standard events to Chainlink's crypto assets and Pyth's traditional financial assets, each step brings more uncertainties from the real world into the on-chain betting scope. Following this logic, future macroeconomic indicators (GDP, CPI, interest rate decisions), central bank interest rate decisions, public company earnings, and even AI model releases could all become market categories for Polymarket.

As long as there is a verifiable data source, corresponding markets can be constructed.

Conversely, for oracle projects, this also means that the wild expansion of prediction markets will not allow any single oracle to enjoy the dividends alone. Each new market will be assigned to the oracle "most suitable for handling that type of data structure," with multiple players sharing the pie without overlap.

Conclusion

By 2026, the oracle sector has essentially evolved from the early "data pipeline" to a "verifiable fact layer" supporting the entire on-chain economy.

Its services are no longer limited to DeFi liquidation and collateral valuation but extend to compliance verification for RWA on-chain, trustworthy transmission of cross-chain information, and the settlement of real-world uncertainties in prediction markets.

And prediction markets serve as a magnifying glass to observe the competition in this red ocean.

Polymarket's three-tier division of labor, along with Kalshi's simultaneous choice in traditional financial assets, reveals a reality: no single oracle can fully serve a mature on-chain application. Every topic on the platform will be assigned to the oracle best suited to handle that type of data structure.

Infrastructure differentiation is already a fact. But when no single project can enjoy the dividends alone, who can truly become irreplaceable?