What distinguishes a centralized (single-source) oracle from a decentralized (consensus-based) one, and why does it matter?

A centralized oracle trusts one source and is itself a single point of failure; a decentralized oracle aggregates many independent sources so no single one can lie or fail unnoticed.

A single-source oracle is simple and cheap, but it re-creates the very thing a blockchain tries to avoid: one party whose word everyone must take, and whose compromise or outage poisons every contract depending on it. A decentralized oracle spreads the job across multiple independent instances and combines their answers (majority vote, median, staked aggregation), so corrupting the feed requires subverting many parties at once rather than one. The trade-off is added cost, latency, and coordination — but for high-value contracts it turns a fragile trusted component into a much harder target.

Go deeper:

• Chainlink — decentralized data model — how a production decentralized oracle aggregates many independent operators into one feed.
• Chainlink 2.0 whitepaper — the design of Decentralized Oracle Networks (DONs) and their cryptoeconomic security.
• Chainlink (Wikipedia) — background on the best-known decentralized oracle network.