AWS failure takes down Mercado Pago, Mercado Livre and affects Pix and balances of millions of users in Brazil this Monday

A failure in Amazon Web Services’ cloud infrastructure caused widespread instability in Mercado Pago on Monday, October 20, 2025, affecting millions of active accounts in Brazil. The digital payment platform, part of Mercado Livre’s ecosystem, experienced disruptions in services like Pix, balance checks, and mobile app operations, starting around 4 a.m. Brasília time.
The issue stemmed from an interruption in the US-East-1 data center in Virginia, United States, where high error rates impacted the DynamoDB service and spread to global integrations. Amazon’s teams identified the DNS API issue as the cause and applied fixes, restoring most operations by mid-morning.
Users reported peak issues from 5:40 a.m., with complaint volumes 300% above normal in the early hours, according to monitoring data.
- Pix transaction errors accounted for 26% of cases.
- Unavailable balances affected 28% of reports.
- Mobile app access failures comprised 46% of issues.

Technical origin of the disruption
A failure in AWS’s domain name resolution system caused delays in data requests, impacting cloud-dependent platforms across Latin America. Mercado Pago, with over 50 million accounts in Brazil, faced rapid error propagation due to its deep integration with Amazon’s storage services.
Engineers confirmed the issue originated in one of the largest data centers, processing billions of daily global transactions.
Other impacted services
The AWS outage extended to multiple digital tools, halting critical functions across sectors. Design platforms like Canva reported slowdowns in collaborative editing, while Roblox multiplayer games stopped for thousands of players.
Prime Video streaming blocked content loading, and Alexa voice commands failed basic responses. Global financial apps like Venmo and Coinbase faced transaction delays, underscoring the interconnectedness of online ecosystems.
- United Airlines reported failures in digital check-in systems.
- Slack experienced errors in team integrations.
Full recovery occurred around 11 a.m., with tests confirming stability in 95% of affected operations.
User reports in Brazil
Freelancers highlighted blocks in bill payments and online sales, with Mercado Livre sellers delaying deliveries due to unconfirmed purchases. One user reported a failed attempt to use a virtual card at a store, receiving a validation error message.
Others awaited balance releases for urgent transfers, resorting to traditional bank cards. The complaint spike aligned with morning peak usage, amplifying daily disruptions.
Contingency measures implemented
Platforms activated redundancy protocols to reroute traffic to secondary servers, minimizing future error propagation. Push notifications updated users on service status via the app, enabling real-time monitoring of restorations.
For Pix transactions, alternatives like PicPay reported minor delays, serving as temporary workarounds. Post-incident security updates prioritized API configurations to reduce vulnerabilities in integrations.
Technical teams adjusted request limits to prevent similar overloads, with internal reports indicating 95% function restoration within six hours.
History of similar outages
AWS incidents have occurred annually, with durations ranging from minutes to hours at key data centers. In 2024, a similar South American outage disrupted streaming services for two hours, exposing regional infrastructure dependencies.
Mercado Pago implemented resilience tests after prior events, boosting recovery rates above 90% in global outages. Multi-region backup configurations now distribute loads to mitigate localized impacts, like this Virginia incident.
Tips for ongoing monitoring
Check the platform’s official status page for real-time service availability updates. Aggregator tools compile user reports, aiding in identifying patterns in isolated failures.
- Verify network connection before reporting issues.
- Update the app to the latest version available.
- Test functionalities on another device to isolate local problems.
Full normalization ensures transaction continuity, with a focus on technical prevention for operational stability.







