AWS – Foto: Erlin Diah / Shutterstock.com
A failure in Amazon Web Services (AWS) cloud infrastructure caused instability in apps like iFood, Mercado Livre, and PicPay on Monday, October 20, 2025, affecting users across Brazil. The issue began around 4:11 AM Brasília time in the US-East-1 region, located in Virginia, USA, and spread globally, disrupting e-commerce, delivery, and digital payment operations. AWS identified increased error rates and latency across multiple services, leading to a surge of 6.5 million complaints on Downdetector worldwide. Local platforms reliant on Amazon’s infrastructure faced crashes in logins, shopping carts, and Pix transfers, prompting users to resort to bank branches as an alternative.
The incident underscored Brazil’s reliance on foreign cloud providers, with iFood reporting disruptions in orders in cities like São Paulo, Rio de Janeiro, and Belo Horizonte. Mercado Livre saw 20% of purchase completions stalled, while PicPay reported errors in 46% of mobile transactions. Normalization began at 8 AM, but setbacks occurred around 2 PM, prolonging the impact for millions of active accounts.
- Key affected services in Brazil: iFood (orders and delivery logins), Mercado Livre (cart and payments), PicPay (Pix transfers).
- Global impacted platforms: Snapchat (message access), Fortnite (game servers), Canva (design editing).
- Business tools: Zoom (video calls), Trello (project management), Wellhub (gym access).
Origin of the issue in AWS infrastructure
The outage stemmed from a DNS error in AWS, which disrupted connections between clients and data services, causing a cascading effect on databases and cloud computing. Amazon engineers mitigated the main issue by 11:35 AM Brasília time, but a backlog of pending requests kept residual instabilities in regions like US-East-1, the largest and most critical in AWS’s global network. This region, handling 32% of AWS cloud traffic, serves thousands of companies, including Brazilian fintechs using services like Elastic Compute Cloud (EC2) and DynamoDB for real-time transaction processing.
Platforms like Mercado Pago, integrated with Mercado Livre, faced disruptions in balances and Pix, affecting over 50 million accounts in Brazil. Stone reported similar spikes in mobile banking errors, forcing users to rely on traditional TED transfers. In delivery, iFood saw three-hour interruptions in delivery logins, coinciding with a 25% growth in delivery volume in 2025 compared to the previous year.
Brazilian services most impacted by the outage
iFood experienced slowdowns in order completions, with delivery workers in major cities reporting tracking and payment failures. The disruption hit during peak lunch hours, amplifying complaints on monitoring apps. Mercado Livre faced login and cart addition issues, halting bulk purchase flows.
Payment platforms like PicPay and Banco Pan saw errors in instant transfers, with Pix stalling at 1:20 PM. Users shifted to offline options, causing queues at physical bank branches. Claro and OLX also reported intermittent access, affecting classifieds and telecom services.
- Specific causes in Brazil: Reliance on US-East-1 servers for local app hosting.
- Average duration per service: iFood (three hours), PicPay (four hours with relapses).
- Alternatives used: TED transfers and in-store payments at partner locations.
Responses from affected companies
AWS issued initial statements at 5:44 AM, confirming investigations into latency errors and promising hourly updates. By noon, the company reported recovery in 90% of services, focusing on clearing pending operation backlogs. Direct clients like Snapchat and Prime Video received progress notifications, while indirect platforms monitored cascading impacts.
In Brazil, PicPay announced gradual normalization, advising users to retry failed transactions. Mercado Livre attributed issues to external infrastructure and activated support teams to manually process orders in distribution centers. iFood prioritized restoring delivery logins, ensuring continuity for urban routes.
These responses involved ongoing monitoring and social media communication, managing user flow during peak complaints. Fintechs like Stone implemented post-incident security updates to prevent recurrences.
List of global platforms with disruptions
Snapchat faced failures in loading messages and stories, impacting real-time user engagement. Fortnite reported multiplayer server crashes, interrupting active gaming sessions. Canva saw design editing stalls, affecting marketing professionals and content creators.
Zoom experienced delays in corporate video calls, peaking during morning meetings. Roblox and Duolingo had unstable access, halting lessons and virtual worlds. Pinterest and Perplexity reported slowdowns in searches and recommendations, while Alexa failed in voice command responses.
- Affected games: Fortnite (servers), Roblox (world access).
- Creative tools: Canva (editing), Pinterest (pinned content).
- Communication: Zoom (video), Snapchat (instant messaging).
AWS’s role in the digital ecosystem
Amazon Web Services operates across 38 geographic regions, providing hosting, databases, and AI tools for third parties since its inception to support Amazon’s operations. Competitors like Microsoft’s Azure and Google Cloud share the market, but AWS holds a 32% global share, making outages like this high-impact events. In 2024, a similar failure disrupted streaming in South America for two hours, highlighting vulnerabilities in central data centers.
In Brazil, a 25% rise in app-based deliveries in 2025 increased exposure to external disruptions, with fintechs processing billions in daily transactions via cloud. Reliance on a single region like US-East-1 underscores the need for diversification, though SLAs guarantee 99.99% annual uptime.
Updates on service recovery
AWS completed mitigation of the main DNS error, restoring connections in services like EC2 by 12 PM PST. However, residual instabilities persisted in support tools, such as API case creation. Platforms like Trello normalized first, followed by payment apps.
iFood delivery workers reported tracking stabilization around 2 PM, resuming pending routes. Mercado Pago automatically processed delayed balances, crediting accounts within 24 hours. PicPay users received alerts about fee refunds for failed transactions.
This final phase involved load testing to prevent new relapses, with AWS promising a detailed report within 48 hours.
