Coinbase Exchange began showing degraded performance at 17:56 PDT on May 7, 2026, and did not fully restore trading until 02:28 PDT on May 8. The root cause was a thermal event at an AWS data center in Northern Virginia, specifically Availability Zone use1-az4 within the US-EAST-1 region. The overheating caused a power loss and hardware damage. AWS confirmed the cause in a public statement: "Coinbase experienced service disruptions due to increased temperatures in the affected AWS service." Customer funds were not at risk at any point during the incident. The same incident also disrupted the CME, the world's largest derivatives marketplace.

The disruption unfolded in phases. Coinbase first placed its markets into "Cancel Only" mode, meaning traders could cancel existing resting orders but could not submit new ones. The exchange then moved into "Auction Mode," accepting limit orders for a minimum of ten minutes before matching them at a single opening price. This mechanism, borrowed from traditional equity exchange practice after trading halts, is designed to prevent sharp price dislocations when liquidity returns to a disrupted order book. Full trading resumed just after 2 a.m. Pacific time. Solana and ALEO network sends and receives were also delayed during the incident. Fiat services, Coinbase Custody, and Coinbase Wallet were not affected.

The timing put Coinbase in an uncomfortable position on multiple fronts. Hours before the outage began, Amazon and Coinbase jointly announced Amazon Bedrock AgentCore Payments, an enterprise payment system for AI agents that settles USDC transactions in roughly 200 milliseconds on Base, Coinbase's Layer 2 network. Base currently handles approximately 90 percent of global onchain agentic stablecoin transaction volume. The same AWS infrastructure at the center of that partnership announcement went offline the same evening. Coinbase also released its first-quarter 2026 earnings after market close on May 7, reporting a GAAP net loss of $394.1 million (driven partly by $482 million in unrealized crypto asset losses), revenue of $1.41 billion (missing analyst estimates of $1.48 billion and down 31 percent year over year), and a planned reduction of roughly 800 employees expected to generate approximately $500 million in annualized savings. The company simultaneously reported an all-time high global crypto trading market share of 8.6 percent. The earnings release arrived against an additional layer of unintended irony: in the days before the outage, Coinbase CEO Brian Armstrong had publicly discussed non-technical teams shipping code to production, a comment that drew wry notice from commentators once the platform went dark.

This is not the first time AWS has taken Coinbase down. In October 2025, a DynamoDB DNS failure cascaded across more than 58 services globally, affecting Coinbase alongside Robinhood, Snapchat, Roblox, and Reddit. That incident and this one have drawn renewed criticism of Coinbase's infrastructure architecture. Tech commentator Gergely Orosz put the concern plainly: "Coinbase seems to have a hard dependency on AWS, and when AWS...goes down." During the October 2025 outage, Binance, Kraken, and OKX remained fully operational. Those exchanges distribute their infrastructure across multiple cloud providers, multiple regions, or on-premises hardware. Context that sharpens the concern: Coinbase's status page recorded eight discrete degraded-performance events in April and May 2026 alone, a pattern that has amplified calls for infrastructure diversification.

---

Regional picture: Africa and South Asia

The impact varied considerably depending on where users were located. Most African users were insulated from the trading halt because Coinbase's primary presence on the continent runs through its Wallet product and a partnership with pan-African exchange Yellow Card, covering 20 countries including Nigeria, Kenya, Ghana, and South Africa. That integration uses self-custody rather than the Coinbase Exchange, so those users were not caught in the halt. However, the underlying exposure remains real. Users across those 20 countries depend on USDC-based remittance flows through Coinbase infrastructure because traditional cross-border transfer costs in Sub-Saharan Africa average 8.4 percent according to World Bank data.

A failure that extended to the Wallet layer could interrupt those flows directly.

In Nigeria, where regulatory friction from the Central Bank has already pushed many users toward peer-to-peer networks, exchange downtime reinforces existing skepticism of centralized platforms. The competitive pressure is concrete: Binance's Wallet has reached 30 million users in Nigeria with 4.5 percent monthly growth, and Binance remained fully operational during the outage, a contrast that further strengthens its competitive position on the continent.

Indian users faced a more direct hit. Coinbase re-entered India in December 2025 after completing Financial Intelligence Unit registration, and Indian traders with exchange accounts would have encountered the disruption at approximately 06:26 IST on May 8, during the morning session. India leads the world in grassroots crypto adoption according to Chainalysis's 2025 index, and the outage may push some retail activity toward locally regulated alternatives such as CoinDCX, WazirX, and Mudrex, which are not known to carry the same single-region AWS dependency.

For Web3 developers in South Asia and Africa evaluating Coinbase's Base network and x402 payment protocol for remittance or micro-payment applications, the incident is a straightforward due-diligence signal. Infrastructure analysts have argued that a thermal event in a single Virginia data center should not be capable of taking offline 90 percent of the world's onchain agentic stablecoin settlement capacity. Until Coinbase demonstrates meaningful infrastructure diversification, the gap between its decentralization narrative and its operational architecture will remain a live question for anyone building production-grade financial applications on its stack.