Google Cloud reported a significant service disruption on April 15, 2026, affecting its Antigravity Integrated Development Environment (IDE). The incident primarily impacted the AI Pro subscription tier, rendering the platform’s advanced coding assistance and model integration features unavailable for a duration exceeding nine hours. The outage began at approximately 04:00 UTC and persisted until 13:15 UTC, during which time developers reported a total loss of functionality for AI-driven workflows.
Technical logs shared by affected users and confirmed by Google’s status dashboard indicated a widespread occurrence of HTTP 503 Service Unavailable errors. Specifically, the IDE’s backend returned MODEL_CAPACITY_EXHAUSTED messages when users attempted to invoke the Gemini 3.5 Ultra or specialized Antigravity-native models. This error typically signifies that the underlying infrastructure is unable to handle the volume of requests or that the specific resource pool for the AI Pro tier has reached its operational limit.
A secondary and more critical technical failure involved a synchronization bug within the IDE’s billing and usage tracking module. Users reported that despite receiving error messages and failed model responses, the system continued to deduct tokens from their daily and monthly AI Pro quotas. In some instances, developers saw their entire daily allotment of 500 high-priority requests depleted within minutes of attempting to reconnect to the service. Google engineers identified this as a race condition where the quota-decrementing service acknowledged the request before the model inference engine returned the capacity error.
The Antigravity IDE, launched in late 2025 as Google’s flagship cloud-native development environment, relies heavily on distributed AI processing. The AI Pro tier, which costs 49 dollars per user per month, is marketed toward enterprise developers requiring low-latency access to large language models. Internal estimates suggest that approximately 1.2 million active subscribers were affected by the downtime. During the peak of the disruption, the Antigravity community forums saw a 400 percent increase in support tickets related to workspace connectivity and model timeouts.
In an official statement released at 14:00 UTC, Google Cloud’s engineering team attributed the failure to a misconfiguration in the regional load balancing layer that manages traffic for the Antigravity AI inference clusters. The company confirmed that a fix had been deployed to resolve the MODEL_CAPACITY_EXHAUSTED loop and that a script was being prepared to restore depleted quotas to affected accounts. Google has committed to providing a full Post-Mortem within 48 hours and indicated that Service Level Agreement credits would be issued to eligible enterprise customers.