GitHub's June 2026 availability report disclosed six separate service incidents in a single month, several tied directly to the load Copilot and other AI features are putting on infrastructure that was not built for it. The same month, Google Cloud's May 1 doubling of CDN Interconnect, Direct Peering and Carrier Peering egress prices started showing up on customer invoices. Neither event alone is unusual for a hyperscaler. Together they show the two pressures platform teams are absorbing at once this year: infrastructure that buckles under AI-driven traffic, and network bills that are rising to pay for the capacity behind it.
GitHub's June Was Rough by Its Own Account
GitHub publishes a monthly availability report, and the June 2026 edition is candid about where things broke. On June 4, between 17:30 and 18:55 UTC, Copilot code review failed for an average of 81.6% of requests, peaking at 93.9%, with roughly 36,800 review requests failing before the incident cleared. GitHub traced the cause to a newly released dependency that its code review processing workflow pulled automatically, without a compatibility check, and that broke on the runtime it shipped into (GitHub Blog, 2026).
Four days later, on June 8, roughly 17% of unauthenticated user requests failed after a spike in automated traffic that GitHub's systems treated as abusive load, according to reporting on the same report (Blockchain.News, 2026). Then on June 10, between 15:05 and 16:25 UTC, sporadic authentication failures hit about 9% of API requests across both REST and GraphQL, producing the "logged out" behavior developers reported that day. GitHub identified the cause as a memcached proxy rollout to its internal API infrastructure that caused the authentication service to load an incorrect host configuration (GitHub Blog, 2026).
Three incidents in a week is a pattern, not a blip. Two of the three traced back to a common root: automated dependency handling and infrastructure rollouts that outran the validation built to catch them. GitHub's own report also noted it paused a planned traffic ramp for about a month following a separate stability incident on May 21, because the environment was not ready to absorb more load. That is an operator admitting its platform engineering had not caught up with demand, in writing.
For any team running CI/CD through GitHub Actions or leaning on Copilot code review as part of its pull request workflow, June was a month to build in retry logic and treat GitHub's own status page as a dependency worth monitoring, not an assumption.
Google Cloud's Egress Prices Doubled, and June Was the First Full Month of Bills
Separately, Google Cloud's May 1, 2026 price change to CDN Interconnect, Direct Peering and Carrier Peering data transfer out reached customers' full billing cycles in June. The list price for data transfer out in North America moved from $0.04 to $0.08 per GB, with smaller but real increases in Europe ($0.05 to $0.08) and Asia ($0.06 to $0.085). Google's own pricing update page describes the change as an alignment of price with the value and performance of its network investments and confirms the increase does not touch Dedicated Interconnect or Partner Interconnect pricing (Google Cloud, 2026).
Coverage from cloud cost analysts filled in the practical detail Google's own page left out: this is roughly a doubling for workloads that route through CDN Interconnect or peering rather than dedicated circuits, and it lands hardest on companies serving large volumes of cached or streamed content out of Google's network (Akave, 2026; Spendark, 2026). Neither Dedicated Interconnect users nor customers on AWS or Azure's equivalent private connections are affected by this specific change, so the immediate financial impact is real but narrow to a particular architecture choice.
What Connects Them
Both stories are downstream of the same forcing function: the infrastructure buildout hyperscalers are running to support AI workloads. Google's network investments are the stated justification for the egress price change. GitHub's own incident writeups point to Copilot-driven load and rushed infrastructure changes as the direct cause of at least two of June's outages. Platform engineering teams are being asked to absorb both sides of that buildout at once, reliability risk from providers moving fast to keep up with demand, and cost increases from providers pricing that capacity into their bills.
Neither trend is confined to GitHub or Google. It is a reasonable proxy for what platform teams elsewhere are watching for in their own vendors this year: dependency and rollout discipline on one side, network and compute pricing on the other.
What to Watch Next
GitHub's July 2026 availability report will show whether the fixes it committed to, pinning dependency versions and adding compatibility checks before the code review workflow consumes a new release, actually reduced incident count. On the pricing side, watch whether AWS or Azure follow Google with their own peering or interconnect price adjustments in the second half of 2026. If one hyperscaler's network cost increase stays isolated, it is a Google-specific decision. If a second follows within a quarter, it is a pricing trend platform teams need to plan around.
Sources: GitHub availability report: June 2026 — GitHub Blog · GitHub Reports 6 Service Incidents in June 2026 — Blockchain.News · CDN Interconnect and Peering pricing update — Google Cloud · GCP Egress Rates Double May 1 — Akave · GCP Pricing Changes in 2026 — Spendark