Cloud Run fully managed or Cloud Functions 2nd gen for infrequent HTTP traffic?

Direct answer

For an infrequently invoked HTTP endpoint with identical duration and memory, Cloud Run fully managed is typically the safer lower-cost choice at very low invocation counts. Both services often fall inside their monthly free tiers, and because Cloud Functions 2nd generation runs on Cloud Run infrastructure, it tends to match or exceed Cloud Run's cost for the same container rather than undercut it. Treat all rates, free-tier limits, and defaults as unverified until you check current Google Cloud pricing for your region.

How concurrency shifts the break-even

Cloud Run exposes an explicit concurrency setting, often defaulting around 80 and capping near 1000, so a burst can reuse warm instances and reduce billable instance-seconds and cold starts. Confirm the current value for your region. Cloud Functions 2nd generation uses per-instance concurrency, and whether it can exceed one concurrent request depends on version and trigger type, so verify the limit before assuming the same burst behavior. Higher concurrency in Cloud Run generally lowers cost per request under bursty traffic; near-serial traffic keeps both services effectively tied inside free tiers.

Baseline charges despite scale-to-zero

Unexpected baseline charges usually come from minimum instances, startup-CPU boosts, provisioned concurrency, or a fronting load balancer or API gateway, not from invocations themselves. Large HTTP responses can make outbound egress the dominant cost, and egress is billed separately on both services outside any compute free tier.

Verification steps

1. Open the Google Cloud Pricing calculator for your target region and model identical requests per month, duration, memory, concurrency, and egress for both services.
2. Deploy a minimal endpoint to Cloud Run and Cloud Functions 2nd gen with scale-to-zero and no minimum instances, send a known request volume, and compare the cost breakdown in Cloud Billing reports after one day.
3. Confirm current settings with gcloud run services describe SERVICE --region=REGION and gcloud functions describe FUNCTION --region=REGION, checking concurrency, min/max instances, CPU allocation, and startup-CPU behavior.

Missing diagnostic

What is your peak concurrent requests per second and average response size? Concurrent bursts or payload-heavy responses favor Cloud Run because concurrency control reduces instance count and egress becomes the main cost. Without those numbers, the recommendation stays conservative.

Assumption: identical container image, runtime duration, and memory allocation on both services. Regional SKUs and quotas can move the break-even independently of invocation count.