Cost + workload fit
Same Remotion job: ~$0.00017 on Mac power vs ~$0.04 on a tiny cloud container.
For a measured 3-minute kinetic Remotion job, M2 Max package electricity was about ~$0.00017 while Cloudflare Containers standard-1 rate-card math landed near ~$0.0435 for 2175 s of box time (~36 minutes for a 3-minute video). Cost and wall-clock both matter. Cloud still wins on ops elasticity and Lambda-style fan-out.
Headline cost and time (2026-07-22 bakeoff)
Mac dollars below are package electricity (ABS Wh), not a RenderMac invoice. Cloud dollars are published Workers Paid rate-card estimates for standard-1 (~$0.00002 per wall-second upper bound), not a Cloudflare invoice and not RenderMac buyer pricing.
| Workload | Best Mac wall | Best cloud / CF-shape wall | Mac elec (ABS) | Cloud $ (rate-card) | Prefer |
|---|---|---|---|---|---|
| Remotion Light 5 s | 3.5 s M2 Max | 54.1 s CF live | ~$0.000003 | ~$0.0011 | Apple Silicon |
| Remotion 3 min kinetic | 111.5 s M2 Max / 134–146 s M1 | 2175 s CF live | ~$0.00017 | ~$0.0435 live | Apple Silicon |
| Remotion 20 min kinetic | 679.5 s M2 Max | ~14502 s PROJECTED CF | ~$0.0010 | ~$0.29 PROJECTED | Apple Silicon |
| Encode 20 min VT | 95 s M2 Max | N/A on CF Linux | n/a (no power sample) | — | Apple Silicon |
| Encode 20 min x264 | 103 s M2 Max | 2371 s Docker CF-shape | n/a | ~$0.0474 | Apple Silicon for speed; containers for ops |
| Massive parallel Lambda-style fan-out | single Mac not comparable | many functions (not run here) | — | — | Distributed cloud if you need that shape |
Source: measured 2026-07-22 bakeoffs. CF $ = wall_s × (4 × $0.0000025 + 0.5 × $0.000020) ≈ wall_s × $0.00002 for standard-1. Mac elec = ABS package Wh × $0.1883/kWh. LongKinetic20m CF cells are PROJECTED (×20/3 from measured 3-minute CF live). Method: benchmark methodology. Full compute table: CF Remotion cost. RenderMac buyer pricing stays quote-first.
Choose Apple Silicon when…
- Per-job container rate-card dollars and wall-clock both hurt (Remotion on tiny 0.5 vCPU shapes).
- You need VideoToolbox or other Mac-native codec paths Linux containers cannot offer.
- You want organization-private pools and macOS execution parity.
Choose cloud containers when…
- Ops elasticity and region sprawl matter more than per-job wall-clock or rate-card cents.
- The workload is pure Linux software encode with no Mac-only API.
- You are already committed to Lambda/Cloud Run style fan-out and will shard aggressively.