Jul 2026 cost
—
BigQuery only
Dec 2030 cost
—
BQ + Archive
2026–2030 total
—
all tiers combined
Saves vs all-in-BigQuery
—
vs keeping all data in BQ forever
Annual cost — 2026 to 2030
BigQuery
GCS Archive
Storage lifecycle phases
Jul '26Dec '26Dec '27Dec '28Dec '29Dec '30
Phase 1 — BigQuery only (M1–6)
All data lands in BigQuery Capacitor format. The BQ window fills month by month. Archive cost is zero — no GCS data yet.
Phase 2 — BQ flat + Archive grows (M7–72)
BQ stays at its fixed window size. Archive receives one month of aged-out BQ data every month and grows permanently — it never shrinks.
Monthly cost over 72 months
Tiering saves (5yr)
—
vs keeping all data in BigQuery
BigQuery cost (5yr)
—
BigQuery tier (5yr)
Archive cost (5yr)
—
cumulative GCS Archive spend
Month-by-month breakdown (Jul 2026 – Dec 2030)
Why BQ → Archive? — storage pricing reference
BigQuery active storage can be cheaper than GCS Standard per raw TB due to superior Capacitor compression (~5× vs ~3.5×). The same price per GB stored ($0.020) buys more raw-TB coverage in BQ. However, for cold data GCS Archive at $0.0012/GB is ~8× cheaper than BQ long-term — making BQ → Archive the cost-optimal two-tier strategy for Telco CDR workloads.
| Storage option | Compression | $/GB stored/mo | Cost per raw TB/mo | Best for |
|---|---|---|---|---|
| BigQuery active <90 days | 5× Capacitor | $0.020 | $4.10 | Hot analytics, real-time SQL |
| BigQuery long-term ≥90 days | 5× Capacitor | $0.010 | $2.05 | Infrequent BQ queries |
| GCS Standard | 3.5× Parquet | $0.020 | $5.85 | Warm, pipeline staging — skipped in this model |
| GCS Nearline | 3.5× Parquet | $0.010 | $2.93 | 30-day minimum storage |
| GCS Coldline | 3.5× Parquet | $0.004 | $1.17 | 90-day minimum storage |
| GCS Archive ← used in this model | 3.5× Parquet | $0.0012 | $0.35 | Cold, 365-day minimum storage |
BigLake option: Store historical data in GCS Archive as Parquet and register it as a BigQuery external table (BigLake). Query via standard BQ SQL at Archive prices ($0.35/raw TB/mo) — no export or data movement needed. Best when historical analytics is occasional rather than daily.