Back to Engineering Insights
Cloud Cost Optimization
May 21, 2026
By Ravi Kanani

Cold Storage Showdown 2026: S3 Glacier vs Google Archive vs Azure Archive vs Wasabi vs B2 (Decision Framework)

Cold Storage Showdown 2026: S3 Glacier vs Google Archive vs Azure Archive vs Wasabi vs B2 (Decision Framework)
Key Takeaway

S3 Glacier Deep Archive ($1/TB/month) is cheapest to store but expensive to retrieve ($20/TB + 12-hour wait). Google Cloud Archive ($1.20/TB/month) and Azure Archive ($0.99/TB/month) are similar to Glacier with comparable retrieval costs. Wasabi ($6.99/TB/month) and Backblaze B2 ($6/TB/month) cost 6-7x more to store but have no retrieval fees and millisecond access. For compliance archives accessed less than once per year, Glacier wins. For warm archives accessed quarterly, Wasabi or B2 win on total cost. Picking by storage price alone leads to 5-20x overpayment on the eventual restore.

We Stored 12 Petabytes Across 5 Cold Storage Tiers. Most Teams Pick The Wrong One.

A growth-stage media company we worked with in early 2026 had 800TB of compliance video archives sitting in S3 Standard at $18,400/month. Their compliance team had told engineering "this data must be retained 7 years." Engineering interpreted that as "this data must be available." It wasn't.

We audited the access patterns. Last 90 days: zero retrievals. Last 12 months: 8 retrievals totaling 600GB. The data needed to be retained, not accessible.

We modeled migration to four cold storage options:

  • S3 Standard (current): $18,400/month
  • S3 Intelligent-Tiering with Archive: $1,800/month (89% savings)
  • S3 Glacier Deep Archive: $792/month (96% savings)
  • Azure Archive Storage: $792/month (96% savings)

After 6 weeks of migration to Glacier Deep Archive, monthly bill dropped to $792/month. Annual savings: $211,000. Over the 7-year retention period: $1.48M total savings.

But here's the twist: a different client we worked with had moved their DR backups to Glacier Deep Archive without analyzing access patterns. When a regional outage required them to restore 50TB from DR, they discovered:

  • Restore took 14 hours (Deep Archive minimum)
  • Retrieval fees: $1,000
  • Egress fees: $4,500
  • Business impact during 14-hour wait: $180,000 in lost revenue

The "cheap" cold storage cost them $185,500 in a single incident because the workload required fast access, not minimum cost.

This pattern is consistent across 32 cold storage audits we ran in 2025-2026: the cheapest storage tier is almost never the cheapest total cost. Picking by per-GB-month price ignores retrieval fees, minimum durations, and business impact of access delay.

This post is the workload-to-tier decision framework: which cold storage option wins for which use case, what each one actually costs at total-cost-of-ownership level, and the migration playbook that doesn't end in disaster.

If your compliance archive is on S3 Standard or your DR backup is on Glacier Deep Archive, you're almost certainly on the wrong tier.


The 5 Cold Storage Tiers That Matter in 2026

TierStorage $/TB/moRetrieval $/TBMin WaitMin Retention
S3 Glacier Deep Archive$0.99$20 (standard)12 hours180 days
S3 Glacier Flexible Retrieval$3.60$0 std + $30 expedited/TB3-5 hrs std, 1-5 min expedited90 days
S3 Glacier Instant Retrieval$4.00$10Milliseconds90 days
Google Cloud Archive$1.20$50Milliseconds365 days
Azure Archive Storage$0.99$201-15 hours180 days
Wasabi$6.99$0Milliseconds90 days
Backblaze B2$6.00$10 ($0 via Cloudflare)MillisecondsNone*

*Backblaze B2 has minimum retention only on Bandwidth Alliance free egress, not on storage itself.

The storage price spread is 7x ($0.99 to $6.99). The retrieval price spread is infinite ($0 to $50/TB). The access time spread is massive (milliseconds to 12 hours). Most teams optimize the small spread (storage) and ignore the large spreads (retrieval and access time).


Why Storage Price Is The Wrong Optimization Target

The biggest mistake we see: picking cold storage based on per-GB-month price. The math goes wrong for three reasons.

Reason 1: Retrieval Fees Compound Surprisingly Fast

Storage cost is paid monthly. Retrieval cost is paid once per restore — but a single restore can equal a year of storage cost.

Example: 100TB stored on Glacier Deep Archive

  • Storage cost: 100 × $0.99 × 12 = $1,188/year
  • One full restore: 100 × $20 = $2,000 (retrieval) + $9,000 (egress) = $11,000
  • One restore costs 9.3x annual storage

If you might restore once per year, Glacier total cost is $12,188/year. Wasabi at $6.99/TB has $8,388/year storage and $0 retrieval — cheaper in total cost despite 7x more expensive per GB.

Reason 2: Business Impact of Access Delay Is Often Larger Than Storage Cost

The 12-hour minimum restore on Glacier Deep Archive is not a minor inconvenience. For business-critical DR scenarios, 12 hours of unavailability can cost millions. Storage cost savings of $200/month (Glacier vs Wasabi for 100TB) are dwarfed by even one hour of business impact.

Reason 3: Minimum Retention Charges On Early Deletion

If you delete data from Glacier Deep Archive before 180 days, you still pay 180 days of storage. Same for Azure Archive. For workflows with short-lived archives (e.g., temporary compliance holds, processed datasets), minimum retention can make "cheap" tiers expensive.


Real-World Cost Modeling: Three Workload Scenarios

We modeled three actual workload profiles. May 2026 pricing.

Workload A: Compliance Video Archive (7-Year Retention, Rare Access)

A media company storing 500TB of broadcast footage for FCC compliance:

  • Retention required: 7 years
  • Expected retrievals: 1-5 per year, ~10TB each
  • Compliance audit deadline tolerance: 24+ hours acceptable

S3 Glacier Deep Archive:

  • Storage (7 years × 12 months × 500TB × $0.99): $41,580 total storage cost
  • Retrievals (5 × 10TB × ($20 + $90 egress)): $5,500 total
  • 7-year TCO: $47,080

Azure Archive Storage:

  • Storage: $41,580 (same per-TB price as Glacier)
  • Retrieval: $5,500
  • 7-year TCO: $47,080 (essentially identical to Glacier)

Wasabi:

  • Storage (7 years × 12 × 500 × $6.99): $293,580
  • Retrieval: $0
  • 7-year TCO: $293,580 (6.2x more expensive)

Backblaze B2:

  • Storage (7 years × 12 × 500 × $6.00): $252,000
  • Retrieval: $0 (via Cloudflare Bandwidth Alliance)
  • 7-year TCO: $252,000 (5.4x more expensive)

Verdict: Glacier Deep Archive (or Azure Archive) wins decisively at $47K vs $250K+. For pure compliance with rare access and 24+ hour tolerance, Glacier Deep Archive is the right answer. The 12-hour restore tolerance is acceptable for compliance audits.

Workload B: DR Backups (Quarterly Test, Business-Critical)

A SaaS company with 200TB of DR backups:

  • Quarterly DR drill restores 5TB
  • Annual full restore test 50TB
  • True disaster: must restore 200TB within 4 hours (RPO/RTO commitment)

S3 Glacier Deep Archive:

  • Storage: 200 × $0.99 × 12 = $2,376/year
  • Quarterly restore + annual: 4 × 5TB + 50TB = 70TB at $20/TB = $1,400 + $6,300 egress = $7,700/year
  • Disaster scenario: 200TB × ($20 + $90) = $22,000 + 12-hour wait that violates RTO
  • Annual cost: $10,076 + violated RTO = unacceptable

S3 Glacier Flexible Retrieval:

  • Storage: 200 × $3.60 × 12 = $8,640/year
  • Restores at $0 standard or expedited: 70TB × $30 expedited = $2,100 + $6,300 egress
  • Disaster scenario: 200TB expedited = $6,000 + $18,000 egress + 5-min wait = meets RTO
  • Annual cost: $17,040 + meets RTO

Wasabi:

  • Storage: 200 × $6.99 × 12 = $16,776/year
  • Restores: $0 retrieval + $0 egress (Wasabi includes free egress)
  • Disaster scenario: $0 + meets RTO instantly
  • Annual cost: $16,776 + best RTO

Backblaze B2:

  • Storage: 200 × $6.00 × 12 = $14,400/year
  • Restores: free via Cloudflare Bandwidth Alliance
  • Disaster scenario: $0 + meets RTO instantly
  • Annual cost: $14,400 + best RTO

Verdict: Backblaze B2 at $14,400/year wins on cost AND meets RTO. Glacier Deep Archive at $10K/year saves $4K but violates 4-hour RTO. For DR backups with strict RTO, B2 is the cost-optimal answer. The "cheaper" Glacier tier is wrong for this workload.

Workload C: Mixed Compliance + DR (Combined Workload)

A healthcare company with 2PB total cold data:

  • 1.5PB pure compliance (HIPAA records, 6-year retention, almost never accessed)
  • 500TB DR backups (must restore within 8 hours)

Naive single-tier strategy (everything on Glacier Deep Archive):

  • 2PB storage: $23,760/year
  • DR restore violations potential: catastrophic
  • Wrong architecture

Optimal multi-tier strategy:

  • 1.5PB compliance on Glacier Deep Archive: $17,820/year
  • 500TB DR on Glacier Flexible Retrieval (meets 8-hour RTO with standard restore): $21,600/year
  • DR restore costs (annual test): 25TB × $0 std restore + $2,250 egress = $2,250
  • Annual cost: $41,670 + meets all SLAs

Verdict: Multi-tier strategy combines Glacier Deep Archive (compliance) + Glacier Flexible Retrieval (DR) for $41K vs single-tier Wasabi at ~$170K/year. Use multi-tier when access requirements differ within your data.


The Decision Framework: 5 Questions

Question 1: How often will you actually access the data?

  • Almost never (less than once per year): Glacier Deep Archive or Azure Archive (storage cost dominates)
  • Quarterly: Glacier Flexible Retrieval (balance) or B2 (no retrieval fee complexity)
  • Monthly: Glacier Instant Retrieval, Wasabi, or B2 (frequent retrieval makes Glacier expensive)
  • Daily or unpredictable: Wasabi or B2 (Glacier becomes wildly expensive)

Question 2: What is your maximum acceptable restore wait time?

  • Days (compliance audits): Glacier Deep Archive, Azure Archive (cheapest)
  • Hours (DR drills): Glacier Flexible Retrieval, Glacier Instant, Wasabi, B2
  • Minutes (production fail-back): Glacier Instant, Wasabi, B2
  • Milliseconds (live access): Wasabi, B2 (Glacier excluded)

Question 3: What is your data volume?

  • Under 10TB: Wasabi/B2 are cheap enough; don't bother with Glacier complexity
  • 10TB-1PB: Glacier becomes meaningfully cheaper if access is rare. Calculate carefully.
  • Over 1PB: Glacier Deep Archive saves enormous money on pure compliance. Multi-tier likely.

Question 4: What is the business impact of one full restore?

  • Trivial (test data, replaceable): Cheapest option works
  • Moderate (development data): Glacier Flexible Retrieval acceptable
  • High (revenue-impacting): Account for restore time AND cost
  • Critical (existential risk): Wasabi/B2 with millisecond access; do not put on Glacier

Question 5: What is your existing cloud commitment?

  • AWS-locked with EDP/SP discounts: Glacier becomes more attractive (potential discounts)
  • Azure-locked with EA: Azure Archive equivalent to Glacier
  • GCP-native: Google Cloud Archive (similar pricing)
  • Multi-cloud or no commit: Wasabi/B2 (cloud-agnostic, simpler)

When To Pick Each Tier (Cheat Sheet)

Use CaseBest TierWhy
HIPAA records, 7-year retentionGlacier Deep ArchiveCheapest storage, 24-hour RTO acceptable
Financial records, regulatory archiveGlacier Deep ArchiveCompliance use cases tolerate retrieval delay
Video surveillance, indefinite retentionGlacier Deep ArchivePure storage workload
Email/document archive (FINRA, SEC)Glacier Deep Archive or Azure ArchiveIndustry-specific compliance
DR backups with 4+ hour RTOGlacier Flexible RetrievalBalance cost and restore speed
DR backups with sub-hour RTOWasabi or B2No retrieval delay, no fees
Database backups (frequent restore tests)Wasabi or B2Frequent access makes Glacier expensive
Application logs (90+ days)Glacier Instant RetrievalBalance access and cost
Build artifacts (regenerable)B2 with One Zone-IA equivalentCheap, regenerable
Tape replacement (off-site copies)Glacier Deep Archive or B2Long-term, rare access
Genomics raw dataGlacier Deep ArchiveCheap, accessed for re-analysis
Media archives (broadcasting)B2 with CloudflareSometimes need to retrieve quickly
Customer historical data (SaaS)Glacier InstantCustomers might query old data
Pre-computed ML datasetsGlacier Flexible RetrievalSometimes regenerated, sometimes restored
Cross-region DR copiesGlacier Flexible or Wasabi multi-regionRegional failover scenarios

Hidden Costs Most Cold Storage Comparisons Miss

Hidden Cost 1: Egress Fees Compound With Retrieval Fees

When you restore from Glacier, you pay retrieval fees ($20/TB) AND egress fees if data leaves AWS ($90/TB to internet). For 100TB restore: $2,000 retrieval + $9,000 egress = $11,000. Most "Glacier vs Wasabi" comparisons forget egress.

Mitigation: If restoring within AWS (e.g., back to S3 Standard), egress is free. If restoring to non-AWS destination, factor in egress. Wasabi and B2 have no egress fees, simplifying the math.

Hidden Cost 2: Minimum Storage Duration Charges

Glacier Deep Archive requires 180 days minimum. Delete data after 30 days, you still pay for 180 days. Same for Azure Archive. For workloads with short-lived archives, this can negate the cheap storage rate.

Mitigation: Use lifecycle rules to ensure data ages naturally before transitioning to Glacier. Don't transition data that might be deleted soon.

Hidden Cost 3: Lifecycle Transition Fees

Moving data INTO Glacier from S3 Standard costs $0.05 per 1,000 objects transitioned. For buckets with millions of small objects, this can be a one-time fee in the thousands.

Mitigation: Aggregate small objects before archiving (combine 1M small files into 1K larger files). Use S3 Batch Operations for bulk transitions.

Hidden Cost 4: Multi-Part Upload Cleanup

Glacier doesn't auto-clean failed multi-part uploads. They accumulate as orphaned uploads charged at standard storage rates.

Mitigation: Enable lifecycle policy to abort incomplete multipart uploads after 7 days. Most teams skip this.

Hidden Cost 5: Wasabi's 90-Day Minimum Duration

Wasabi's 90-day minimum retention is less well-known than Glacier's 180-day minimum. Delete data after 30 days, pay for 90.

Mitigation: For high-churn archives, B2 has no minimum duration on storage (only on Bandwidth Alliance free egress).

Hidden Cost 6: Vault Lock and Object Lock For Compliance

For regulated archives requiring tamper-proof retention, S3 Object Lock and Glacier Vault Lock add complexity. Setting them up wrong can prevent legitimate access. Setting them up loosely fails compliance.

Mitigation: Use S3 Object Lock in Compliance mode for true regulatory requirements. Test the configuration before relying on it.

Hidden Cost 7: Cross-Cloud Egress For Migrations

Migrating 1PB from S3 Glacier to Wasabi requires retrieving (12 hours, $20K retrieval fee) plus egress (1PB × $90 = $90K). Same migration in reverse requires only Wasabi-side ingress (free).

Mitigation: Choose your cold storage vendor carefully upfront. Migrating away from Glacier is expensive.


Migration Playbook: S3 Standard → Glacier Deep Archive

For teams with TBs of data on S3 Standard that should be on cold storage:

Phase 1: Access Pattern Analysis (Week 1)

  1. Enable S3 Storage Lens for affected buckets
  2. Pull 90-day access patterns: which objects accessed, how often
  3. Categorize data by access frequency:
    • Hot: accessed weekly+ → stay on Standard
    • Warm: accessed monthly → S3 Intelligent-Tiering or Glacier Instant
    • Cold: accessed quarterly → Glacier Flexible Retrieval
    • Frozen: accessed yearly or less → Glacier Deep Archive

Phase 2: Lifecycle Rules (Week 2)

  1. Add S3 lifecycle rules to transition objects:
Rules:
  - Status: Enabled
    Transitions:
      - Days: 30
        StorageClass: STANDARD_IA
      - Days: 90
        StorageClass: GLACIER_IR # Instant Retrieval
      - Days: 180
        StorageClass: DEEP_ARCHIVE
  1. Apply lifecycle rules to historical data via S3 Batch Operations
  2. Monitor lifecycle transitions via CloudWatch

Phase 3: Validate (Week 3-4)

  1. Confirm objects transitioned to Deep Archive correctly
  2. Test restore process with a sample object
  3. Document restore procedures in runbook
  4. Validate compliance access controls

Phase 4: Lock In Savings (Week 5+)

  1. Calculate actual cost reduction in next billing cycle
  2. Set up monthly review of storage class breakdown
  3. Adjust lifecycle rules if access patterns change
  4. Apply same patterns to other buckets

Typical outcome: 70-95% storage cost reduction for archives.


When To Stay On Hot Storage

Don't migrate to cold storage if:

  • Annual savings under $5K: Migration effort exceeds savings
  • Access patterns are unclear or unpredictable: Cold storage retrieval fees can blow up if patterns differ from expectation
  • Data is regenerable cheaply: Better to delete and regenerate than archive
  • Compliance requires immediate access: Hot storage is the right answer regardless of cost
  • You don't have engineering capacity to test restore procedures: Untested cold storage is a disaster waiting

For about 15-20% of "cold storage candidates" we audit, staying on hot storage with lifecycle to S3 IA is the right answer.


A 30-Day Cold Storage Audit

If you have 100TB+ of S3 Standard or equivalent that might be cold-storage-eligible, run this audit.

Week 1: Inventory

  1. Pull S3 Storage Lens reports for all buckets
  2. Identify total storage by storage class
  3. Calculate current monthly storage cost
  4. Identify access patterns (last 90 days minimum)

Week 2: Categorize

For each bucket and prefix:

  • Frequency of access
  • Maximum acceptable retrieval time
  • Compliance requirements
  • Business impact of unavailability

Week 3: Cost Model

For each cold-storage-eligible workload:

  • Calculate cost on each tier (Glacier Deep Archive, Wasabi, B2, Glacier Flexible)
  • Include retrieval scenarios in TCO calculation
  • Calculate net annual savings

Week 4: Execute

  1. Apply lifecycle rules for new data
  2. Use S3 Batch Operations for historical data migration
  3. Validate restore procedures
  4. Document runbook for future restores
  5. Set up alerts for unexpected retrieval volumes

After 30 days, expect 50-90% reduction in archive storage costs.


The Bottom Line

Cold storage in 2026 is a workload-fit decision, not a price-comparison decision. Glacier Deep Archive wins for true compliance archives accessed less than yearly. Wasabi and B2 win for DR scenarios where restore time matters more than storage cost. Glacier Flexible Retrieval and Instant Retrieval fill the middle for warm archives. Picking by per-GB-month price alone leads to 5-20x overpayment on the eventual restore.

The discipline most teams skip: modeling actual retrieval cost in addition to storage cost, and testing restore procedures before relying on them. Untested cold storage is just a future disaster waiting to happen.

If your archive storage exceeds 100TB and you haven't run a tier audit in the last 12 months, you are very likely on the wrong tier for at least part of your data. Our cloud cost optimization team runs free cold storage audits and typically captures 50-90% archive cost reduction within 60 days. Run a free Cloud Waste Scorecard to find your biggest storage cost leaks first.


Further reading:

Frequently Asked Questions

Stop Overpaying for Cloud Infrastructure

Our clients save 30-60% on their cloud bill within 90 days. Get a free Cloud Waste Assessment and see exactly where your money is going.

Related Insights

Cloud Cost Optimization
10 Ways Teams Overpay On AWS Fargate in 2026 (And How To Fix Each One This Week)
May 21, 2026
10 Ways Teams Overpay On AWS Fargate in 2026 (And How To Fix Each One This Week)

AWS Fargate is the second-most-overprovisioned compute service on AWS after Lambda. We audited 64 production Fargate deployments in 2025-2026 and found the average bill was 50% higher than necessary due to 10 specific waste patterns: missed ARM/Graviton, oversized task definitions, no Spot usage, missing Compute Savings Plans, unused capacity providers, and more. This is the fix list with real cost math for each.

Cloud Cost Optimization
AWS Savings Plans vs Reserved Instances 2026: Pick Wrong, Lose 60% (Real Commitment Decision Framework)
May 21, 2026
AWS Savings Plans vs Reserved Instances 2026: Pick Wrong, Lose 60% (Real Commitment Decision Framework)

AWS offers four commitment types in 2026 (Compute Savings Plans, EC2 Instance Savings Plans, Standard Reserved Instances, Convertible Reserved Instances) plus SageMaker Savings Plans for ML workloads. We optimized 47 commitment portfolios in 2025-2026 and found teams consistently pick the wrong type, losing 40-60% in either savings or flexibility. This is the workload-to-commitment decision framework based on real production portfolios.

Cloud Cost Optimization
Cast AI vs Spot.io 2026: Automated Kubernetes Cost Tools Compared (We Saved a Client $720K/Year)
May 20, 2026
Cast AI vs Spot.io 2026: Automated Kubernetes Cost Tools Compared (We Saved a Client $720K/Year)

Cast AI and Spot.io are the two leading automated Kubernetes cost optimization platforms in 2026. We deployed both on production EKS clusters across 12 clients and found the cost gap for identical workloads averaged 40%. This is the head-to-head decision framework based on real production deployments, including pricing transparency that vendor pages obscure.