15. Competitive battlecards

Position on fit, never on their engineering. Every competitor here is a good product for someone. Find the workload where they are the wrong answer.

Card format: When they win / When we win / Landmines to plant / Landmines to avoid / Objections / Proof.

AWS S3

When they win: deep AWS-native apps using IAM roles, S3 event to Lambda, Athena, object tagging, S3 Express One Zone latency; customers burning down an EDP commit; data that is genuinely cold (Glacier Deep Archive at ~$0.99/TB/mo).

When we win: egress-heavy workloads (restores, media reuse, multi-cloud AI, CDN origin to non-AWS CDN); teams tired of tiering and lifecycle tuning; COGS-sensitive SaaS.

Landmines to plant (questions):

• “What percent of last month’s S3 bill was data transfer?”
• “Have you paid early-delete fees on IA or Glacier?”
• “How long would a full Glacier restore take, and what would retrieval cost?”
• “How many storage classes do you manage, and who tunes the lifecycle rules?”

Landmines to avoid:

• Do not claim feature parity. Gaps: IAM roles, object tagging, object ACLs, website hosting, browser POST.
• Do not compare to Glacier on storage price.
• Do not forget EDP discounts; ask for effective rates.

Objections:

• “We are all-in on AWS.” → “Keep compute there. Put the copy that must survive an AWS account compromise somewhere else. That is also the cheapest copy to restore from.”
• “S3 has 11 nines too.” → “Yes. Durability is table stakes. Differences are egress, tiers, and who can delete your backups.”

Proof: Tribute ($15K/month saved, “drop-in API replacement”); Hume AI consolidated from multiple hyperscale object stores; Cal Poly Humboldt shrank Google footprint.

Wasabi

When they win: flat-rate buyers with egress reliably below stored volume; existing Wasabi reseller relationships; customers who want the “no egress” headline.

When we win: reuse-heavy data (egress 1x to 3x stored), short-retention data (no minimum duration on B2), AI workloads and GPU-cloud partnerships, CDN partner egress, Overdrive needs; public-company transparency; Drive Stats credibility.

Plant:

• “What does Wasabi’s fair-use policy say if your monthly egress exceeds your stored volume?”
• “Do you delete data before 90 days? What is billed when you do?” [verify Wasabi’s current minimum storage duration policy]
• “Your price went from $6.99 to $7.99 this year. How do you budget the next increase?”

Avoid: calling Wasabi unreliable or making claims about their finances or outages without a cited source.

Objections: “Wasabi has no egress fees at all.” → “As long as egress stays within their fair-use terms. B2 gives 3x stored free and unlimited to partners, with a published $0.01/GB if you go over. Predictable both ways.”

Math: 1 PB: B2 $6,950 vs Wasabi $7,990 per month = $12,480/yr difference at list.

Cloudflare R2

When they win: extremely egress-heavy internet delivery (above ~4x stored per month) without a partner CDN; teams already deep in Workers; small datasets where storage cost is trivial.

When we win: storage-dominant workloads (R2 is $15/TB, B2 $6.95); write-heavy or small-object workloads (R2 Class A $4.50/M); backup (Object Lock maturity and backup ISV certifications); you can use Cloudflare CDN with B2 and get $0 egress anyway.

Plant:

• “How many PUTs and LISTs a month? R2 bills them.”
• “Do you need Object Lock compliance mode for backups, and does your backup vendor certify R2?” [verify R2 Object Lock status at time of call]
• “If you are on Cloudflare already, why not keep the CDN and put a cheaper origin behind it?”

Avoid: saying R2 is not durable or not enterprise. Stay on economics and fit.

Math: 1 PB, 10% egress: B2 $6,950 vs R2 $15,004/month. Writes: 100M PUTs/month on R2 = $450.

Best line: “Cloudflare is our partner. B2 behind Cloudflare gives you R2’s egress story at less than half R2’s storage price.”

Azure Blob Storage

When they win: Microsoft-centric shops with MACC commitments, Entra ID integration, Azure-native analytics, Archive tier for truly cold data.

When we win: off-Azure backup copy (Veeam, M365 backup to a non-Microsoft target for separation), egress-heavy use, simplicity (Hot/Cool/Cold/Archive tiers with minimum durations and retrieval fees vs one class).

Plant: “If your Microsoft tenant is compromised, is your backup in the same blast radius?” “What do Cool and Archive early-deletion and retrieval fees cost you per year?”

Numbers: Hot LRS first 50 TB ~$20.80/TB/mo (sample table) vs B2 $6.95. [verify current Azure list]

Google Cloud Storage

When they win: BigQuery/Vertex-centric data platforms, GCP credits, dual-region and turbo replication needs.

When we win: training on GPUs outside GCP (egress), media archives, backup copies outside Google.

Plant: “Where are your GPUs actually coming from this quarter?” “What does pulling the dataset to that provider cost?”

Numbers: Standard regional ~$20/TB/mo vs B2 $6.95; Nearline 30-day, Coldline 90-day, Archive 365-day minimums (GCS pricing page).

On-prem: Pure Storage (FlashBlade), NetApp (StorageGRID / ONTAP S3), Dell ECS/ObjectScale

When they win: latency-critical primary storage, air-gapped or sovereign sites, existing large installed base, capex-friendly budgets, customers who need sub-ms or file protocols.

When we win: off-site copy (on-prem cannot be its own off-site), capacity growth without refresh cycles, archive tier behind flash, budget moving from capex to opex, staff constraints.

Plant: “When is the next refresh and what is the quote?” “Who manages drive failures and firmware?” “Where is your off-site immutable copy?”

Avoid: pitching B2 as primary storage for databases or VMs. Pitch it as the tier behind them. NetApp, Pure, and Dell all support tiering or replication to S3 targets; validate the specific product’s support for B2 before claiming. [verify per-product B2 certification]

Line: “Keep the flash. Stop buying disk shelves for data you touch twice a year.”

MinIO / Ceph (self-built object)

When they win: strict on-prem or edge requirements, Kubernetes-native S3 inside a data center, teams with storage engineers who want control.

When we win: when the team counts the real cost: hardware, 3x or erasure-coded overhead, drive replacement, upgrades, on-call, and a second site for off-site copies. For neoclouds, B2 Neo’s whole pitch is “do not build this.”

Plant: “How many people are on call for the cluster?” “What happens when a 20 TB drive fails during a rebuild?” “Where is the off-site copy?”

Math prompt: self_built_$/TB/mo = (hardware/36 + power + space + support + FTE×salary/12) / usable_TB. Fill with their numbers. Include usable vs raw overhead.

Quick comparison matrix (list, checked 2026-10-14)