Tigris vs Cloudflare R2 vs Backblaze B2: Object Storage Compared in 2026

Object storage looks like a boring commodity until you run the numbers on your monthly S3 bill. For a media-heavy app (user uploads, video, images, PDFs, backups), storage and egress will quietly consume 20 to 45% of your AWS spend by the time you hit product-market fit. That is often a bigger line item than compute, and it is the first place I look when I audit a startup's cloud costs.

The specific line items that kill budgets:

• Egress fees: S3 charges $0.09 per GB egress to the internet from most US regions. For an app serving 50TB of media per month, that is $4,500 in bandwidth alone.
• Request costs: GET, PUT, and LIST requests each have a per-thousand price. A high-traffic app can easily rack up $200 to $800 per month in request fees on top of storage and egress.
• Cross-region replication: If you replicate your buckets across regions for redundancy, you pay for both storage and cross-region transfer.

The three providers covered in this article (Tigris, Cloudflare R2, and Backblaze B2) all price aggressively against S3, but they each do it differently. Picking the wrong one for your workload can cost you an extra $2K to $15K per month at scale. Our cloud cost reduction guide covers the broader optimization playbook, but object storage deserves its own deep dive.

Cloudflare R2 is the headline-grabbing option because of one killer feature: zero egress fees. You pay for storage ($0.015 per GB per month in 2026) and requests ($4.50 per million Class A operations, $0.36 per million Class B operations). Every byte you serve out of R2 is free, no matter how much traffic you push.

That changes the math dramatically for apps where egress dominates the bill. Consider a photo-sharing app storing 10TB with 200TB of egress per month. On S3: $230 for storage + $18,000 for egress = $18,230 per month. On R2: $150 for storage + $0 for egress = $150 per month. That is a 121x savings. The request costs are slightly higher than S3 for write-heavy workloads, but for read-dominated apps, the total cost difference is almost absurd.

The tradeoffs with R2:

• Single-region storage by default: Your bucket lives in one Cloudflare region. You can enable multi-region replication at an additional cost, but the default setup is less geographically redundant than S3.
• API compatibility: R2 implements the S3 API but not all of it. Most common operations work, but advanced features (object tagging filters, inventory, some lifecycle policies) may not match. Test your workload.
• Integration with Cloudflare ecosystem: R2 shines if you use Cloudflare Workers, Cloudflare Images, or Stream. You can serve objects directly to Workers without egress penalties, which enables cool architectures that are expensive on AWS.

R2 is the default choice for any app where most of your spend is egress and you are already using Cloudflare for DNS, CDN, or Workers. If you are still on Route 53 and CloudFront, migration is trickier but often still worth it. Our CDN strategy guide goes deeper on Cloudflare's broader stack.

Backblaze B2 is the old reliable in the budget storage category. Pricing in 2026: $0.006 per GB per month for storage (cheaper than R2), $0.01 per GB for egress to the internet (dramatically cheaper than S3 but not free), and generous free allowances for requests. They also have the Bandwidth Alliance deal with Cloudflare, meaning B2 to Cloudflare egress is free.

That last point is the clever architecture play. You can store data in B2 at rock-bottom prices and serve it through Cloudflare CDN with zero egress fees from B2, plus zero egress fees from Cloudflare to end users. This stack beats R2 on storage cost for very large datasets, while matching R2 on egress cost for Cloudflare-fronted traffic.

B2 is great for:

• Backups and cold storage: The low storage cost makes it ideal for archives, database backups, and anything you touch infrequently.
• Large file libraries: If you store tens of terabytes of static content that is served through a CDN, B2 is the cheapest option.
• Video and media workflows: Combined with Cloudflare Stream or your own CDN, B2 is the budget champion for media-heavy apps.

Where B2 struggles: latency and geographic distribution. B2 has fewer regions than R2 or Tigris, and first-byte latency is higher than the premium options. If you are serving users globally and do not want to rely on a CDN in front, B2 feels slower. For most production workloads, that is a non-issue because you always want a CDN in front of object storage anyway. But if you are building a low-latency multi-region app without a CDN, B2 is not the right choice.

One underrated B2 feature: their S3-compatible API is genuinely compatible. Almost every SDK and tool that works with S3 works with B2 out of the box, which reduces migration friction.

Tigris Data is the 2026 entrant that changes the conversation. Built on top of FoundationDB and distributed across dozens of regions, Tigris offers a globally distributed object store where every bucket is automatically replicated to multiple regions and read/write requests are served from the closest one. The pitch is "S3 compatibility plus global latency plus reasonable pricing."

Tigris pricing in 2026: $0.020 per GB per month storage, $0.02 per GB egress (though this is heavily discounted for traffic that stays inside the Fly.io network), and simple request pricing. On raw cost it sits between R2 and B2: more expensive than both per GB but cheaper than S3 on egress for most workloads.

What Tigris does uniquely well:

• Built-in global replication: Writes go to the closest region and propagate asynchronously. Reads are served from the closest region. This eliminates a huge class of latency problems for global apps.
• Fly.io integration: If you run on Fly.io (a growing choice in 2026), Tigris is the object store designed for that stack. Traffic between Fly Machines and Tigris is free, and latency is sub-10ms within the same region.
• Strong consistency within a region: Read-after-write consistency is guaranteed within a region, even across multi-part uploads.

The downsides: Tigris is the youngest of the three, has a smaller ecosystem, and less battle-tested operational history. I have personally run production workloads on Tigris for 14 months without any incidents, which is encouraging but not the same as S3's track record. For mission-critical applications where downtime of the storage layer is catastrophic, R2 or B2 are still safer bets.

Tigris is the right pick if you are building on Fly.io, have a globally distributed user base, need low-latency reads from multiple regions, and are comfortable with a newer provider.

Price comparisons are only meaningful at your specific traffic profile. Here are the three workload shapes I see most often and what each provider costs at scale:

Workload A: SaaS with user uploads (5TB stored, 20TB egress/month).

• S3: $115 storage + $1,800 egress = $1,915/month
• Cloudflare R2: $75 storage + $0 egress = $75/month
• Backblaze B2: $30 storage + $200 egress (or $0 through Cloudflare) = $30 to $230/month
• Tigris: $100 storage + $400 egress = $500/month

Workload B: Media-heavy app (50TB stored, 500TB egress/month).

• S3: $1,150 storage + $45,000 egress = $46,150/month
• Cloudflare R2: $750 storage + $0 egress = $750/month
• Backblaze B2: $300 storage + $0 egress through Cloudflare = $300/month
• Tigris: $1,000 storage + $10,000 egress = $11,000/month

Workload C: Backup and archive (200TB stored, 1TB egress/month).

• S3: $4,600 storage (standard) or $920 (Glacier) + $90 egress = $1,010 to $4,690/month
• Cloudflare R2: $3,000 storage + $0 egress = $3,000/month
• Backblaze B2: $1,200 storage + $10 egress = $1,210/month
• Tigris: $4,000 storage + $20 egress = $4,020/month

The takeaway: for high-egress workloads, R2 is the clear winner. For cold storage, B2 wins. For globally distributed apps on modern stacks, Tigris is worth paying extra for the latency benefit. None of these workloads make S3 a smart default anymore.

Migrating object storage sounds scary but is usually one of the easiest infrastructure changes a startup can make. All three providers implement the S3 API, which means rclone and s3cmd can copy data between providers without any code changes on your application side.

The practical migration playbook I use for clients:

• Step 1: Provision a bucket on the new provider and configure credentials.
• Step 2: Use rclone (with parallel transfers) to copy historical data in the background. Expect 10 to 50 Gbps throughput on a reasonable VM. A 20TB migration takes 1 to 4 hours of wall clock time.
• Step 3: Configure your app to dual-write to both old and new buckets during a transition window of a few days. This catches any files that change during the migration.
• Step 4: Run a reconciliation check to confirm byte equality on all objects.
• Step 5: Flip reads to the new bucket and monitor error rates. Keep the old bucket as a fallback for at least 7 days.
• Step 6: Decommission the old bucket.

The one real migration risk is presigned URLs. If you have long-lived presigned URLs distributed to users or cached in a CDN, flipping providers will break those URLs because the signature is provider-specific. Plan to invalidate or regenerate them as part of the cutover. For most apps, presigned URLs live for minutes or hours, so this is a minor issue.

Lock-in is real on two fronts. First, bucket-level features diverge. Lifecycle policies, event notifications, and advanced access controls are slightly different on each provider. If you lean heavily on these, document your dependency before migrating. Second, IAM models differ. R2 uses Cloudflare API tokens. B2 uses its own application keys. Tigris uses access keys compatible with AWS signature v4. None of these are drop-in replacements for AWS IAM. Budget a week of engineering to update your permission infrastructure.

Here is the final decision tree I use with clients in 2026.

• High-egress app (SaaS with uploads, media platform, image gallery): Cloudflare R2. The free egress alone justifies the slightly higher storage cost, and the Cloudflare ecosystem is a massive bonus.
• Cold storage and backups: Backblaze B2. Nothing beats their storage price for data you rarely touch, and the Cloudflare bandwidth alliance makes egress cheap when you need it.
• Globally distributed reads with strict latency: Tigris. The multi-region architecture is genuinely differentiated and valuable.
• Fly.io-native applications: Tigris. The integration with Fly Machines is seamless and saves you infrastructure glue.
• Already deep in AWS, need 100% S3 feature parity: Stay on S3 or move to R2 for specific high-egress buckets only.
• Maximum redundancy with multi-cloud: Combine two providers (R2 primary, B2 backup) and replicate between them with rclone on a schedule.

One last note: the savings from moving off S3 to any of these providers are real, fast, and lasting. I have seen clients cut $8K per month into $800 per month after moving a single media bucket. The migration takes a week of effort. It is one of the highest ROI infrastructure changes you can make as a startup, and most teams wait way too long to do it.

If you want a second opinion on which provider fits your specific access patterns, traffic profile, or CDN setup, Book a free strategy call and we will walk through the numbers for your real workload.