Summary

  • Backblaze sells two related forms of storage confidence: B2 Cloud Storage, where customers pay mainly for object-storage capacity and access, and Computer Backup, where individuals and businesses pay a flat subscription for endpoint backup.
  • The public financial record supports the core thesis. Backblaze can publish low storage prices only if drive purchasing, data-centre capacity, support labour, bandwidth, finance leases and customer behaviour remain tightly controlled.
  • The strongest evidence is not a marketing claim. It is the combination of the 2025 Form 10-K, the Q1 2026 Form 10-Q, current pricing pages, service terms, drive-stat disclosures and visible substitute pricing from AWS, Microsoft Azure, Google Cloud, Wasabi and Cloudflare R2.
  • The value proposition is real but conditional. B2 looks economically persuasive for workloads that value simple hot storage, S3-compatible access and predictable egress. Computer Backup looks persuasive for ordinary endpoint loss, but less complete for users who expect the word "unlimited" to cover every data pattern, synced folder, external-drive edge case or urgent multi-terabyte restore.
  • The missing proof is account-level unit economics. Public filings show revenue, cost of revenue, customer counts, retention signals, lease exposure and infrastructure risks, but they do not disclose gross margin by cohort, support tickets per terabyte, average restore volume, egress intensity or hard-drive procurement cost by generation.

The paid unit is a fear-reduction account

The customer does not start by buying a disk. The customer starts with a fear: a laptop dies, a studio loses a shoot, a software company has to retain logs, an MSP needs a storage target, an AI or media workflow needs cheap hot data, or a small business wants a copy of its data somewhere other than the office. Backblaze turns that fear into a paid account.

The paid unit is different in the two main services. In B2 Cloud Storage, the customer buys a cloud object-storage account that can be used through S3-compatible tooling, APIs, web console, command-line tools and third-party integrations. The visible tariff is capacity plus download behaviour. On the current Backblaze pricing page, pay-as-you-go B2 storage is presented at $6.95 per TB-month, with free egress up to three times average monthly data stored, and overage at $0.01 per GB. The same page also says Class A, B and C API calls are free for pay-as-you-go customers, while Class D transactions have a small charge after a daily free allowance. B2 Reserve is a committed-capacity product with bundles starting at 20 TB and annual pricing, with bundled egress and support language aimed at larger customers.

In Computer Backup, the customer buys a per-computer backup subscription. The current pricing page presents Personal Backup and Business Backup at $99 per year, with monthly and two-year options. The product language emphasizes unlimited data backup, access from anywhere, web-based restores, restore software, mailed hard-drive restore options and optional longer file-version retention. The paid unit is an endpoint, not a bucket. That distinction matters because Backblaze is trying to make two unlike risks look similar on the buyer's invoice. One risk is metered storage infrastructure. The other is a flat-price promise to protect a machine whose owner may or may not understand retention windows, external-drive rules, exclusions or restore mechanics.

The first economic question is what is expensive about that promise. A storage account is expensive because the customer pays for visible capacity, while the provider pays for invisible redundancy, unused headroom, failed drive replacement, support, bandwidth, power, rack space, software, billing, compliance work, security controls and restore operations. Durability is not just a statistic. It is a workflow. Customer data has to be broken into parts, placed across physical drives, checked, reconstructed when pieces fail, moved as hardware ages and served back in the moments when the customer is most impatient. If the system works, the customer experiences nothing dramatic. That is the business problem. Backblaze is paid to make the hard part feel boring.

The second question is whether public evidence proves that Backblaze is good value. It proves part of it. The company reported more than 500,000 customers, operations in more than 175 countries and about 5 billion gigabytes of data storage under management in its 2025 annual report. It reported 2025 revenue of about $145.8 million, cost of revenue of about $57.0 million, gross profit of about $88.8 million and a net loss of about $25.6 million. In Q1 2026, it reported $38.7 million of revenue, $15.1 million of cost of revenue, $23.5 million of gross profit and a $6.1 million net loss. Those figures show that the storage platform has material gross profit, but the business is still investing, still leasing infrastructure and still not proving durable net profitability.

The third question is who carries the tail risk. Customers carry some of it through service terms, restore practices and the practical limits of their own workflows. Backblaze carries some through reputation. If an ordinary storage service is slow, a customer may complain. If a backup service fails at restore time, the customer remembers it as a trust breach. That asymmetry is why Backblaze's cheap pricing is not just a growth tactic. It is a controlled bet that ordinary customer behaviour will remain predictable enough to fund a storage system whose failures are mostly handled before the customer sees them.

Backblaze's public identity and ownership context

Backblaze, Inc. is a Delaware corporation. Its 2025 Form 10-K says it was incorporated in Delaware in 2007 under the name Backblaze, Inc., completed its initial public offering in November 2021 and lists Class A common stock on the Nasdaq Global Market under the ticker BLZE. The same filing gives a San Francisco principal executive office address and identifies the website as www.backblaze.com. The public filings do not present a parent company controlling the business. It is a public company, not a private subsidiary inside a hyperscaler.

The company's public description is carefully positioned. It calls the Backblaze Storage Cloud a purpose-built, web-scale software architecture operating on commodity hardware. It frames its two main offerings as B2 Cloud Storage, an infrastructure-as-a-service storage product, and Computer Backup, a software-as-a-service backup product. The filings repeatedly emphasize cost efficiency, transparent pricing, openness and a move upmarket toward larger enterprise, AI, media, MSP, reseller and neocloud opportunities.

That positioning matters because Backblaze is not trying to beat Amazon Web Services, Microsoft Azure or Google Cloud by offering a broader cloud platform. It is trying to win accounts where storage is important enough to be bought separately and where the buyer resents the complexity or egress exposure of larger vendors. The company can therefore be both smaller and sharper. It does not have to provide every compute, database, analytics and identity service. It does have to convince buyers that specialized storage is not more fragile because it is specialized.

The 2025 10-K says the company had over 500,000 customers at year-end 2025, including about 400,000 Computer Backup customers and about 120,000 B2 Cloud Storage customers, with about 20,000 customers using both services. It also says no single customer accounted for more than 10 percent of total revenue in 2025 or 2024. This is a useful concentration signal. It suggests Backblaze is not visibly dependent on one buyer. But it does not prove the absence of product-line concentration, usage concentration or future dependence on a small set of large AI or platform customers.

The Q1 2026 10-Q adds an important recent wrinkle. It says Backblaze continued to have more than 500,000 customers as of March 31, 2026, and describes a strategy around larger enterprise customers, AI supply-side platform relationships, B2 Overdrive, B2 Neo and developer-led adoption. It also states that, effective May 1, 2026, Backblaze implemented B2 Cloud Storage pricing changes, including an increase in pay-as-you-go storage pricing and the elimination of API transaction fees. That sentence is critical because it shows that the low-price offer is not static. It is being tuned as usage patterns, competition and infrastructure costs change.

The financial record shows a growing storage business, not a finished profit machine

Backblaze's economics are easier to understand if the revenue line is separated from the capital and support machinery underneath it. In 2025, revenue rose to about $145.8 million from about $127.6 million in 2024 and about $102.0 million in 2023. Cost of revenue was about $57.0 million in 2025, down slightly from about $58.3 million in 2024, while gross profit increased to about $88.8 million. That improvement is consistent with operating leverage, price actions, infrastructure efficiencies or product mix. It does not by itself identify which customers are most profitable.

The company still lost money on a net basis. Its 2025 net loss was about $25.6 million, narrowed from about $48.5 million in 2024 and about $59.7 million in 2023. Q1 2026 continued the pattern: positive gross profit, negative net income. This is not unusual for a smaller public cloud infrastructure company, but it matters for the storage buyer because pricing discipline cannot be judged only at the product page. A provider can sell cheap storage if gross margin funds the rest of the organization, if capital can be financed at acceptable cost and if customers grow into accounts that do not overwhelm support or bandwidth.

The balance sheet and lease disclosures show why "cheap storage" is not asset-light in the ordinary software sense. At December 31, 2025, Backblaze reported about $29.2 million of cash and cash equivalents, about $57.3 million of property, plant and equipment plus finance-lease right-of-use assets after depreciation and amortization, about $35.7 million of finance lease liabilities and about $25.4 million of operating lease liabilities. At March 31, 2026, those balances were about $26.3 million of cash and cash equivalents, about $61.3 million of property and finance-lease assets, about $38.5 million of finance lease liabilities and about $23.5 million of operating lease liabilities.

This is the storage account's hidden invoice. The customer sees a price per TB-month or a backup subscription. Backblaze sees equipment leases, operating leases, bandwidth, data-centre rent, customer support and service-availability staff. In Q1 2026, the company described cost of revenue as including data-centre spaces, network and bandwidth costs, depreciation of equipment and finance-leased equipment, and personnel costs tied to customer support and maintaining service availability. That description connects the buyer's account directly to the cost envelope in the thesis. The platform is cheap only if those inputs remain predictable relative to stored data and downloads.

Remaining performance obligations are also relevant. Backblaze reported about $66.2 million of remaining performance obligations at the end of 2025, up from about $41.3 million at the end of 2024. This indicates a larger contracted revenue base, which can help plan capacity. But the public record also says many customers can reduce usage, not renew, or change storage behaviour. In storage, a contract can improve visibility without eliminating operational uncertainty. A customer can keep paying while changing the shape of the load: more egress, more small files, more restore events, more support tickets, more geographic demands.

Why durability costs money even when drives are commodities

The strongest operational evidence Backblaze publishes is its drive-stat material. The company's 2025 drive-stat report says that, at the end of 2025, it monitored 349,462 hard drives used to store data and, after applying annual criteria, analyzed 344,196 drives across 30 models. It reported a 2025 annualized failure rate of 1.36 percent, down from 1.55 percent in 2024, and a lifetime annualized failure rate around 1.30 percent. This material is valuable because it admits the thing that cloud-storage marketing often hides: durability is built on hardware that fails constantly, in patterns that require procurement, monitoring, replacement labour and software compensation.

The drive-stat reports do not prove customer-level durability. They do not show how any specific customer's files were placed, restored or protected. They do not disclose every procurement price, repair workflow, erasure-coding overhead or operational incident. They do, however, prove that Backblaze runs at a scale where drive failure is not an exception. Failure is the normal work.

Backblaze's 2025 Form 10-K describes the storage architecture in functional terms. It says the software splits uploaded customer files into several data parts, adds redundant parts and stores those parts across discrete hard drives in different servers in a data centre, so that data can be reconstructed from remaining parts if some are lost or offline. That is a concise explanation of why commodity drives can be turned into a durable service. It is also a reminder that the low price depends on software and operations being good enough to keep redundancy overhead economical.

The same filing says the platform manages hundreds of thousands of hard drives across multiple data centres. It also says the company leases data centre spaces in California, Arizona, Virginia, Toronto and Amsterdam. Geographic choice is part of the product value for customers with locality, latency, sovereignty or compliance preferences. But locality increases the planning problem. Capacity has to exist in the right region before all of the revenue is known, and the provider must avoid both underbuilding and overcommitting. The 10-K explicitly warns that if Backblaze underestimates needed capacity, or lacks enough capacity at existing data centres, it may fail to meet customer needs; if it overestimates, it may bear excess lease and equipment obligations.

The discipline is therefore not simply buying cheap drives. It is deciding when to lease equipment, when to add data-centre space, when to refresh aging hardware, which drive models to trust, which regions to expand and how much redundancy overhead is acceptable. A lower TB-month price can survive only when those decisions are made repeatedly without large mistakes.

Egress is the bill that tests the story

Backblaze's object-storage proposition leans heavily on egress economics. The B2 pricing page says users get free egress up to three times average monthly data stored, with additional egress at $0.01 per GB, and free downloads through many integrated CDN and compute partners. The cloud-storage page describes B2 as S3-compatible storage with free egress in specific circumstances and positions it as a way to avoid restrictive legacy pricing.

This is not just a discount. It is a behavioural assumption. If a customer stores 100 TB and downloads 20 TB a month, the economics are straightforward under the published B2 tariff. If a customer stores 100 TB and tries to move 800 TB repeatedly to arbitrary destinations, the egress cost and network burden become a different business. Backblaze's 2025 10-K recognizes this directly. It warns that, because the company offers Computer Backup at a fixed price and offers free egress for B2 subject to limitations, customers who back up unusually large amounts of data, use excessive egress, or increase backed-up data faster than storage costs decline could adversely affect costs, gross margins and infrastructure.

The Q1 2026 Form 10-Q makes the issue more current. It says Backblaze changed B2 Cloud Storage pricing effective May 1, 2026, including increasing pay-as-you-go storage pricing and eliminating API transaction fees. That is a trade. The company appears to be simplifying one friction point while raising the base storage price. The public record does not say whether that change was driven by bandwidth, support, drive procurement, competitive positioning, AI workloads or customer feedback. But it confirms that the tariff is an instrument of cost control.

Competitor pricing clarifies why this matters. AWS S3's pricing page is organized around storage classes, regions, request types, retrievals and transfer behaviour. Google Cloud Storage and Azure Blob Storage likewise present multi-dimensional tariffs by class, region, operation and egress. Wasabi advertises hot cloud storage starting at $7.99 per TB-month with no egress or API request fees, while Cloudflare R2 is positioned around zero egress fees and S3-compatible APIs. Each substitute moves the economic pressure to a different place. Hyperscalers charge for breadth and detailed usage. Wasabi simplifies around capacity. Cloudflare R2 competes aggressively on egress. Backblaze tries to sit between those models: cheap hot storage, some free egress, low overage and a supportable infrastructure footprint.

The practical buyer question is not "Who has the lowest price on one row?" It is "Which bill matches my data movement?" Backblaze looks stronger when data is stored hot, accessed predictably and integrated with partners that reduce egress cost. It looks less proven when the workload is bursty, regionally complex, extremely latency-sensitive or likely to require urgent large restores to arbitrary destinations.

Support discipline is part of the margin

Storage support is not generic customer service. A support ticket often arrives when the customer thinks data is missing, a restore is too slow, a drive was disconnected too long, a bucket policy failed, a backup client excluded files, a ransomware event triggered panic, or a migration bill is larger than expected. The emotional load is heavier than in many software categories because the customer's fear is irreversible loss.

Backblaze's support and product terms make that margin discipline visible. The Terms of Service state that users are responsible for internet access, data network services and supported software environments, and that services can be used only for the number of users, computers or drives permitted under the plan and documentation. For Computer Backup, the terms say the service is designed for individual computers and directly attached external drives, and that "unlimited" use does not permit inconsistent use such as servers, data centres, shared storage systems or enterprise infrastructure not expressly covered by the service. The terms also reserve rights around disproportionate resource consumption after notice.

Those clauses are not incidental legalism. They are unit economics. Flat-rate backup is economically possible only if the average endpoint behaves like an endpoint. If users treat a consumer or small-business backup client as an unlimited data-centre archiving system, the provider's costs detach from the subscription price. The more Backblaze has to police edge cases, the more the apparent simplicity depends on boundaries that customers may not read until something goes wrong.

The published Computer Backup pricing page adds another support dimension. Backblaze offers web-based restores, a dedicated restore app and mailed hard-drive restores. These features are valuable precisely because restore is the moment of truth. But they also cost money. Preparing large restores, shipping devices, handling encryption and guiding non-technical users through recovery are not the same as serving a small object through an API. A buyer should treat restore mechanics as part of the product, not a footnote.

B2 Reserve support language is also revealing. Backblaze's B2 pricing FAQ says the standard support path is available during Pacific business hours, while premium support options can extend coverage, and B2 Reserve includes Tera support with a four-business-hour response-time commitment. That separates cheaper self-service storage from higher-touch accounts. Backblaze can publish low pay-as-you-go prices because not every account receives enterprise support economics. Larger accounts can pay for capacity, predictability and response commitments.

The three substitute tests a buyer should run

The first substitute test is hyperscaler hot storage. Backblaze's own pricing calculator uses published U.S. West rates that, at the time captured, model AWS S3 Standard storage at roughly $0.026, $0.025 and $0.024 per GB-month across storage tiers, with egress tiers at about $0.09, $0.085, $0.07 and $0.05 per GB. It models Google Cloud storage around $0.023 per GB-month with higher egress tiers and Azure Blob storage with tiered hot-storage and egress rates. Even if exact regional rates differ, the structure is the point. A buyer comparing B2 to the hyperscalers is comparing a simpler storage-focused bill with a broader cloud platform bill.

The second substitute test is specialist low-cost storage. Wasabi's public pricing page advertises hot cloud storage starting at $7.99 per TB-month with no egress or API request fees. Cloudflare R2 advertises S3-compatible storage with zero egress fees. IDrive E2 has also competed in low-cost S3-compatible storage. These substitutes mean Backblaze cannot rely only on being cheaper than AWS, Azure or Google. It has to prove that its durability, integrations, support, performance, migration help and public operational transparency justify choosing it over other low-cost specialists.

The third substitute test is self-managed or bundled backup. A small business can use local NAS hardware, external drives, Microsoft 365 or Google storage, Synology or QNAP tools, MSP-managed backup, CrashPlan, IDrive, Acronis, Veeam targets, or a mix of cloud and local copies. Backblaze Computer Backup competes against inertia as much as against another tariff. Its simplicity is powerful: install the client, pay per computer, restore when needed. But a sophisticated customer may prefer more control, clearer versioning, no broad exclusions, local-first restores or managed service support.

These substitutes create a ceiling on Backblaze's pricing power. The company acknowledged this in the 2025 10-K risk factors, naming AWS, Google Cloud, Microsoft Azure, EMC/Dell and NetApp, and noting newer storage offerings from companies such as Cloudflare, CoreWeave, DigitalOcean and Vultr. The filing warns that competitors could offer lower prices or bundle services in ways Backblaze does not. That is exactly the pressure a specialized storage provider faces. It can be cheaper because it is focused, but it is also exposed because larger rivals can bundle and smaller rivals can copy the simple-storage message.

Customer retention and switching costs cut both ways

Storage has a natural switching cost: moving data is slow, operationally risky and sometimes expensive. Once a customer stores tens or hundreds of terabytes in a bucket, the practical cost of change includes engineering time, egress, validation, application reconfiguration, compliance review and confidence that the new provider will not fail at the same moment. That should help Backblaze retain accounts.

Backblaze's Q1 2026 filing gives some retention evidence. It says gross customer retention rates have remained approximately 90 percent across revenue products as of both March 31, 2026 and the comparison date. It also says B2 Cloud Storage net revenue retention was 110 percent as of March 31, 2026, while explaining that customers who store data with Backblaze generally increase stored data over time. The company said B2 annual recurring revenue grew by $20.5 million, or 28 percent, compared with March 31, 2025, while Computer Backup ARR declined by $1.7 million.

This is a mixed signal in a useful way. B2 appears to be the growth engine. Computer Backup appears flatter and more mature. That makes economic sense. Object storage can grow with data-heavy customers, partner integrations, AI workflows and media workloads. Endpoint backup has a large installed base but is constrained by per-computer pricing, consumer sensitivity, support burden and the practical limits of ordinary user data.

Switching cost can also hurt Backblaze. If customers believe the service is hard to leave, they may be more sensitive to egress terms, portability, APIs and support. Backblaze tries to counter that with S3 compatibility, data-migration assistance and an open-cloud message. The promise is not merely "we are cheap"; it is "we will not trap you the way others do." That is a powerful message only if the customer's real restore or migration experience matches it.

Data sovereignty and locality are part of the product, but not fully visible

Backblaze's public filings and product pages indicate geographic choice. The 2025 10-K says the storage platform offers multi-region geographic choice, including East and West Coast regions in the United States, Canada and Europe. The same filing identifies leased data-centre spaces in California, Arizona, Virginia, Toronto and Amsterdam. For customers in regulated or latency-sensitive environments, this is meaningful. It lets a buyer discuss locality, disaster recovery and sovereignty with more precision than a generic "cloud" promise.

But public evidence does not show everything a sovereignty-sensitive buyer would need. It does not publish full data-flow diagrams for every service, full subprocessor details in the source material reviewed here, exact replication topology by region, or customer-level residency guarantees beyond the service documentation and contract terms available to customers. It also does not prove that every integration, CDN path, support workflow or restore process keeps data in the same jurisdiction.

The correct reading is therefore conditional. Backblaze has public region claims and visible data-centre geography. That is enough to identify the locality dimension of the product. It is not enough for a bank, hospital, public agency or cross-border enterprise to skip contractual review. Data sovereignty is a commercial feature, but the public page is only the start of the evidence.

Public technical records show boundaries, not the inside of the platform

Public DNS and RDAP records help define the visible service boundary. A DNS check on July 6, 2026 showed backblaze.com and www.backblaze.com resolving to Cloudflare edge addresses, with Cloudflare name servers for the domain. Mail exchange records pointed to Proofpoint-hosted mail filtering. Public hostnames used for Backblaze services showed a mixed picture: secure.backblaze.com resolved to Cloudflare edge addresses, while f001.backblazeb2.com and api.backblazeb2.com resolved to addresses that ARIN RDAP associates with Backblaze Inc. IP ranges. IPinfo similarly labelled the checked B2 service addresses as AS40401 Backblaze Inc in San Jose, California.

This evidence proves only the public-facing boundary. It shows that Backblaze uses Cloudflare for some web-facing surfaces, Proofpoint for mail filtering and Backblaze-numbered resources for some B2 hostnames. It does not prove where customer data is stored, how erasure coding is deployed, how traffic is routed after ingress, which data-centre building holds a given file, or how internal failover is designed. Treating DNS as architecture proof would overstate the evidence.

The technical record is still commercially useful. It shows Backblaze is not a fully self-contained island. Like most internet services, it depends on external edge, security, mail and network ecosystems. The buyer should therefore separate platform durability from public access dependencies. Data can be durable while a web front door is impaired; a web property can be reachable while a storage region has trouble. The public technical record helps ask better questions, not answer all of them.

Drive transparency is both evidence and marketing

Backblaze's drive-stat franchise is unusual because it exposes a working storage operator's hardware reality. That transparency has value for customers and for the company's brand. It builds credibility with technical buyers, researchers and journalists. It also creates an implicit message: if the company publishes drive failures, perhaps it is serious about operational measurement.

But drive transparency is not the same as account transparency. The drive-stat report is about hard-drive models and annualized failure rates in Backblaze's environment. It does not reveal per-customer restore success rates, unresolved ticket ageing, region-specific availability, object-level durability incidents, failed migration rates, support escalations or how many customers are affected by any particular hardware issue. It also may reflect Backblaze's own workload and procurement mix, not the universal reliability of a drive model in every environment.

The useful conclusion is narrower and stronger. Backblaze has evidence that it operates a large drive fleet, tracks failures and uses those observations publicly. That supports the claim that durability is an operational discipline inside the company. It does not prove that every customer will experience a painless restore or that low storage prices have no hidden tradeoffs.

The support edge cases are not noise

Unofficial customer signals should not be treated as audited fact, but they are important market colour because backup is judged at the edge. Reddit threads, Trustpilot reviews, G2 comments and specialist articles show a split experience. Many users praise simple setup, low price, useful restores and predictable B2 billing. Other users complain about external-drive retention confusion, slow restores, support frustration, exclusions for synced folders or uncertainty after terms changed around disproportionate usage.

The most commercially relevant 2026 signal is not that every complaint is correct. It is that the phrase "unlimited backup" invites expectations that are broader than the product's cost envelope. Tom's Hardware reported in April 2026 on criticism around changes to Backblaze's Terms of Service and exclusions involving synced folders such as Dropbox, OneDrive and Google Drive. The report is not a substitute for the terms themselves, but it captures a public customer perception problem: users who thought a backup tool was protecting all local data may interpret exclusions or usage limits as a change in the bargain.

Backblaze's own Terms of Service are the stronger evidence. They define the service boundary and reserve rights around unsupported use, outdated software and disproportionate resource consumption. The unofficial signals show why those boundaries matter commercially. The customer's fear is not abstract data loss. It is the moment after a deletion, ransomware event, failed disk or mistaken folder sync, when the customer discovers the exact scope of what was backed up.

B2 has a different support edge. Developers may be less surprised by API, bucket, lifecycle and performance limitations, but they are more sensitive to rate limits, egress, migration behaviour and interoperability. Backblaze's Q1 2026 performance-stat blog explicitly discusses rate limits and bandwidth caps in testing. That transparency is helpful. It also tells buyers that high-throughput storage is not magic. Large workflows may need account tuning, support engagement and realistic testing before a production migration.

Regulation, privacy and geopolitical risk are ordinary but material

Backblaze is a U.S. public company selling storage globally. That creates ordinary privacy, cybersecurity, export-control, sanctions, tax, trade and data-locality exposure. The 2025 10-K discusses privacy and security law risk, cybersecurity risk, international expansion and foreign-currency exposure. It also discusses global trade policy and tariffs, including the possibility that technology-related goods and components could increase direct or vendor costs.

For this economic unit, tariffs and hardware supply matter more than they might for a pure software company. A storage provider needs hard drives, servers, networking gear, racks, power, colocation and replacement parts. If trade barriers, component shortages or supplier consolidation raise equipment costs, the pressure eventually reaches pricing, margins, capacity planning or all three. The 10-K names hard drives and semiconductors as key components with limited sources of supply and notes past industry events involving Toshiba and Western Digital. It says shortages, pricing fluctuations or supply interruptions could increase costs, lease liabilities, interest, depreciation and inventory levels.

Geopolitical risk also affects customer demand. A U.S. storage provider can be attractive to North American buyers and to global customers who want U.S. infrastructure. It can be less attractive to buyers with strict data-residency, procurement sovereignty or public-sector cloud requirements. Backblaze's Canada and Amsterdam regions partly answer that concern, but public region presence is not the same as a country-specific sovereign cloud.

The price works only if "boring" stays measurable

The thesis is that Backblaze's low published storage prices work only if drive procurement, data-centre operations, support, egress expectations and customer fear of data loss stay inside a disciplined cost envelope. The public record supports that thesis.

Drive procurement is visible in finance leases, drive-stat scale and supplier-risk language. Data-centre operations are visible in leased facilities, operating lease liabilities, expansion language and risk factors around capacity. Support is visible in cost-of-revenue definitions, support-plan distinctions and restore mechanics. Egress expectations are visible in B2 pricing, free-egress limits, partner egress claims and explicit risk-factor language about excessive egress. Customer fear is visible in the whole Computer Backup product and in unofficial customer signals around retention, exclusions and restore speed.

The thesis is not fully proven because the public record lacks granular metrics. Backblaze does not disclose average storage per Computer Backup customer, average restore size, support tickets per customer, egress-to-storage ratio, gross margin by product after support, failed restore rate, region-level utilization, drive procurement cost per TB, or the number of accounts that trigger disproportionate-use controls. Without those metrics, an outside observer cannot know exactly where the low-price envelope is tightest.

But the available evidence is consistent with a disciplined low-cost operator rather than a simple loss leader. The company has meaningful gross profit, a growing B2 revenue engine, large storage under management, public drive operations, and a clear willingness to adjust pricing and terms. The risk is that the next growth wave, especially AI, media and neocloud workloads, may have heavier throughput and support demands than the older mix of backup and predictable object storage.

What would change the judgement

The judgement would improve if Backblaze published product-line gross margin including support and bandwidth, egress ratios by customer cohort, restore success and speed distributions, region-level availability, average support response times by tier, and customer-visible incident history by service. It would also improve if the company showed that B2 Overdrive, B2 Neo and Powered by Backblaze accounts have attractive economics after network and support costs.

The judgement would weaken if the company had to raise B2 prices repeatedly without corresponding value improvements, if Computer Backup complaints about exclusions and restores became a larger public pattern, if finance-lease costs rose faster than revenue, if large AI or neocloud customers created concentration risk, or if drive procurement and data-centre capacity tightened in a way that reduced gross margin. It would also weaken if the low-cost specialist market, especially Wasabi, Cloudflare R2 and other S3-compatible alternatives, forced Backblaze into a price war that it could not offset with operational efficiency.

The most important missing metric is egress intensity. Storage providers can survive cheap capacity. They struggle when the traffic shape, support shape and restore shape do not match the tariff. If Backblaze's customers mostly store data and download within the free or low-cost envelope, the model is credible. If a growing share of customers use the platform as a high-churn data-movement layer, the model has to be repriced, bundled or limited.

Public evidence used for this assessment

Conclusion

Backblaze's low-price story is credible because it is not only a slogan. The public filings show a company with a large customer base, billions of gigabytes under management, rising revenue, meaningful gross profit, public drive operations and explicit awareness of the costs that could break the model. The pricing pages show why buyers pay attention: B2 is simple enough to compare, Computer Backup is cheap enough to feel like insurance, and the egress policy is friendlier than many hyperscaler bills for the right workload.

The same evidence keeps the judgement bounded. Backblaze is not selling free durability. It is selling a disciplined approximation of it, financed through commodity hardware, software redundancy, leases, support tiers, customer behaviour assumptions and a willingness to define the limits of "unlimited." The public record supports the thesis that cheap storage works only while drive procurement, data-centre operations, support load, egress behaviour and customer fear remain inside the envelope. The thesis remains unproven at the account level because Backblaze does not disclose cohort margins, restore outcomes, egress intensity or support burden by product.

For the customer, the best reading is practical. Backblaze is compelling when the job is predictable hot object storage, offsite backup, ransomware-resistant copies, media archives, MSP storage targets or developer storage that values simple pricing and S3-compatible access. It is less proven as a blank cheque for extreme downloads, ambiguous "unlimited" endpoint behaviour, urgent multi-terabyte consumer restores or regulated workloads that need detailed contractual sovereignty guarantees. The company makes durability look boring. The buyer still has to ask which parts of the boring machine are included in the price.