DDN
DDN is an enterprise data platform for AI and HPC workloads, pairing the Infinia data engine, the EXAScaler parallel file system and dedicated hardware to keep very large GPU fleets fed and fully utilized.
What is DDN?
DDN, the brand of DataDirect Networks, Inc., sells the data layer that sits underneath large GPU estates. The company presents itself as the data intelligence platform behind AI factories, sovereign AI programmes, hyperscalers and the most demanding accelerated workloads, and it has been in high-performance storage since 1998, when it was formed by the merger of MegaDrive and ImpactData.
The platform has two software pillars. Infinia is described as an AI data engine: it unifies fragmented silos, indexes metadata at high speed for instant search and filtering, keeps embeddings, vectors and inference state close to compute through a KV store and KV cache, and orchestrates placement across core, cloud and edge. EXAScaler is the parallel file system, derived from Lustre, that delivers multi-terabyte-per-second throughput and is claimed to run in more NVIDIA SuperPODs than any competing solution. Around them sit hardware platforms such as HyperPOD, AI400X3 and AI400X2 Turbo, plus the Horizon, IndustrySync and Insight software. Access is offered through native S3 and POSIX on the same platform, with NFS also referenced.
The numbers on the site are vendor claims rather than independent benchmarks, and they are ambitious: up to 99% GPU utilization, 10x return on investment, 70% more GPU utilization, 50% power savings per workload, 75% lower AI costs, 100x faster insights, 15x faster checkpointing, and for Infinia specifically a 75% reduction in cost per token, 18x more tokens per watt and 22x faster RAG.
What gives the pitch weight is the reference list. NVIDIA is quoted directly by Jensen Huang, xAI is credited with building a 200,000-GPU environment for Grok, and Ford, Roche, Jump Trading, Core42, Scaleway and Yotta all appear as customers. DDN also claims 40% of the Global 500, 11,000 customers, revenues of $400 million reported for 2021 and more than $500 million invested in research and development. Industries served span automotive, energy, financial services, healthcare and life sciences, higher education, manufacturing, media, security and surveillance, telecommunications and the US public sector.
What it does
- Unify structured, semi-structured and unstructured data into a single high-performance platform
- Feed GPUs continuously during training and cut checkpointing time, claimed at 15x faster
- Serve real-time inference and RAG pipelines, with claims of 22x faster retrieval and 25x lower time to first byte
- Index metadata at high speed so massive datasets stay searchable and filterable
- Orchestrate and place data across core, cloud and edge environments
- Isolate thousands of tenants in one namespace with QoS, granular access controls and real-time encryption
- Expose the same data through native S3 and POSIX access without copies or separate infrastructure
When to use DDN / When not to
A quick filter to help you decide if DDN is the right fit.
When to use DDN
- AI platform and infrastructure teams training large language models on hundreds or thousands of GPUs
- Cloud and GPU-as-a-service providers who need multi-tenant isolation, QoS and predictable throughput
- Sovereign AI programmes that must keep national datasets encrypted, isolated and inside their own jurisdiction
- HPC and supercomputing centres running parallel file systems at multi-petabyte scale
- Life sciences, financial services and automotive R&D groups whose pipelines stall on data access rather than compute
When not to use DDN
- Individuals and small teams looking for a ready-to-use AI assistant rather than storage infrastructure
- Anyone who needs a published price list: nothing can be budgeted without going through a sales conversation
- Buyers expecting self-service sign-up, a free tier or a downloadable trial
- Mobile-first users: there is no iOS or Android application and no browser extension
- Organisations without data centre skills or a partner to handle installation, migration and tuning
How to use DDN
A typical end-to-end flow, from setup to results.
- Start from the Contact us or Contact a storage specialist form, or the Talk to an Expert button on a product page
- Size the opportunity with the GPU efficiency ROI calculator published on the site
- Read the pre-sales material: AI Readiness Checklist, Infinia solution brief, EXAScaler data sheet
- Book a session in the DDN Demo Center to see the platform running
- Work through the technical fit with a DDN specialist or an authorised partner (GSI, VAR, cloud or OEM)
- Negotiate the quote directly or buy through the partner channel; EXAScaler Cloud is also listed on Google Cloud Marketplace
- Activate your software licence through the license activation page once the order is placed
- Engage Professional Services for planning, migration, AI workload tuning and run or operate support
- Train the team through the DDN Education offering
- Run day-to-day operations from the support.ddn.io customer portal, using support plans, technical bulletins and the DDN Connect diagnostic utility
Pros & Cons
Pros
- Reference customers named openly on the site, including NVIDIA, xAI, Ford, Roche, Jump Trading, Core42 and Scaleway
- Long track record: founded in 1998, with more than $500 million claimed in research and development
- End-to-end coverage from hardware and file system to professional services, training and support plans
- File and object workloads run against the same data through native S3 and POSIX, with no copies
- Multi-tenancy and isolation designed for sovereign AI and cloud providers rather than retrofitted
- Genuine worldwide footprint, with offices across North America, Europe, the Middle East, India, Asia and Australia
- Partner ecosystem and marketplace availability, EXAScaler Cloud being listed on Google Cloud Marketplace
Cons
- No public pricing at all and no pricing page anywhere on the site, so nothing can be budgeted independently
- No free plan and no free trial announced, only a demo centre and a pre-sales ROI calculator
- Every performance figure is a vendor claim, with no measurement methodology published on the pages reviewed
- No public API documentation is reachable without an account
- No data processing agreement, no subprocessor list and no Article 27 EU representative published
- No security certification such as SOC 2 or ISO 27001 is mentioned anywhere on the public site
- Support has no published email address: everything goes through the customer portal
Pricing & Plans
DDN publishes no pricing whatsoever. There is no pricing page on the site or in its sitemap, no permanent free plan is described and no free trial is announced. Commercial terms are established through a sales conversation or an authorised partner, with EXAScaler Cloud also available through Google Cloud Marketplace. Support plans exist as a separate offering, but no amount, currency or billing unit is disclosed. Prospective buyers should therefore expect an enterprise quotation process rather than a published rate card.
Data, GDPR & hosting
A consolidated view of how DDN handles your data.
GDPR overview
Neither the privacy policy nor the terms ever use the words GDPR or General Data Protection Regulation, so DDN makes no explicit compliance claim. Several GDPR-shaped mechanisms are nonetheless documented. The policy, effective 8 December 2025, lists its legal bases (performance of a contract, legal obligation, consent, legitimate interests) and the rights it will honour: confirmation of processing, access, rectification, portability, objection and restriction, withdrawal of consent and erasure, all exercised by writing to privacy@ddn.com. Cross-border transfers out of the European Economic Area are covered by the European Commission's standard contractual clauses. What is missing is equally clear: no Article 27 EU representative is named, no data protection officer is identified, no data processing agreement is published and no subprocessor list exists. The California CCPA notice, by contrast, is detailed.
Who owns the data?
DDN's Terms of Use, effective 8 December 2025, state that all rights, title and interest in the Solutions and in the Content belong to DDN, its licensors or certain third parties. Visitors receive only a limited, non-exclusive, revocable, non-transferable and personal licence, restricted to non-commercial use unless an active customer agreement says otherwise, and that licence grants no ownership or intellectual property interest. Anything sent to DDN as feedback is treated differently: the company, its affiliates, service providers, licensees and assigns may use, reproduce, modify, display and distribute those ideas for any purpose. For personal data, the privacy policy names DataDirect Networks Inc. as the controller, reachable at privacy@ddn.com.
Reuse rights
Reuse rights are narrow. A visitor may download, display, view and keep a single copy of the Content as it is shown, for personal and non-commercial purposes only, unless an active customer agreement extends that scope. Harvesting information from the site, reverse engineering or modifying the Content, interfering with security features and implying an unauthorised association with DDN are all expressly forbidden, and every right not granted is reserved with no implied licence. Product-specific Additional Terms may apply on top and take precedence where they conflict. One caveat deserves attention: DDN warns users not to enter confidential or proprietary information into the site's AI features, because submissions may be used to keep training them.
Data retention & training
Hosting summary
For the website and related services, DDN's privacy policy states that the Solutions are hosted in the United States and in other countries where DDN and its subsidiaries are located or do business. Personal data may therefore be transferred and processed outside the visitor's own country, under laws that differ from local ones; for flows originating in the European Economic Area, DDN says it enters into the European Commission's standard contractual clauses. The controller is DataDirect Networks Inc., based in Chatsworth, California. Beyond the United States, no specific hosting country or region is named, and the site publishes no trust or security page and no hosting certification such as SOC 2 or ISO 27001. One technical point is worth separating out: ddn.com resolves to a Cloudflare anycast address, which is a content delivery point of presence and says nothing about where customer data is actually stored.
Things to keep in mind
Risks and trade-offs to weigh before adopting DDN.
- DDN's terms state that information entered into the site's AI features, including the Pax AI chatbot, may be used to keep training them, and warn against submitting confidential or proprietary material
- Those AI features are described as experimental and unsupervised, capable of returning inaccurate content or reproducing copyrighted training data
- The terms impose individual binding arbitration and waive class actions unless the user opts out within 30 days of accepting them
- The California notice acknowledges selling and sharing personal information with cookie vendors and adtech providers for cross-context behavioural advertising
- With no published data processing agreement or subprocessor list, contractual compliance has to be negotiated case by case
- Concentrating the data layer of an entire GPU estate on one vendor creates real lock-in and a single point of dependency
- Vendor-supplied performance multipliers are easy to carry straight into a business case; sizing a project on them without independent testing invites disappointment
Setup & Integrations
Technical difficulty
High, and unavoidably so. This is data centre infrastructure: hardware such as HyperPOD or AI400X3, a parallel file system, and integration with existing network and GPU estates. There is no self-service sign-up and no online registration; a licence must be activated after purchase. DDN offers Professional Services for planning, migration, AI workload tuning and operations, plus an Education programme to build in-house skills, and it claims deployment and scaling are straightforward even for very large workloads. Expect to need storage and HPC expertise internally, or a partner supplying it.
Deployment
Integrations
Behind DDN
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What exactly does DDN sell?
Who is it aimed at?
How much does it cost?
Is there a free plan or a free trial?
Which data access protocols are supported?
Can it be deployed outside a private data centre?
Which organisations already use it?
Is there public API documentation?
Where is data hosted, and what about GDPR?
How do I get support or close an account?
Should you pick DDN?
DDN occupies an unglamorous but decisive position: the data layer that determines whether an expensive GPU estate actually works. Its case is built on longevity, having been in high-performance storage since 1998, and on a reference list most vendors cannot match, with NVIDIA, xAI, Ford and Roche all named on the site. The technical proposition is coherent too, pairing the Infinia data engine with the EXAScaler parallel file system so that object and file workloads share the same data through native S3 and POSIX access, across core, cloud and edge.
The reservations are about transparency rather than capability. Every headline figure, from 99% GPU utilization to a 75% reduction in cost per token, comes from DDN itself, with no published methodology on the pages reviewed. Nothing is priced publicly, there is no free tier and no trial, so even a first estimate requires a sales conversation. On the compliance side the picture is partial: standard contractual clauses are cited for European transfers, but the GDPR is never named, no Article 27 representative is designated, no data processing agreement is published, no subprocessor list exists and no SOC 2 or ISO 27001 certification is mentioned. Buyers with a procurement checklist will have to obtain those documents directly. It is also worth noting that DDN's terms allow input to the website's own AI features to be reused to train them, and warn against entering confidential material there.
The practical verdict is straightforward. If your models are waiting on data rather than on compute, and you operate at a scale where a percentage point of GPU utilization is worth real money, DDN belongs on the shortlist. If you are looking for an AI tool to use tomorrow morning, this is the wrong layer of the stack entirely.
- Choosing a selection results in a full page refresh.
- Opens in a new window.