At a glance
Who gets named when buyers ask AI about storage? · US · November 2026Buyers ask AI about storage, and AI names WEKA before DDN
Ask AI which storage to buy without naming a company, and it names DDN in 29 of 81 answers. WEKA gets 39. On 13 of 30 questions no AI names DDN, including KV cache, sovereign AI and AI data platforms.
30 buyer questions on each engine, 2 Nov 2026: 90 answers, all returned. ChatGPT and Claude were not part of this scan.
Who is DDN, and who buys?
- What it is: DDN (DataDirect Networks) sells data storage for AI and high-performance computing (HPC): EXAScaler, a fast file system built on Lustre, and Infinia, its AI data platform.
- Who buys: AI infrastructure architects, heads of HPC, research computing directors and CIOs who build or run large GPU clusters.
- Main rivals: WEKA, VAST Data, Pure Storage, IBM, NetApp and Dell in AI answers; MinIO and the big clouds on object storage.
What buyers search for
Before a team picks a storage vendor, someone types a question into Google or asks an AI. Most of these questions do not name any company. (Sources 1, 4)
1What words do people type?
Bars use a log scale: each step to the right is about 10 times more searches, so the small searches stay visible.
The 15 biggest of the 27 searches behind our open questions. Together the 27 add up to 13,270 a month in the US. The three biggest are about the technology, not a vendor: “kv cache” (2,900), “sovereign ai” (2,400) and “ai factory” (1,900). The searches where buyers pick a vendor are smaller, but one buyer here can mean a large project.
2What do buyers ask an AI? 


DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 15 named in all.
DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 7 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 16 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 8 named in all.
DDN is named by 2 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 12 named in all.
DDN is named by 1 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 8 named in all.
DDN is named by 1 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 10 named in all.
DDN is named by 2 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 7 named in all.
DDN is named by 2 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 15 named in all.
DDN is named by 2 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 12 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 21 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 16 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 7 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 17 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 21 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 8 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 9 named in all.
DDN is named by 1 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 21 named in all.
DDN is named by 2 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 19 named in all.
DDN is named by 1 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 6 companies named most, of 6 named in all.
DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 20 named in all.
DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 12 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 7 named in all.
DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 15 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 33 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 18 named in all.
DDN is not named by any of the 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 7 companies named most, of 9 named in all.
DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 4 companies named most, of 4 named in all.
DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 2 companies named most, of 2 named in all.
DDN is named by 3 of 3 AI engines.
The number is the place in the AI’s list (1 = first). Showing the 5 companies named most, of 5 named in all.
Bars use a log scale: each step to the right is about 10 times more searches, so the small searches stay visible.
The 30 questions we put to AI. Press “AI answers” to see who each AI named, in order. On 13 of the 27 questions that do not say “DDN”, no AI names DDN at all.
Where DDN loses buyers
Most of DDN’s real Google visits come from people who already know the name. In AI answers DDN is 4th, and AI reads other sites about storage. (Sources 1, 4)
1Where ddn.com’s Google visits land
Estimated Google visits a month, US, by address (Semrush model, 5.9K in all).
Semrush gives 46% of ddn.com’s visits (2,727 a month) and 6.7K of its 8.2K keywords to welcome.ddn.com. That address no longer works (checked 2 Nov 2026), and its pages were about recipes, calories, news and local places. The real storage site, www.ddn.com, gets about 3,160.
2The searches that bring visits to www.ddn.com
Estimated Google visits a month, US, the 10 biggest searches, with ddn.com’s place on Google. Bars use a log scale, so the small numbers stay visible.
About 88% of these visits come from people who already type DDN’s name. Where buyers search for the thing itself, ddn.com sits lower: 5th for “hpc storage” (14 visits) and 13th for “best sovereign ai server” (6,600 searches, 33 visits).
3The sites AI 

reads before it answers
Links AI showed with its answers to the 27 questions that do not name DDN: 648 links (Google’s own links left out). Bars use a log scale, so the small numbers stay visible.
AI learns about storage from YouTube, Reddit and WEKA’s own site. On the open questions, ddn.com is 4 of the 648 links; weka.io is 26.
4In AI 

answers
We asked Gemini, Google AI Overview and Google AI Mode the 27 questions that do not name DDN: 81 answers. Who got named?
DDN is named in 29 of 81 answers (36%), 4th. WEKA is named in 39 and VAST Data in 37. When the question says “DDN”, AI names it in 9 of 9.
5By type of question
Share of answers that name DDN, by group of questions (each asked on 3 engines).
Strong on vendor comparisons, AI training and HPC. Weak on object storage, inference and RAG, and missing on industry and sovereign AI.
6By buyer
Share of all 90 answers that name DDN, by the buyer who asks.
Heads of HPC see DDN in 3 of 4 answers. The ML platform engineer, who keeps the GPUs fed, sees it in none: 0 of 12.
Fix the first and buyers who do not know DDN yet find it on Google. Fix the second and AI quotes ddn.com, not a rival’s blog.
Who takes your buyers
On the 27 open questions WEKA and VAST Data are named most, then Pure Storage. On the questions DDN misses, AI names MinIO, the big clouds and database companies. (Sources 1, 3, 4)
1Who gets named most by AI 


Answers naming each company, out of 81 (the 27 questions that do not name DDN, on 3 engines).
The 13 questions where no AI names DDN, and who it names
| Question no AI answers with DDN | Searches a month | Who AI names instead |
|---|---|---|
| How does KV cache offload to storage work for AI inference? | 2,900 | vLLM, LMCache, SGLang |
| What storage does sovereign AI need? | 2,400 | WEKA, VAST Data, IBM |
| What is an AI data platform? | 880 | Databricks, Oracle, Snowflake |
| What storage do I need for large language model training? | 590 | IBM, Lustre, NVIDIA |
| Which S3-compatible storage is best for AI? | 590 | Backblaze B2, MinIO, Scality |
| Where should AI teams keep training data and model checkpoints? | 320 | Amazon S3, Google Cloud Storage, Azure Blob Storage |
| What is the best object storage for AI workloads? | 140 | MinIO, Pure Storage, Google Cloud Storage |
| What storage suits genomics and life sciences research? | 90 | WEKA, Lustre, Dell Technologies |
| What storage works for a RAG system with a billion vectors? | 90 | Milvus, Pinecone, Qdrant |
| Pure Storage vs NetApp for AI: which is better? | 90 | Pure Storage, NetApp, VAST Data |
| How do I stop GPUs sitting idle while they wait for data? | 50 | NVIDIA, WebDataset, Kornia |
| What storage do banks and financial firms use for AI and risk analytics? | 20 | Google Cloud Storage, Snowflake, Pure Storage |
| What storage is used for autonomous driving data? | 0 | Azure, VAST Data, WEKA |
DDN, WEKA, VAST Data and Pure Storage on Google (Semrush)
| Site | Authority (0 to 100) | Google visits a month | Keywords | Sites linking in | AI mentions (Semrush) | AI answers naming it (of 81) |
|---|---|---|---|---|---|---|
| vastdata.com | 37 | 10.5K | 3.7K | 9.8K | 243 | 37 |
| ddn.com | 40 | 5.9K | 8.2K | 8.1K | 146 | 29 |
| weka.io | 38 | 5.8K | 3.8K | 6.7K | 256 | 39 |
| purestorage.com | 35 | 3.6K | 1.3K | 13.3K | 774 | 30 |
On Google page one for “DDN vs VAST Data vs WEKA for AI”, TechnologyMatch, Reddit, Gartner Peer Insights and a LinkedIn post hold the places. No ddn.com page is there (checked by hand, 2 Nov 2026).
The gap
Six things the data shows that the SEO and AI program would work on. (Sources 1, 2, 3, 4)
Thirteen questions with no DDN
The biggest: KV cache for AI inference (2,900 searches a month), sovereign AI (2,400), what an AI data platform is (880), S3 storage for AI (590), storage for training large AI models (590). About 8,160 searches a month in all.
One buyer never hears DDN
of answers name DDN for the ML platform engineer. “Where should AI teams keep training data and checkpoints?” and “RAG with a billion vectors” name no DDN, though DDN has a checkpointing paper and a RAG post.
AI reads others about storage
are ddn.com on the open questions. YouTube (49), Reddit (30) and weka.io (26) shape the answers.
No page compares DDN with rivals
in the sitemaps compare DDN with VAST Data or WEKA. Buying pages are short: 217 words on AI as a Service, 368 on the Data Intelligence Platform. Posts are signed “DDN Team”, not a named expert.
The traffic number is off
of the Google visits Semrush gives ddn.com go to welcome.ddn.com, an address that no longer works. Reporting should start from www.ddn.com: about 3,160 visits a month.
Week 1 check
How many enquiries come from search, and from which pages? That is not public: we read it from your analytics and CRM.
The plan: 30, 60 and 90 days
First turn the strongest pages into answers AI can quote for the biggest missing questions. Then comparison pages and proof where AI reads. Then measure against WEKA and VAST Data every month.
Five answer pages, and a baseline
- Write or rewrite five answer pages: KV cache for AI inference, storage for sovereign AI, what an AI data platform is, S3 storage for AI, storage for training large AI models
- Each page opens with a two-line answer, then proof (speed tests, customers, design), then an honest comparison with WEKA and VAST Data
- A named expert signs each page; one fact sheet (speed, scale, customers) feeds every page, so AI reads the same numbers everywhere
- Link the five pages from the home page and the product pages; add an llms.txt file (ddn.com has none)
- Baseline: Google positions for the 30 searches, and reporting set to www.ddn.com
- Report at day 30
Missing buyers, comparisons, proof
- Pages for the ML platform engineer: where to keep training data and checkpoints, keeping GPUs busy, RAG at a billion vectors, genomics research
- Honest comparison pages: DDN vs VAST Data vs WEKA, NetApp alternatives for AI, Lustre vs GPFS vs BeeGFS
- YouTube: pair each Infinia and EXAScaler video with the written answer under it, plus three side-by-side speed videos (your team adds the videos)
- Real people on your team, signed, in the Reddit and review threads where storage is compared
- Run the 30 questions again and report at day 60
Scale and measure
- Run the 30 questions again: AI answers naming DDN against WEKA, VAST Data and Pure Storage, and Google positions
- The next 30 questions: financial services, self-driving cars, genomics, inference at scale, security
- A monthly report: visits to www.ddn.com, enquiries from search, AI answers naming DDN
- Hand the playbook and the monthly report to your team, so the work keeps running
- Run the 30 questions again and report at day 90
What this can bring
Two things we can measure every month from day 1: how often AI names DDN, and how many buyers find ddn.com on Google before they know the name. (Sources 1, 4)
Google visits a month the plan could add
Estimates. Top 3 = searches x 20%, top 10 = searches x 3%. Click rates are our assumptions, not forecasts. Bars use a log scale, so the small numbers stay visible.
What we track every month
| What we track | Today | Where we look |
|---|---|---|
| AI answers naming DDN, on the 27 questions without its name | 29 of 81 (WEKA 39, VAST Data 37) | Xtrusio scan |
| Questions where no AI names DDN | 13 of 30 | Xtrusio scan |
| Answers naming DDN for the ML platform engineer | 0% (head of HPC 75%) | Xtrusio scan |
| Google place for “hpc storage” and “best sovereign ai server” | 5th and 13th | Semrush |
| Google visits to www.ddn.com (US) | about 3,160 a month | Semrush and Google Search Console |
| Enquiries from search | Measured with your team in week 1 | Your analytics and CRM |
Targets are from stated assumptions, not forecasts. DDN already gets part of these visits today. Revenue is not estimated, because the step from a visit to an enquiry is not public.
Let's meet
In 30 minutes we'll walk your team through these findings and the plan, open any of the 90 AI answers on screen, and agree where to start.
Reply to this report to choose a time.
Data sources
Where every number in this report comes from.
Show the sources and the method
1Semrush
Domain Overview, Organic Rankings (subdomains, positions) and Keyword Overview for ddn.com, weka.io, vastdata.com and purestorage.com, US database, desktop, data of 1 Nov 2026, read 2 Nov 2026. Semrush traffic is a model of Google clicks, not DDN’s own analytics.
2ddn.com
Home, solution, product and blog pages, the sitemaps (582 addresses), robots.txt and llms.txt (not found), read 2 Nov 2026 without running scripts.
3Google page one, by hand
Two searches checked by hand on 2 Nov 2026 (US, English, first page only): signals, not a measurement.
4Xtrusio AI scan
30 questions asked to Gemini, Google AI Overview and Google AI Mode on 2 Nov 2026 (US, English): 90 answers, all returned. Collected and scored by the Xtrusio app (xtrusio.com).
Method and full tables, for the curious
Searches a month are Semrush US estimates. Difficulty is Semrush keyword difficulty (0 to 100); n/a means Semrush showed none.
AI answers: an answer counts when its own text names DDN. Three questions name DDN on purpose; the company shares on this page use the other 27 (81 answers). The “by buyer” chart counts all 90 answers, as the Xtrusio app does. “Weka” and “WEKA” are counted as one name. Links: the links AI showed with its answers, Google’s own links left out.
Visit estimates use our click-rate assumptions: about 3% of searches at positions 4 to 10 and about 20% at positions 1 to 3.
What we could not check: who published the welcome.ddn.com pages and when (the address no longer works and the web archive was rate limited); NetApp’s authority and AI data (the Semrush page did not load). ChatGPT and Claude were not part of this scan.
Every one of the 30 questions
| # | Question | Search behind it | Searches a month | Difficulty (0 to 100) | AI engines naming DDN |
|---|---|---|---|---|---|
| 1 | What is the best storage platform for AI training on large GPU clusters? | ai storage | 880 | 58 | 3 of 3 |
| 2 | Which storage works with NVIDIA GPUDirect and DGX SuperPOD systems? | gpu direct storage | 210 | 35 | 3 of 3 |
| 3 | What storage do I need for large language model training? | storage for ai | 590 | 52 | 0 of 3 |
| 4 | How do I stop GPUs sitting idle while they wait for data? | gpu storage | 50 | 22 | 0 of 3 |
| 5 | What is the best high performance storage for AI and HPC? | high performance storage | 210 | 23 | 2 of 3 |
| 6 | What storage does an AI factory need? | ai factory | 1,900 | 51 | 1 of 3 |
| 7 | What is a parallel file system, and which one should I choose? | parallel file system | 110 | 22 | 1 of 3 |
| 8 | Which is better for HPC: Lustre, GPFS or BeeGFS? | lustre file system | 390 | 43 | 2 of 3 |
| 9 | What is the best storage for HPC clusters? | hpc storage | 480 | 18 | 2 of 3 |
| 10 | Which vendor sells commercial Lustre storage with support? | exascaler | 260 | 17 | 2 of 3 |
| 11 | What storage suits genomics and life sciences research? | genomics data storage | 90 | 10 | 0 of 3 |
| 12 | What is the best object storage for AI workloads? | object storage for ai | 140 | 19 | 0 of 3 |
| 13 | What is an AI data platform? | ai data platform | 880 | 44 | 0 of 3 |
| 14 | Which S3-compatible storage is best for AI? | s3 compatible storage | 590 | 39 | 0 of 3 |
| 15 | Where should AI teams keep training data and model checkpoints? | ai data storage | 320 | 50 | 0 of 3 |
| 16 | What storage works for a RAG system with a billion vectors? | rag storage | 90 | 16 | 0 of 3 |
| 17 | How does KV cache offload to storage work for AI inference? | kv cache | 2,900 | 34 | 0 of 3 |
| 18 | What is the best storage for AI inference? | inference storage | 20 | n/a | 1 of 3 |
| 19 | What are the best AI storage solutions for enterprises? | ai storage solutions | 320 | 53 | 2 of 3 |
| 20 | VAST Data vs WEKA: which is better for AI? | vast data vs weka | 30 | n/a | 1 of 3 |
| 21 | Who are the main competitors of VAST Data? | vast data competitors | 50 | 9 | 3 of 3 |
| 22 | What are the best NetApp alternatives for AI workloads? | netapp alternatives | 40 | 11 | 3 of 3 |
| 23 | Pure Storage vs NetApp for AI: which is better? | pure storage vs netapp | 90 | 10 | 0 of 3 |
| 24 | Which companies are the top AI storage vendors? | ai storage companies | 210 | 25 | 3 of 3 |
| 25 | What storage do banks and financial firms use for AI and risk analytics? | financial data storage | 20 | n/a | 0 of 3 |
| 26 | What storage is used for autonomous driving data? | adas data storage | 0 | n/a | 0 of 3 |
| 27 | What storage does sovereign AI need? | sovereign ai | 2,400 | 71 | 0 of 3 |
| 28 | What is DDN Infinia, and is it good? | ddn infinia | 90 | 18 | 3 of 3 |
| 29 | DDN vs VAST Data vs WEKA: which is better for AI? | ddn vs vast | 20 | n/a | 3 of 3 |
| 30 | Is DDN EXAScaler good for AI and HPC? | ddn exascaler | 30 | n/a | 3 of 3 |