PCR Tech and IT retail, distribution and vendor news
Caeves is a young company, but its founders are anything but new to the world of enterprise storage. They have a history that stretches back more than two decades, through earlier ventures such as Talon Storage, later acquired by NetApp.
The Caeves team has spent years working on distributed storage, file systems, caching, and cloud data services. Now, they are trying to solve a problem they believe the industry has ignored for too long: the rising cost of storing huge volumes of unstructured data, and the difficulty of making that data usable for modern AI‑driven workflows.
“I thought we would all be flying around in jet packs by now, but we’re still talking about storage,” Head of Product and Customer Experience, Jaap van Duijvenbode says, “and I thought storage would be free as a commodity, but it’s not.”
Caeves’ pitch is straightforward: cheap, deep, easy storage that keeps decades of archived data accessible, searchable, and ready for AI tools without forcing organisations to overhaul their entire infrastructure.
Van Duijvenbode describes himself as “a technologist” with 30 years in IT, starting with “managing Novell servers, pulling cables, crunching coaxial cables.” He later moved into enterprise storage and distributed environments, helping organisations consolidate sprawling file systems long before the cloud era.
Caeves’ founding team includes several long‑time collaborators, especially Sharish Phatak. “Sharish and I have worked together for about 20 years. Sharish was basically the inventor of the storage caching over IP protocol,” he says. Their previous company, Talon Storage, focused on edge caching and file consolidation. NetApp acquired it in 2020, and the team spent several years inside NetApp’s cloud data services group before leaving to start Caeves in late 2024.
The company now has around 25 to 31 employees across the US, the Netherlands, and India. “We launched our product in November last year, and we have really great adoption from Microsoft, as well as the Microsoft ecosystem,” van Duijvenbode says.
The company’s core argument is that organisations are drowning in unstructured data they rarely touch, but must keep for compliance, legal, or business reasons. “We know that 80 to 90% of data is unstructured – 85% of unstructured data is only created once. It’s dark and untouched by AI or analytics,” van Duijvenbode says.
He describes a customer still running “48 Amazon storage gateways… 1.8 petabytes… I’m paying for it, but I have no clue what’s inside there.”
This, he argued, is the industry’s iceberg problem: the vast majority of data sits below the surface, expensive to store and difficult to access. “Data is often locked away in silos, but your forgotten data could hold tomorrow’s breakthroughs, and it deserves to be alive,” he said.
Caeves’ view is that organisations need a data foundation before they can realistically adopt AI. “They shouldn’t be forced to put all their data into AI; they should create a data foundation first,” he said.
Caeves is built on Microsoft Azure, using Azure Blob Storage as the deep‑storage layer. The company provides a Windows‑based software layer that presents a normal SMB/NFS file system to users and applications, while automatically tiering data into object storage behind the scenes.
The system uses caching, tiering, and stubbing to move data into Azure’s hot, cool, cold, or archive tiers. Metadata is kept locally, so files remain visible and searchable even when the content is stored cheaply in the cloud.
The company also built a connector that indexes this data into Microsoft Graph, enabling Microsoft 365 Search and Copilot to query decades‑old files. “This data could have been sitting in your archives for 30 years, but now you have it readily accessible,” he says.
The company emphasises that permissions flow through the entire pipeline. “All of the permissions cascade through object storage, so that if I’m searching for something I don’t have permission to, I’m not getting any results,” he said.
Caeves positions itself as a cost‑reduction tool. Many customers, van Duijvenbode says, are paying “eight to 10 cents per gigabyte per month” for data that is rarely accessed.
By contrast, he claims Caeves customers can reach “one or two cents per gigabyte per month, total cost of ownership, including Caeves licensing.”
The pricing model is consumption‑based, with the first 5TB free. “We price based on consumed capacity,” he says. “The more you use, the more cost‑efficient it’s going to be.”
A major question about using US-based companies is US jurisdiction over data. Van Duijvenbode doesn’t dismiss this concern. “Data residency is one thing, data sovereignty is another,” he says, while acknowledging that the US Cloud Act “spans across many borders” and that European organisations are increasingly wary of US‑based providers.
Caeves’ answer is portability. “The software should be portable. It should be able to run in an on‑premises data centre, in traditional object storage environments, even dark‑site operations,” he says.
With this in mind, while the company is currently Azure‑first, it plans to expand to S3‑compatible and full on‑prem object stores.
Currently, Caeves can run fully in Azure, with compute and storage co-located; hybrid, with a Caeves instance on‑prem and object storage in Azure; and Edge‑to‑cloud, with small branch‑office deployments syncing to object storage.
“Think of the Caeves instance as a scratch pad – it’s there as a migration ingestion point, and if you don’t need it, you shut it down,” van Duijvenbode says.
Caeves sees itself competing less with primary storage vendors and more with deep‑storage and file‑tiering platforms. Van Duijvenbode mentions NetApp, Nasuni, Panzura, and others, but argues that many existing solutions still rely on expensive hot tiers or proprietary file systems.
“We believe a lot of source data sits on primary storage that shouldn’t be sitting there,” he says.
He also notes that some cloud‑native storage gateways leave customers locked into specific vendors. “Disparate data sources often equal vendor lock‑in, and organisations want to avoid that,” he says.
Caeves’ approach, he argues, gives customers the option to walk away. “If we do not provide enough value, then the customer can choose to walk away, and we’re willing to take that risk,” he says.
For the channel, Caeves is positioning itself as a partner‑friendly platform. The company is already in the Microsoft Marketplace and is working toward CSP resale support. “The partner ecosystem is in development – the channel allows you to grow and scale,” van Duijvenbode says.
He also sees potential for OEM‑style integrations in future, especially as the company expands beyond Azure. But he was clear that Caeves is still early in its journey: “We’re just getting out of startup, so we’re not ready for that yet,” he says.
Van Duijvenbode repeatedly stresses that Caeves is not trying to sell AI magic. “We’re not going to talk about quantum computing or very tough AI stuff, I want to keep it practical,” he says.
The company’s message is that AI is only useful if organisations can access and understand their historical data. “The winners will make all their data intelligent,” he says. But only after building the right foundation.
He references Nvidia’s Jensen Huang to reinforce the point: “Unstructured data has been completely useless to the world – we cannot query it – we cannot search for it.” Caeves is trying to change that, slowly, pragmatically, and with a focus on cost.
