‘Last year, we saw 1,000 percent growth in our sales. We are growing very fast. [An IPO] could be soon. And I believe that we need to be a public company as soon as possible because we sell to very large companies,’ says Hammerspace founder and CEO David Flynn.

Hammerspace, developer of a high-performance platform for storing, managing and migrating data for AI, this week unveiled a massive $100 million round of funding.

The funding round is unique in this industry in that it was led not by venture capital companies but by investors normally focused on public companies, said Hammerspace founder and CEO David Flynn.

“What attracts me is I’m working with long-term investors who will be supportive and hold or even accumulate a bigger position once we IPO, unlike the pure venture funds where IPO is the end of the journey and they want to sell,” Flynn told CRN.

[Related: The 2025 Storage 100]

That long-term view is also important as Hammerspace looks at holding an IPO, a move Flynn said he would like to see happen in the relatively near future.

“I believe that we need to be a public company as soon as possible because we sell to very large companies,” he said. “After Snowflake’s IPO, there literally was an inflection point [of high growth] once they became public. The same happened with Data Domain. When you sell to large enterprises, they feel much more comfortable buying from you when you’re a public company. It shows that you’re there to stay. So we anticipate needing to be a public company to help accelerate growth and doing that in fairly short order.”

In the meantime, Hammerspace will continue to focus on developing the high-performance storage technology that Flynn said is needed as AI goes mainstream.

“The thing about AI is it demands performance,” he said.” People don’t have time to build out exotic HPC [high-performance computing] file systems. … Because AI is going mainstream, there’s not enough talent on the planet for everybody to be able to install it and make it work. So it has to be plug-and-play, easy to use.”

There’s a lot going on at Hammerspace and the quest for AI-focused storage. Here is more of CRN’s discussion with Flynn.

How do you define Hammerspace?

We’re a high-performance data platform. We speed up the use of data in AI systems, and uniquely we do that at every stage, from the installation, setup and configuration to getting data out of existing storage systems, to orchestrating the data across data centers and then feeding the high-performance data to GPUs, and even positioning the data into the GPU servers for local access. It’s a data platform that accelerates data inside of storage. We accelerate that data to be available at unprecedented levels of performance that are necessary for GPU computing.

This isn’t just a high-performance file system, so to speak. This is uniquely the only one that’s actually native to Linux, so you don’t have to install any software. It’s the only one that can use your existing storage. It’s the only one that can use the local flash inside of the GPU servers as part of it. So we can incorporate your existing storage. We can incorporate the storage inside of GPU servers without deploying any special software. It uses native Linux because we are the first to be using the new parallel NFS [network file system] that we uniquely built in the standard and into the kernel. My CTO is the kernel maintainer of the NFS stack. We enhanced NFS, and that allows us to turn NFS into the first-ever high-performance HPC-class parallel file system that’s native to Linux. That’s why I can make the claim that, unlike anything on the planet, we are much easier to deploy because you can come to the table with your existing storage and with your existing Linux, and you don’t have to add any software or new storage. Our system can then orchestrate data and move it across systems and across data centers from behind a single pane of glass and namespace. We can create a globally unified global file system. That means you don’t have to wait to copy things around. There’s no downtime, not even to get that data into Hammerspace. You can simply point to the existing data in your storage, and we can immediately start serving it without waiting, even to scan all of the metadata. …

Meta has been using this for building Llama. They’ve blogged about that. Other big players in the AI arms race building foundation models have used Hammerspace. Numerous companies use Hammerspace not just for AI and GPU but for running their businesses. It has been a dream in the industry to be able to separate or abstract data from the storage and make data available everywhere with the utmost crazy performance levels and from a single namespace.

You mentioned working with GPUs, but the technology wasn’t originally developed with GPUs in mind, right?

Actually, some of our earliest customers were in the media and entertainment space doing rendering and special effects and animation. They were using GPUs before they were a sexy part of AI. So yes, our technology was designed for GPUs but focused on doing special effects animation rendering as one of many use cases. It works not just with GPUs, but any data-intensive or HPC kind of workloads, or anything that requires high performance. Data feeding AI just happens to be what’s pushing that into the mainstream.

What did you have to do to make the technology AI-ready?

Beyond the feat of having a parallel, scalable file system, the thing about AI is it demands performance. People don’t have time to build out exotic HPC file systems. We’ve had file systems made to be super-fast, like Luster that was created by Los Alamos or IBM’s GPFS, for supporting HPC. But because AI is going mainstream, there’s not enough talent on the planet for everybody to be able to install it and make it work. So it has to be plug-and-play, easy to use. It needs the convenience of enterprise NAS with the performance of these HPC file systems. Everybody knows how to configure and run NFS, and with Hammerspace you can have that convenience and yet get the scalable performance that used to require those exotic file systems to get. AI going mainstream means this has to be convenient. It has to just work. We can’t have everybody building out their own science project to make it work.

There’s an interesting story behind the name ‘Hammerspace.’ Talk about that.

A ‘hammerspace’ is an extra-dimensional universe where things can be instantly retrieved and is infinite in size. Think of it as when a magician pulls something out of the hat. Where did it come from? That was hammerspace. We use the concept of hammerspace to represent data being pulled out of thin air. It’s infinitely fast because it’s right there, and it’s infinite in size, and it’s available from wherever you go. It has the characteristic of decoupling data from the physical world, just like the hammerspace concept. One way to think of it is, hyperspace is sci-fi for traveling faster than the speed of light by using an alternate universe, right? ‘Hammerspace’ is for storage where hyperspace is for travel.

We didn’t create the term. It’s a fan fiction invention. It was actually talked about extensively in the ‘Spider-Man Into the Spider-Verse’ cartoon. Our company was founded well before that. So it was great to have Disney come out with ‘Spider-Man Into the Spider-Verse’ and talk about hammerspace. … The term originated from some Japanese cartoons where a character would pull a big mallet out of nowhere to whack another character. And the joke was, where did that come from? Where was he carrying that big hammer? And so they started saying, ‘Oh, it was in hammerspace, the space for carrying hammers in this alternate universe.’ It’s like you are gaming your first-person shooter games when you’re carrying tons of massive weapons. How can you fit all of that? Well, it’s in a magic pack that lets you fit an infinite amount of stuff in a finite amount of space and carry it with you in a weightless kind of way. That’s the way we think about it. Long story short, it is data not bound by data gravity, not stuck in storage. That data lives in hammerspace because it’s no longer a prisoner to its own mass or to data gravity.

Talk about the big investment in Hammerspace.

Hammerspace in the past was somewhat uniquely funded. My success with my former company, Fusion-IO, allowed me to fund Hammerspace myself in the early stages. I personally invested nearly $20 million of my own capital to get the company off the ground. We have taken many partners along the way. Some of those strategic partners, like Saudi Aramco, led an investment. But this is the first round of investment in Hammerspace that’s led by a purely financial institution. And it’s not just any. It’s a major, major player in the AI world, Altimeter Capital, founded by Brad Gerstner. They led this new $100 million investment round. That’s not a small amount of capital. And in a way, it’s technically our B round. In addition, there’s Cathie Wood and ARK Invest along with Millennium Capital. The unique thing is that all three of these—Altimeter, ARK and Millennium—are actually in the public markets. They venture into the venture capital world only as a side part of their businesses. They’re not venture-first. They’re public-company-first. These are their crossover funds.

That’s the exciting part for us. They’re focused more on investing in the public markets, which means they would be supportive through an IPO. Pure venture guys tend to want to sell at an IPO, and that leads to shareholder turnover. I like the fact that these guys are public company investors because that makes them buyers when we have our IPO, and that’s a good thing. They’re looking at it for the long term, not just to get to an IPO and then sell.

Unlike traditional venture capital, all of these players have major stakes in public companies. Most of their trading is in public companies, and so that’s sort of the starting gate. They’re reaching back into the private world just to get a jump on public investments. And what attracts me is I’m working with long-term investors who will be supportive and hold or even accumulate a bigger position once we IPO, unlike the pure venture funds, where IPO is the end of the journey and they want to sell.

Any idea when Hammerspace might be ready for an IPO?

At our growth rate? It could be soon. And that’s really all I could say. Last year, we saw 1,000 percent growth in our sales. We are growing very fast. It could be soon. And I believe that we need to be a public company as soon as possible because we sell to very large companies. After Snowflake’s IPO, there literally was an inflection point [of high growth] once they became public. The same happened with Data Domain. When you sell to large enterprises, they feel much more comfortable buying from you when you’re a public company. It shows that you’re there to stay. So we anticipate needing to be a public company to help accelerate growth and doing that in fairly short order, assuming the whole market doesn’t melt down with tariffs and everything.

Are you seeing any impact from the tariffs? Your company has a hardware component, right?

There is. And we’re always fighting delays in getting hardware delivery in, and that is likely to be exacerbated by all of this. But I can’t say we’ve seen it yet. Impacts will happen if tariffs persist, but it hasn’t yet at this point.

Have you had to change your pricing because of the tariffs yet?

No, we have not. We’re a U.S.-based company, so U.S. [hardware] doesn’t matter. The reciprocal tariffs in China maybe will come into play but hasn’t yet. [And] let me point out that we don’t sell hardware. We’re a software-only company. That gives us a lot of flexibility. But the tech does need to run on hardware, and so hardware supply chains can stall deployments. As a software-defined storage company, you’ve got to wait for the hardware. So we don’t have to worry about the hardware cost side of it, but it becomes an issue for our customers.

What part of Hammerspace’s business comes from indirect channels?

All of our business goes through the channel. We do not sell direct. We are like VMware. This is a new paradigm. VMware allows you to manage servers in the virtual. We allow you to work with data abstracted from storage and, like VMware, that’s a new paradigm. It requires that people rethink how they manage servers. Instead of racking and stacking and feeding them CD-ROMs, they could now use vSphere in software. The same thing is true for working with storage data through Hammerspace. Why do I bring that up in the context of the channel? Because the channel is where you can leverage that evangelism. Once partners understand the power of this, then it gets propagated because they bring that into their customer base. So we are focused on being extremely channel-friendly and making sure all deals go through the channel.

Is Hammerspace profitable yet?

I don’t think we’re commenting on that. That would be a major announcement. So let’s right now say we are definitely predisposed for growth. And bringing in a war chest with this investment will definitely help set the stage for further growth.

What are your strategic priorities for 2025?

The strategic priorities for 2025 really drive home the fact that you can have your cake and eat it too. That performance is not just about delivering data from storage into GPUs. It’s about everything from getting the environment up and running—the installation, the ease of use, the plug and play, the standards, the native and built-in—through sourcing existing data using existing storage systems and automating the movement of data across those systems. People think performance is about moving storage to compute. What we’re saying is, performance is how long it takes you to get things going, how long you spend, how much time you waste setting it up, configuring it, moving data, all of that. So our mission this year is to help people understand there’s a better way than the manual logistics of setting up exotic file systems, copying data sets from here to there, and so forth. This is, again, a new paradigm for working with data in the virtual to where it can move freely while you’re using it through the swipe of a mouse, but behind a single pane of glass. So just getting the message out there that what we do actually works.

Companies that had this dream in the past failed because it kills your performance. We had to introduce new protocols to make it work. This is very sophisticated stuff. So my main mission in the next 12 months is to get the word out that you can have your cake and eat it too. You can solve the data logistics and have high performance at the same time. …

Our new capital is mainly about the go-to-market reach and getting our vision of the future across, where this stuff can be automated and it can be very high performance and speed things up at every stage.