When the large cloud providers have excess compute capacity, they tend to discount it through programs like AWS’s and Azure’s spot instances. Any time a server idles, it isn’t making the company money, after all. NodeShift aims to take this concept and expand it well beyond the big clouds — and with stronger guarantees — by providing a single API for access to excess compute, storage and graphics accelerators from independent data center operators and through connections to low-cost decentralized web services like Akash and Filecoin.
NodeShift today announced that it has raised a $3.2 million seed funding round led by Inovo.vc, which focuses on startups based in Central and Eastern Europe, with participation from Notion Capital, 10x Founders and Kestrel0x1.
The company was founded by Andrey Surkov and Mihai Mărcuță. They first met when they were interns at Cisco back in 2016. While Surkov stayed at Cisco (and also was in crypto at one point), Mărcuță went on to work at Microsoft Azure, Twitter and Epic Games out of London for the next few years. They remained friends, though, and the project came to be when they accompanied another friend of theirs to Turkey, where he was getting a hair transplant. While in the recovery room, Surkov shared his idea for a company that would make this excess data center capacity available to developers. How is that for a founding story?
“A lot of data centers have spare capacity just sitting out there — about 10 to 20% of spare capacity just sitting there — and there’s hundreds of data centers like this,” Mărcuță explained. “The price is very, very affordable. If you compare it to traditional cloud providers, we’re talking about prices that are 70 to 80% cheaper.”
NodeShift promises that it can save its users well over 70% in compute costs compared to the large cloud providers. That includes access to Nvidia accelerators like the ever-in-demand A100 GPUs at a significant discount. And while the large cloud vendors can’t offer it, the web3 services do also offer access to high-end consumer-grade accelerators like the RTX 4090 gaming GPUs, which may be slower but also far more affordable.
Because of his experience with networking at Cisco and decentralized systems in crypto, Surkov had become interested in why there is so much friction in using decentralized compute. “I wanted to understand how I can host my DAP fully decentralized and then I understood that there is a huge amount of friction. Hosting fully decentralized storage is like one thing. Compute is another thing — and then you combine this all together and remove the friction of the tokens. You need to get the token for each of the crypto projects just to fund infrastructure. It’s a huge amount of friction,” Surkov said.
Indeed, I would assume it’s hard to convince a CFO to allow a developer team to buy a random cryptocurrency even if it’s just to experiment with these services. So even though the prices may be great, using these services becomes to cumbersome in a business context.
It’s worth noting that the founders are very clear-eyed about using web3 and blockchain technologies. “We personally feel like web3 and blockchains oftentimes goes into projects where you don’t necessarily need them just for the sake of having that,” Mărcuță said.
In addition to these decentralized projects, NodeShift has also set up deals with independent data center operators. The team stressed that developers get to choose exactly where their projects are located (down to the data center) and that these data centers were certified by the Uptime Institute and have all of the standard SOC 2 and ISO 27000 certifications an enterprise user would expect. When using those data centers, the company is also able to offer those customers an SLA — something that would be hard to do (or rather costly) in a web3 context.
What this combination of traditional infrastructure and web3 allows for, though, is bursting out capacity to those projects when needed and at a low price, turning them into what is essentially a spot instance on NodeShift. Just on the data center side, the company says, it currently has access to about 400,000 CPUs and 28 million terabytes of storage.
Soon, NodeShift will launch a Kubernetes platform that will sit on top of all of this, making it easy to shift workloads around as needed.
On top of that, the team plans to use the funding to build out its supply, as well as its go-to-market efforts.
Some of the company’s competitors include the likes of Germany-based Impossible Cloud, though that team focuses only on storage, and Salad, which focuses heavily on AI/ML workloads on consumer-grade hardware.
Comment