Kubernetes software startup Spectro Cloud Inc. has bagged $100 million in a late-stage funding round as it looks to solve a significant and growing problem in artificial intelligence.

Powerful processors might be hard to come by, but they are available for a price. However, what’s really lacking is the software needed to squeeze the most value out of those chips.

Spectro Cloud, founded by Chief Executive Tenry Fu and Chief Technology Officer Saad Malik in 2019, sells a software platform that helps organizations to deploy, govern and operate AI infrastructure at scale, across data centers, edge locations and on-premises.

Today’s oversubscribed Series D round was led by Growth Equity at Goldman Sachs Alternatives and saw participation from AMD Ventures, Ericsson, LG Technology Ventures and Maximus. It brings Spectro’s total amount raised to date to over $260 million.

Spectro says that enterprises are spending billions of dollars on AI hardware, yet despite these massive expenditures, many are struggling to get their AI projects up and running at full speed. According to Gartner Inc., global AI spending will reach $2.59 trillion this year, up 47% from just 12 months ago. But the vast majority of that cash is going towards hardware, with the software that sits between the AI chips and the working AI applications accounting for only a fraction of that amount.

AI infrastructure headaches

The challenge Spectro wants to solve is that running AI in production involves more than just installing applications on powerful servers. The graphics processing units and other chips that live inside those servers require coordinated management, as well as storage, networking, security policies and governance rules, and these must be extended across multiple physical locations and cloud environments. For businesses in regulated industries such as healthcare, defense, manufacturing and telecommunications, their strict regulatory requirements make AI infrastructure management even more difficult. Most rely on the open-source Kubernetes system to manage these environments, but it can be difficult to scale to the degree necessary to support AI workloads.

This is where Spectro Cloud believes it can make a difference. Its main product is the Palette platform, which acts as a single control plane for information technology teams to deploy and manage complex computing environments, no matter if they exist in the public cloud, on bare metal servers, in private data centers or edge locations such as retail stores and oil rigs. The company also sells PaletteAI, which is a new service that debuted in October, expanding the Palette platform to cover AI infrastructure use cases such as GPU management and distributed inference.

Spectro’s software acts like an all-encompassing management layer that sits atop of all of the computing environments and hardware used by an organization. IT engineers use the software to configure the hardware, networking, storage and all of the security and policy rules, and then they can monitor everything when it’s up and running. It does this regardless of where the physical hardware lives.

Fu said software is often neglected, yet it’s the secret sauce that transforms expensive AI infrastructure into business outcomes. “No two customers are starting from the same place,” he said. “Some are modernizing legacy infrastructure, some are scaling edge or Kubernetes operations and others are building AI factories or neocloud services. Spectro Cloud gives them one consistent platform to manage that complexity and adapt AI faster without losing control.”

Future plans

With regards to the money raised, Spectro has three priorities. The first is to expand the PaletteAI platform and enhance its functionality so companies can get more out of their GPUs and keep a lid on the costs of running AI models while scaling production. Second, it’s targeting geographic and market expansion. It’s particularly interested in expanding to neocloud and sovereign cloud providers, which are regionally focused cloud infrastructure services that enable companies to keep their AI and data local.

The third priority is all about deepening Spectro’s integration with the hardware it wants to manage, which means collaborating with chipmakers, server makers and systems integrators.

The participation of Advanced Micro Devices Inc.’s investment arm is key to that last priority. With AI inference, or the process of running trained AI models to generate predictions and answers, rapidly becoming the core workload for most organizations, companies need to be able to run their AI applications on a whole bunch of different processors and server hardware, as opposed to using only Nvidia Corp.’s GPUs.

“As AI moves into production, inference is becoming one of the most important drivers of infrastructure demand,” said Patrick Rundell of AMD Ventures. “Spectro Cloud’s platform approach addresses a critical challenge for enterprises deploying production inference workloads at scale.”

Images: Spectro Cloud