After about three years of development, Carbon Relay says it’s ready to help data center operators use artificial intelligence to optimize power consumption, reduce energy costs, and predict and prevent outages.
The AI startup came out of stealth Tuesday to announce a $5 million Series A funding round from Foxconn Technology Group and four private-individual investors.
The Boston-based company, founded in 2015, also announced release of its software, which continually monitors cooling systems and other data center equipment and can automatically adjust temperatures and other variables to improve energy efficiency and detect operational anomalies to avoid downtime.
“Our goal is to not only drive efficiencies and thereby reduce cost as it relates to electricity, but also to predict and get ahead of failures,” said Carbon Relay CEO and co-founder Matt Provo.
Carbon Relay joins a growing number of startups and established companies that are building AI capabilities into software and services to improve data center efficiency and their overall effectiveness. AI has the potential to transform data center management, from forecasting demand and managing IT workloads to ensuring uptime and controlling costs.
Companies including Schneider Electric, Eaton, and Nlyte Software have all launched what they say are AI-powered data center management as a service (DMaaS) offerings, while Maya Heat Transfer Technologies says it has added AI into its on-premises data center infrastructure management (DCIM) software.
To compete successfully, Carbon Relay and other specialist data center AI startups, such as California Data Science, Litbit, and Solecular, will need to partner with these DMaaS and DCIM providers, because their technologies are complementary, said Rhonda Ascierto, VP of research at the Uptime Institute.
Carbon Relay offers compelling technology, she said, but the company – along with the other startups – needs to emulate the strategy of Vigilent, an established data center software company that uses machine learning to optimize cooling.
Vigilent, which launched in 2008, has partnerships with companies like Schneider, Siemens, and NTT Facilities and has its technology installed in 600 data centers and other facilities worldwide, she said.
“Carbon Relay is not offering DCIM and not necessarily offering a full-stack DMaaS service. They are a specialist AI data center provider, and their technology is designed to be complementary to these other offerings,” Ascierto said. “In order to penetrate the market, they will need to partner with one of those larger suppliers.”
Carbon Relay’s software does integrate with DMaaS and DCIM vendors’ products, and the company plans to announce strategic partnerships this year, Provo said. In the meantime, it offers standalone software solutions.
Optimization and Prediction
Carbon Relay announced commercial availability of two software products. Optimize uses an AI agent, APIs, and connections to IoT sensors to ingest and analyze data from customer data centers and automatically controls HVAC systems and IT equipment. A second software product called Predict performs the same analytics, but without the automation. Both applications are available on-premises and in the cloud, Provo said.
“We are flexible with how much control we require the client to give over to the AI agent,” he said. “If they are not as comfortable yet, they can get Predict with a sprinkle of Optimize at first. And over time, as they get more comfortable, they will ultimately use the Optimize solution.”
While Carbon Relay’s technology does do anomaly detection and prevent failures of IT and cooling equipment in data centers, the company is focusing its initial marketing effort on helping data center operators improve energy efficiency and reduce their carbon footprint.
Its software, which it says leverages advanced data science, deep reinforcement learning, and augmented intelligence, can track and control 37 variables in HVAC and IT equipment, including humidity, temperature, and the electrical load of CPUs, fans, and lights, Provo said.
Unlike existing solutions that require six to 12 months of historical data, Carbon Relay says it can create a fully simulated version of a data center within 48 hours by analyzing the facility’s blueprints and electrical layout diagrams. The technology can improve energy efficiency by 13 to 25 percent, Provo said.
“We can build a simulation quickly, train the AI agent to navigate that physical environment, and ultimately control the machinery,” he said. “Within 48 hours, you will see a very clear indication of your savings potential, and then across an addition eight to 10 weeks the system gets fully trained, and you will really start seeing significant savings that consistently grows across a year or two of operating in any environment.”
Temperature, humidity, and other variables can shift every few seconds. The software can automatically regulate the different variables to get efficiency gains fast without human intervention.
“These are small, but continuous changes. It could be 100ths of a degree here or shifting some computing process to another part of the data center,” he said.
As an example of its anomaly detection capabilities, Carbon Relay’s software can identify a hard drive that’s about to reach capacity and automatically move data to another drive with available space, Provo said.
“We utilize intelligent forecasting techniques to predict and prevent failure,” he said.