Cloud vs Edge AI

Cloud vs Edge AI

What’s Best for Your Facility? by Sastry Malladi, Vice President and Head of OpenBlue AI, Johnson Controls and Francisco Ruiz, Global Director of IoT & Infrastructure Strategist, Oracle

Building managers are integrating smart technology into the properties they’re responsible for faster than ever before. According to Juniper Research, the number of smart buildings worldwide will grow by 150% by 2026, from 45 million buildings this year to more than 115 million. There’s a solid reason for the steep rise in deployment. Cutting-edge automation software and systems provide the opportunity for building owners to continuously monitor operational parameters like occupancy numbers, indoor air quality (IAQ) and utility use to help achieve unprecedented safety and efficiencies.

However, integrating smart technology into a facility can make some building managers feel uneasy. The decisions that must be made when adopting automation systems are complex and may include elements unfamiliar to them. But just like they mastered HVAC, lighting controls and chillers, building managers can come to know internet of things (IoT), networking and artificial intelligence (AI).  

An AI-enabled IoT (AIoT) system can be especially intimidating, but it can be one of the most powerful ways to maximize building efficiency, safety and sustainability. AI can be applied either at the edge (Edge AI) or in the cloud (cloud AI). Both have their advantages depending on the application’s goals and needs, and building managers who understand when to use which one – or a combination of both – are at an advantage.

Understanding the difference between remote and on-premise storage

The AI now being deployed started life as a cloud computing technology. The machine learning algorithms under the hood of these systems require significant computing power, both to train the algorithms and to call on them to deliver insights – a process called inferencing. Until recently, on-premises infrastructure rarely had the resources to effectively do those things and, as such, building operators had to run their AI applications out of data centers.

Yet, running smart building applications out of remote data centers has its own limitations. Connectivity, bandwidth costs, security and latency – the time it takes to send data to the cloud and back – can impact a system’s efficacy. If a machine, or an entire building automation system, is going to fail, the alarm and automated response need to be as immediate as possible.

That issue has largely been mitigated by a new generation of edge computing technology: infrastructure installed in facilities with the processing power demanded by these compute-intensive workloads.

Companies like FogHorn, founded seven years ago, have developed an Edge AI technology that creates new possibilities to digitally transform building operations. This includes advanced technology (known as Edgification) to optimize AI models to run efficiently on low-cost edge compute devices. Johnson Controls acquired FogHorn at the start of 2022 and has now integrated the edge technology into its OpenBlue platform.

By closing the on-premises capability gap, edge devices provide an architectural component important for achieving the goal of running a building as efficiently and effectively as possible.

Choosing between the cloud and Edge AI

With the availability of Edge AI, building managers thinking about implementing smart automation technology now almost inevitably confront the question of whether to deploy AI on-premises or in the cloud. For those facing this question, there are some simple rules of thumb to consider.

Edge AI is best when:

  • Actions need to be executed in real-time, or close to it. Smart automation systems that detect operational problems and automatically alert or respond to them tend to work best when latency is minimized as much as possible.
  • Local control of a system is required. Turning off a machine or adjusting a control system from the cloud often runs into security and latency challenges.
  • There are limitations to data transit and storage costs. Take for example a video monitoring system in which high-fidelity images from multiple cameras are analyzed by a computer vision AI model, a popular AI application. Sending to and storing all that data in the cloud can quickly become cost prohibitive.

The cloud may be better when:

  • Completing rigorous data analysis. Often building managers want a deeper understanding of how they’re operating based on AI analytics, or to run simulation exercises on a ‘digital twin’ version of their facilities. That kind of data analysis typically doesn’t need to happen in real time, so it’s best executed in the cloud, where the managers can harness at any scale the most powerful hardware and software tools for the job.

A combination of both may be best when:

  • Running multiple buildings and corelating information between them. The cloud allows for a centralized data clearinghouse and command center. As a practical matter, a hybrid approach is typically employed where some initial processing in the individual buildings happens through Edge AI and then cloud AI is run on the aggregated data from multiple buildings, possibly combining other data sources.

Taking the first steps to adopt AI

It’s important to remember these are decisions that building managers don’t need to make alone – there are expert technology vendors who can ensure AI is deployed where it will serve your unique needs best. Building managers don’t need to be data scientists and fully understand all aspects of the AI and its underlying machine learning algorithms, but rather can partner with expert technology vendors and allow the AI to do its magic behind the scenes.

Oracle, like many organizations now initiating return-to-work policies at a large scale, sees the aftermath of the pandemic as a unique moment in which to introduce smart building systems. After a couple years of pandemic-induced closures, employees insist on a physical workplace where amenities are at their fingertips, collaboration tools are ubiquitous, air quality is monitored, crowding is limited, and their companies are meeting sustainability goals in their use of energy and water and reduction in waste. And with buildings still at historically low occupation rates, powering off systems that don’t need to be running helps deliver considerable gains in efficiency. 

These shifting workplace dynamics and expectations can be an opportunity to assess new investments in IoT technologies, the advanced networks that connect them, and the AI systems that control them. It’s also an opportunity to evolve workplaces to make better decisions based on occupancy, employee experience needs, ownership of the sites and their use case criticality (e.g., a research lab compared to office space).

When deciding whether to invest in automated control systems, building managers have historically prioritized schedules. Not anymore. The new key consideration is utilization metrics. They can’t take for granted that everybody will come back, and many companies are adopting hybrid work policies.

Creating smarter, safer and more sustainable spaces

For the first time, the office needs to compete against the home as an attractive and productive work environment. People want to feel confident knowing that the indoor air quality (IAQ) in their office is being monitored, resources like water and energy are being used efficiently and the rooms they occupy are comfortable. AIoT systems can help make buildings more energy efficient, healthy, autonomous, secure and responsive to occupant needs.

In response, new and seasoned building managers are seeking the support of smart technology providers to help them acquire the new skills required to implement AIoT automation systems and optimize their operations. One valuable lesson is when to deploy AI on premises or in the cloud. Once they’ve determined whether the edge or cloud AI is aligned with their building goals and application needs, informed building managers can be confident that AI will help them ensure healthy air, comfortable spaces and efficient operations that can help refill their buildings.

Contact Us

MEB Media Publishing (UK) Ltd

13 Princess Street,


ME14 1UR

United Kingdom

Our sister publications

Campus Estate Management Magazine


In Security Magazine