Oh no, not another article about AI, you may be thinking. But this time, we're addressing where cloud computing and AI converge, and what that means for us as system integrators in the Pro AV space in our mission to provide the best service to our clients.
First, some definitions so we’re all speaking the same language. The classical definition of cloud is a method of doing compute. Cloud refers to a group of computers working in concert, and that term has been around long before Amazon or Google existed. For systems integrators today, it’s important to return to that classical definition, because cloud means a group of computers working in a coordinated way that could be on premise, in a co-location, in a client's data center, or in a public hyperscaler or a neoscaler (tailored equipment and software to a particular clients' needs).
In the classic sense of cloud compute, up until now we’ve been talking about methods for transforming signal workflows and workloads from dedicated hardware to software running in compute. As the capabilities of compute have continued to evolve and we’ve moved to software-defined infrastructures, we could be running them in a hyperscaler, on our client's premises, or in a third-party data center.
This opens the door for us to begin to add services that would not be possible in a traditional SDI, HDMI, or NDI-based hardware environment—and where the application of AI is both useful to our clients and commercially viable today. Much of what is written about AI is still very forward facing. However, today we see applications that we can install for our clients in useful ways.
AI Tech Deployable Today
This is a turning point in the system integration world. Previously, much of our talk about what AI might or might not do has been very specific to a vendor embedding a technology in a device, such as tracking a speaker with a camera. But, today, via the work that NVIDIA and others have done, we have practical applications that can run in real-time systems to assist our clients now. These applications are not only useful to our clients but create recurring revenue for us and help spread the cost as an OpEx load.
Of the low hanging fruit today are tools for closed captioning, translation, and localization. All of us who’ve installed systems for multinational corporations have been grappling with these topics for decades, starting as requirements around captioning, then translation and localization. AI technology has matured to the point where it is now deployable in our clients’ spaces running on local hardware, which eliminates many of the security concerns that previously existed.
Through the NVIDIA Holoscan for Media initiative, AI modules called NVIDIA NIM (NVIDIA Inference Microservices) are being deployed that provide containers to self-host microservices for customized AI models across clouds, data centers, and RTX AI compute. Holoscan for Media was built to enable multi-vendor applications with compressed and uncompressed media support through ST 2110 and other interfaces.
As this year progresses, we’ll see more products coming into general availability.
When we started our work in partnership with NVIDIA on Holoscan for Media, we had a much larger vision for replacing hardware systems with completely software-defined workflows in mind. But in today's hyper-inflated compute cost environment, that's not as practical as we and the market expected it to be. It doesn’t mean that the concept of using modular software-defined products in our real-time production and AV worlds is off the table. It just means, as always, using the right tool for the right job at the right cost.
We cannot solve the issue of translating a CEO’s speech into multiple languages in a video switcher, video router, or any other part of the video ecosystem, but we can solve that by running AI on prem for our clients. At the 2026 NAB Show, NVIDIA demonstrated the ability to translate a commentator into three different languages, closed caption those languages, and adjust the lip sync so that each language looked natural. All running a set of NIMs in a scaled Holoscan for Media environment on a single computer in real time. These are the beginnings of applications we’ll see throughout this year that are practical, deployable, and will fit within our clients’ budgets and cost structures.
Stay Aware of What's Available
What is that going to look like for system contractors as we go to sell these products and services to our clients? It's going to break down into multiple parts, as every system does.
First and foremost, we're going to need to learn the services that are available today to sell to our clients, and that will include understanding who the vendors are, what the risks of deploying those vendors are, and what the business models look like. This is the same as all products we’ve deployed throughout our careers.
Some of the AI-enabled applications becoming available in our real-time production world include the ability to use AI to upconvert material with greater quality than any previous technology provided. And AI provides a more nuanced set of tools for automatically reframing 9:16 for vertical streaming. In terms of audio, Ross Video showed a widely deployed AI audio mixing tool at NAB called Lama, which provides sophisticated audio mixing with less user interaction. These are tools that can be deployed in real time, on-prem client facilities today.
As this year progresses, we’ll see more products coming into general availability. These are continuous opportunities for us to help our clients evolve their systems and workflows, and provide better capabilities at lower operational costs.
A Sea Change for Integrators
How does this work in today's on-prem systems? It means you’ll be configuring and installing compute, either individual computers or small clusters in our client spaces, where we have traditionally put other devices, such as video switchers and other products that have evolved to become computer based. It also means that these systems are going to need real-time audio and video I/O.
It will be important to understand the signal flow in your existing system that you're adding these capabilities to, so the compute systems you’re building can be configured optimally. Is it SDI? Is it NDI? Is it HDMI?
This is where we take a different approach to the marketplace, because now we’re looking at augmenting existing on-prem, real-time systems with small clouds of compute within the existing environment. We’re bending the compute to fit the existing environment, as opposed to bending the workflow to fit the compute. This is a sea change in opportunities for system integrators—it brings these sophisticated capabilities into a place that doesn't require a major pivot in operations.
What do you need to do to get ready for this? As we begin to deploy this into our client sites, the most important set of skills for integration firms is that we have staff trained and ready to deploy virtualized and containerized applications on premise. Most traditional AV integrators have not had to do this previously.
The skills you will require will depend on your vendor partners. Some may choose to run their products in virtualized Windows environments using Hyper-V, others may choose Docker or Kubernetes environments. All of these are related but require different certifications and skills.
Then, of course, there will be the specific training on each vendor, how their product deploys, and how we train our customers to use it. These are skills we’re all familiar with and have been doing for many years.
This is the first time since I’ve written this column where I can say that everything we're talking about this quarter are things you can deploy for your clients next week. I encourage you to seek out vendors, learn the marketplace, and understand these ecosystems, as they will be highly valuable to our clients in the near term.
