As new advances in efficiency and vendor interoperability fuel momentum behind broadcasters’ reliance on production in the hybrid private/public cloud domain, cost challenges posed by AI resource scarcities are moving to the top of industry suppliers’ to-do lists.
The latest, frequently AI-enhanced solutions that came to light at April’s NAB Show have made the hybrid cloud model more attractive than ever amid waning enthusiasm for exclusive reliance on either public cloud or proprietary hardware resources.
In the case of Grass Valley’s customers, for example, “whether broadcasters choose to host aspects to their production workflows on premises or in the cloud can be decided on the basis of architectural complexity and what makes sense for the tasks at hand,” Chief Product Officer Adam Marshall said. “That’s what they’re looking for.”
Like just about everywhere else, AI is playing an ever-greater role at Grass Valley. Marshall said the company uses AI “to accelerate our ability to deliver what our clients need” and to provide a “third-party front door” that allows customers to add their own AI application layers.
The need to secure cloud operations, including vulnerability analysis of the type delivered by Anthropic’s new Mythos AI model, has added to AI processing requirements. Grass Valley will make the technology available if and when Anthropic lifts current restrictions on its release, Marshall said.
Inference Costs
A late 2025 report from Deloitte Insights underscored the immensity of the AI usage challenges facing all industries. That report found enterprises everywhere “are discovering their existing infrastructure strategies aren’t designed for AI’s demands,” Deloitte said.
The mismatch results from “inference,” or the incessant processing required as AI models generate new information based on changing conditions. “While inference costs have plummeted, dropping 280-fold over the last two years, explosive growth in overall AI spending… has dramatically outpaced cost reduction,” Deloitte said.
“Dependence on high-performance compute in the public cloud can get expensive,” Jonathan Gryckiewicz, business development director for cloud solutions at Ross Video, said.
Dependence on high-performance compute in the public cloud can get expensive.”
Jonathan Gryckiewicz, Ross Video
Businesses are adjusting to the new realities “incredibly fast,” he said, but for now, “it seems largely the case that to get benefits from AI, we need to connect the media environment to the [public] cloud” where processing is “gobbling up GPU capacity and prices are getting higher and higher.”
Ross Video’s latest cloud- and AI-related contributions include support for its XPression real-time browser-based graphics creation and monitoring tools via a scalable on-demand service called XPression.cloud, which can be used instead of local XPression hardware. XPression.cloud uses AI to identify appropriate XPression templates and expedite graphics-related asset retrieval, creativity and generation.
XPression.cloud offers customers “a more sophisticated way to manage graphics on their own” without “shipping boxes around,” Gryckiewicz said.
Matrox Video is also riding the market surge toward hybrid cloud-optimized converged production solutions. For the past two years, Matrox has promoted its ORIGIN media-aware content transfer fabric as a way to control high-performance multiprotocol transport and scalability without dedicated hardware. ORIGIN was again a big focus for Matrox at this year’s NAB Show, but with a different emphasis this time, Product Manager Daniel Robinson said.
“We hadn’t seen very many use cases emerging until now,” Robinson said, noting uptake this year involves the likes of industry suppliers Amagi, EVS and TSL, as well as the Canadian multiservice giant Rogers Communications. The difference lies with Matrox’s success in using AI to make it easier for customers to put ORIGIN to work in their products and services, Robinson said.
“We’re not running AI on the software but using AI tools to build things quickly,” he explained, which means executing on projects that might have taken days or longer in a matter of hours. “It’s about shortcutting discovery of information. If you didn’t do that, you’d have AI searching through thousands and thousands of lines of documentation with outcomes that may or may not be the right ones.”
AI Supply Chain
Such uses of AI on the build side of supplier innovation are fueling acceleration in product development for cloud- and hardware-based solutions alike. While these applications don’t incur the ongoing inference processing costs associated with AI applications used in live production workflows, they are vulnerable to hardware capacity and cost trends, noted Stephan Stadler, chief product officer at Appear.
“Being a hardware manufacturer, the AI supply chain is top of mind,” Stadler said, citing Appear’s use of AI not only in “how we plan and build things” but in discovering new opportunities and creating sales messaging. “The big challenge is memory chips for everyone who builds hardware products in this industry,” he added.
Appear has moved into the hybrid cloud production space with the recent introduction of the VX Media (VXM) Gateway, which uses open-source interfaces to enable interoperability with other vendor solutions in commodity cloud support for the dispersed REMI production versatility enabled by its hardware-based X Production portfolio.
“VXM doesn’t replace XP,” Stadler noted. “It’s just an extension of that.”
The company’s release of VXM—even as it registered another big hardware win, with Fox Sports using its X platform to deploy its REMI production workflow for the 2026 FIFA World Cup—reflects the growing pull of the virtualized hybrid cloud opportunity across the traditional broadcast-production hardware supply chain.
Another case in point is the expanded support for cloud production introduced by Sony Electronics’ Nevion unit. Nevion has integrated MOXELA, a new cloud-native software platform, with its cloud-based VideoIPath media orchestration layer to transport, process and monitor production workflows encompassing output from cameras and other equipment supplied by Sony and by other vendors.
Mounting demand for a next-generation video streaming solution that surpasses the latency and unidirectional limitations of the currently dominant streaming architecture is generating strong support for a new standard known simply as “MOQ” (Media over QUIC).
By all accounts, it will take few more months for the Internet Engineering Task Force to finalize the MOQ specifications. But multiple projects testing this new multidirectional, multilatency approach to streaming are now underway. “There’s a buzz and excitement in the air about MOQ we’ve never seen before,” said Will Law, chief architect for the cloud technology group at Akamai and a leader in the IETF MOQ initiative.
Infrastructure Support
This was evident at April’s NAB Show, where MOQ demos abounded on the exhibit floor and an overflow crowd filled the large meeting room devoted to the show’s first-ever session on the standard.
Entities announcing infrastructure support for MOQ included public content-delivery network operators Broadpeak, CacheFly, Cloudflare, CDN77 and Vindral, as well as cloud-compute platform providers Akamai, Amazon Web Services (AWS) and Oracle Cloud Infrastructure (OCI). Streaming software system suppliers demonstrating delivery of payloads and advanced features over MOQ-enabled facilities included Ant Media, Ateme, Bitmovin, EZDRM, Nomad Media, Norsk Media, Oracle Media Services, Synamedia, Red5 and Wowza.
While proprietary real-time streaming systems supporting video conferencing and a host of other applications have been around for several years, MOQ is meant to satisfy the need for a standardized, ubiquitously available two-way superhighway where video flows in any direction and at any scale, with latencies attuned to specific use-case requirements. That’s why Red5, a provider of a highly scalable WebRTC-based streaming platform, is throwing its weight behind MOQ, according to CEO Chris
Allen.
“WebRTC was designed as a peer-to-peer protocol, which means there’s no standard for scaling the way we do by building out clustered cloud resources,” Allen said. “MOQ was designed from the get-go as a scalable broadcast technology for the internet, which we believe will accelerate adoption of next-generation streaming.”
The foundational set of MOQ specifications, known as MOQ Transport (MOQT), marks an incompatible break with conventional streaming modes like HLS and MPEG-DASH, which are based on the Hypertext Transfer Protocol (HTTP). As a so-called connectionless platform, MOQT employs a simple publish/subscribe approach to streaming by signaling the start and end points of a session while avoiding the thousands of request/response exchanges that transpire between clients and servers in HTTP streaming.
Latency Tune-Up
The new standard also provides a way to automatically tune end-to-end latencies to whatever levels meet user requirements, including real-time latencies of 200 to 400 milliseconds that are imperceptible to humans; what Law calls “interactive live” at 2 seconds for use in sports betting, online casino gambling, banking and similar applications; and “conservative live,” supporting ironclad HD, 4K and eventually 8K quality persistency at 5 seconds.
Paralleling MOQT development, there are other initiatives within IETF and the OpenMOQ Software Consortium, a closely affiliated ad hoc organization aimed at creating a practical operating environment for the transport protocol. They include media-layer streaming formats emerging from IETF, a new client player template developed by Red5 and endorsed by the consortium, and several ready-to-deploy players.—Fred Dawson
“The big problem we’re solving with MOXELA is producers’ need for flexibility to adapt to different workflows by running various processing functions for different streams in parallel across on-prem and public cloud clusters,” Nevion Head of Product and Solutions Jan Helgesen said.
The Impact of MXL
MOXELA aligns with common frameworks like SMPTE ST 2110 and NDI, as well as the emerging industry guidelines shaped by the European Broadcast Union’s open-source Digital Media Facility (DMF) initiative, including the Media eXchange Layer (MXL) specifying containerized microservice support for real-time, in-memory exchanges of uncompressed audio, video and metadata across multiple vendor software applications. This is a big step toward enabling cloud-based live sports and other productions, but there’s much more to come, Helgesen said.
“We’re currently on a cloud production journey from Tiers 3 and 2 to Tier 1,” he explained, noting that getting to Tier 1 still “requires some technical development that’s not completely there yet.” It’s a staged road map that will also bring AI into play to automate the complex troubleshooting, planning and setups that big cloud productions will require, he added.
Implementations of MXL in vendor product lines—facilitated by the Linux Foundation’s release of the first MXL Software Development Kit (SDK)—are fast becoming the norm. This is because the shift toward hybrid cloud production has made just-in-time activations of COTS resources with cross-vendor interoperability essential to cost containment.
Sony, along with a new commitment to the protocol by Ross Video, adds to a list of MLX supporters that includes Appear, AWS, Grass Valley, Imagine Communications, Intel, Lawo, Matrox, NVIDIA, Riedel Communications and Telos Alliance.
Pay As You Go
For Lawo, the growing consensus around the hybrid cloud approach to broadcast production is right in stride with its longstanding promotion of the Lawo HOME management and orchestration platform. HOME runs on commodity servers in private or public clouds to enable discovery, authentication, security management and monitoring of all production-connected devices from a single user interface.
To address cost uncertainties as the hybrid cloud production model takes hold, Lawo has introduced Flex, a payment system built on credits that makes it easier for customers to adjust to changing dynamics in private vs. public cloud resource usage.
“You can buy and use Flex credits like chips in a casino to spend on any HOME app,” Chris Scheck, head of marketing content for Lawo, said. The hybrid cloud model “is where we’re heading,” he added, but uncertainties mandate offering customers as much flexibility as possible in choosing how to spend on cloud resources.
For example, while customers would rather maximize use of private resources, “right now, servers have almost tripled in price because of chip shortages, and it can take six months to get them,” Scheck said. At the same time, as Grass Valley’s Marshall observed, in the public cloud space “we’ve seen processing costs rising at record rates, up 40% during the first part of the year and now up 80%.”
Factor in the complications added by the AI squeeze and it’s clear the hybrid cloud model might not be the cost-saving slam dunk it once appeared to be. But in a live-production environment that needs to accommodate fragmented workflows delivered through ST 2100, NDI, SDI and other pipelines while enabling browser support for executing editing tasks across dispersed locations, the advantages of cloud-based accessibility, agility and efficiency are too compelling to be ignored.
This puts a premium on cloud cost-containment know-how. As Marshall noted, at a moment of growing confusion among producers, figuring out what will work for each customer has much to do with knowing which questions to ask.
