IBC
article here
The elimination of OB trucks is just the start of the light
revolution. For the media industry, a rewiring of the transport network from
electrons to photons promises to unlock AI‑driven production, immersive
formats, and globalised workflows while dramatically cutting energy
consumption.
“Bandwidth demand has exploded,” says Dr. Masahisa
Kawashima, IOWN Technology Director at Japanese telco NTT. “What used to be
less than one gigabit per customer is now tens of gigabits for media
applications, and hundreds of gigabits for AI workloads.”
Momentum is building behind photonics – laser light - as
replacement for electronics. An All Photonics Network (APN) has gained the
support of 170 major telcos, device and chip vendors and internet powerhouses
including Nokia, Ericsson, Orange, KDDI, Intel, Nvidia, Cisco, Ciena, Samsung,
Sony, Microsoft and Google under the Innovative Optical and Wireless Network (IOWN) Global Forum.
“UHD contribution, multi‑camera remote production, VR
capture, and AI‑assisted workflows have all accelerated bandwidth demand far
beyond what legacy electrical switching architectures were designed to handle,”
says Kawashima, who is also Chair of the IOWN Global Forum’s Technology Working
Group. “At the same time, AI workloads, particularly large‑scale model training
and inference, have introduced traffic patterns that are both bursty and
massive, often exceeding hundreds of gigabits per second per node. The bottleneck
is the network.”
Optical transport technologies have long been used to connect
routers and switches in data networks. Traditionally, these routers and
switches connect electrically through a telecom carrier’s optical transport
system. But bandwidth demand is reaching a point where this model is no longer
efficient.
The IOWN initiative proposes to move long‑distance optical
transport capabilities directly into customer premises equipment (CPE),
enabling end‑to‑end photonic paths with minimal electrical conversion.
“Optical transport technology has evolved to the extent that
it’s both practical and cost effective to deploy long‑distance optical systems
directly at customer sites,” he says.
For decades, electrons have been the carriers of data inside
devices, performing all computation and routing. But electrons come with
limits: heat generation, power inefficiency, and bandwidth bottlenecks. Every
time a signal switches between optical and electronic domains, latency and
energy cost increase.
Photonics replaces some of these electrical pathways with
optical ones, allowing data to travel at the speed of light while reducing
energy consumption dramatically.
The result, according to NTT, is to reduce power consumption
to one-hundredth of existing output, increase data capacity 125 times and slash
network latency to a fraction of a percent of its current levels.
True virtual remote production
For broadcasters and live‑production companies, the
implications are profound. Media networks are increasingly indistinguishable
from data‑centre networks. Today’s live production workflows still rely heavily
on outside broadcast vans, specialist crews, and on‑site infrastructure.
“OB vans are expensive, and broadcasters can only own a
limited number,” Kawashima says “Skilled editing crews are another bottleneck.”
APN changes the equation, he claims. By enabling direct,
high‑bandwidth optical connectivity from venues to centralised production hubs,
“APN removes the need for OB vans and dramatically reduces on‑site staffing.
The result is a more flexible, scalable, and financially sustainable model.”
By eliminating the buffering resulting from optical-to-electrical-to-optical
conversions in current networking latency is deterministic – ideal for live cloud
switching.
“In media production, from multi-camera sports coverage to
immersive, free-viewpoint experiences, precise synchronisation is critical,”
says ,” Katsutoshi Itoh, Chair of the Use Case Working Group at IOWN and head
of Sony’s Swedish R&D lab. “Even minor latency variations can disrupt 3D
reconstruction and real-time interactivity. APN means predictable, tightly
controlled timing across the network.”
This shifts the bottleneck from physical logistics to
network provisioning and unlocks the long‑tail of live content. “With lower
production overheads, broadcasters can cover more events (local sports, niche
competitions, cultural performances) that were previously uneconomical.”
Tests have been made: Sony and Japanese broadcaster TBS claimed the
first successful remote production a live music event using APN. 64 audio
streams from The Japan Record Awards held at Tokyo’s New National Theatre were
relayed for remote production to TBS’s Akasaka studio with a roundtrip of 5
milliseconds.
Enabling AR, holography, and 6DoF video
Emerging formats such as 6DoF video, volumetric capture, holographic
replay and AR overlays for live events require multi‑camera arrays and
significant AI compute. These workloads generate enormous traffic between
capture nodes and compute clusters.
“APN is essential for these,” says Kawashima. He points to
the VAR-style Automated Balls and Strikes (ABS)
challenge system introduced for MLB baseball coverage this season. It uses 12
Hawk-Eye cameras placed around the stadium that continuously track the baseball
in 3D.
“AI synthesises entirely new viewpoints—angles that no
single camera could capture,” he says. “This six‑degree‑of‑freedom video
experience is a glimpse of what’s coming. APN provides the bandwidth and low
latency needed to make them practical.”
How APN integrates with today’s networks
Crucially, APN does not require ripping out existing carrier
infrastructure. Instead, it changes where the optical transceivers live.
The key component are Dense
Wavelength Division Multiplexing (DWDM) transceivers that convert electrical
signals into optical signals and vice versa, using multiple wavelengths to send
information through a single fibre optic cable.
Carriers already operate optical transport systems, and
those remain in place. The difference is where the optical transceivers sit.
Today, they are installed in the carrier’s transport equipment. In APN, the
transceivers are installed directly in the customer’s routers and switches,
which then connect directly to the carrier’s optical infrastructure.”
“The most efficient method,” Kawashima explains, “is to
connect customer routers and switches directly to a DWDM optical network. Directly
connecting customer routers and switches with optical transport transceivers is
already feasible today.”
Optical transceivers themselves are currently more
expensive, but Kawashima expects the market to evolve quickly. “Think of
smartphones,” he says. “The components inside are extremely expensive, yet
business models make them accessible. The same will happen with APN
transceivers.”
Industry consensus required
The next challenge is ecosystem alignment. IOWN was preceded
in 2015 by OpenROADM which similarly aims
to define and promote open optical data plane specifications. Its supporters
include AT&T, Deutsche Telekom, NTT and Cisco.
Other vendor and carrier‑led groups have emerged to promote
their own compute and optical technologies. These include the Open Compute Project (OCP - which has
Meta, Microsoft, Nvidia and Google on its steering committee) and the OCI MSA (Optical Compute Interconnect
Multi-Source Agreement) group which intends to establish an open, interoperable
optical interconnect specification for AI. It’s also backed by Meta, Microsoft,
Nvidia plus Broadcom and OpenAI.
“We have a strong collaboration with the OCP,” says Kawashima.
“Together we’ve launched the AI
Computing Continuum, which aims to define technology standards for a
computing space spanning multiple clouds and edges.
“What differentiates IOWN is that we define concrete use
cases—financial data centres, remote media streaming—and then specify end‑to‑end
system designs. With that foundation, we collaborate with consortia like OCP
and OpenROADM.
OCI MSA launched in February and IOWN is yet to engage
directly. But Kawashima believes it will appreciate IOWN’s work “because we
define use cases for their optical devices and show how customers can build
full end‑to‑end systems using them.”
Interoperability for APN requires standardised long‑reach
coherent transceivers suitable for CPE as well as operational models for
provisioning photonic paths directly to enterprise sites.
“Once interoperability standards are established, the cost
impact will be modest,” he says.
2030 APN roadmap
IOWN’s 2030 roadmap is ambitious: a global computing and
communications fabric built on photonics, low‑latency architectures, and
distributed AI. Parts of this vision are already commercially deployed.
NTT launched its first APN services in Japan two to three
years ago, initially offering 100 Gbps connectivity in limited regions.
Coverage and performance have been expanding ever since. Higher throughput,
with per‑customer links scaling to 100G, 400G, and beyond is envisaged.
“The basic concept is already in service,” he says. “We
continue to upgrade, but the foundation is real.”
Emerging architectures such as co-packaged optics (CPO)
bring optical data transceivers directly next to compute chips, cutting power
and latency by shortening conversion distances.
Nvidia
is building CPO networking switches to scale AI for industrial use. Other
developers are lining up to incorporate optical semiconductors, switchers and
interconnects including Twinstar Technologies, Delta
Electronics, and Corning
Incorporated.
Beyond this, technologies like optical interposers and fully
all-optical computing systems, where photons handle processing without
conversion, are rapidly progressing.
One of the most urgent use cases is sovereign AI
infrastructure. Nations are racing to build regional AI clouds to maintain
competitiveness, but high‑performance compute is only half the story. “You must
also connect customer locations to these AI systems with enough bandwidth,”
Kawashima says.
Financial services are another early adopter. As banks
transform into digital‑platform‑driven businesses, ultra‑low‑latency, high‑bandwidth
connectivity becomes essential. Photonics offers both.
Photonics and the Quantum future
Looking further ahead, photonics will be essential for
quantum computing. Quantum systems will be centralised in specialised data
centres, not distributed to every enterprise. The challenge will be moving
enormous matrix‑based datasets in and out of those facilities.
“High‑bandwidth, low‑latency connectivity will be critical,”
Kawashima says. “APN is well suited to that requirement.
Oriole Networks, which spun out of UCL in 2023, set out to
build the highest-performing AI network imaginable, “pushing toward
theoretical limits” with photonics as a key enabler.
“It is a massively performant AI network based on incredibly
efficient photonics,” says
CEO James Regan of his company’s PRISM platform which boasts a 50 Exabit per
second throughput.
The Netherlands is another photonics hotspot.
Eindhoven-based Photon Bridge’s
laser‑on‑silicon approach “redefines photonic integration to make
light scalable, manufacturable, and infrastructure-ready.”
“Photonics is the physical foundation of the next digital economy,” says Mark Rushworth, founder and CEO of UK start up Finchetto which is working on an optical packet switch. “AI, quantum computing, and cloud networking all depend on the ability to move and process data faster, cooler, and smarter. In the ongoing race to optimisation, light wins every time.”