Optical networks to bridge the AI compute-consumption gap
With AI spurring gigawatt-scale datacentre builds across APAC, Ciena is deploying ultra-fast, energy-efficient optical networking and AI-driven automation to ensure AI services can reach consumers
Artificial intelligence (AI) is often imagined as an ethereal network of algorithms and advanced computing chips in the cloud. But behind the scenes, the ability to generate a poem, render a video, or run complex predictive models relies on pulses of light travelling through glass fibres across oceans and continents.
As big tech firms and enterprises race to build multi-billion-dollar datacentres to fuel the demand for AI, the telecoms infrastructure connecting them is under immense pressure. This has created significant opportunities for companies like Ciena, a global networking systems and software provider.
“AI is only good when it is consumed,” said Amit Malik, vice-president and general manager of Ciena in Asia-Pacific, Japan, and India. “We are the bridge between compute and consumption because we provide those pipes.
“For a while, people thought we were the derivative layer, because we were getting the benefits of what’s happening on the upper layers in AI, but the world has started believing that we are the mission-critical layer because if the network is down, nothing flows,” he added.
To meet the insatiable demand for bandwidth, Ciena has leaned heavily into its leadership in coherent optics, which synchronises light waves to reduce noise and carry more data over long distances. Its flagship WaveLogic 6 modem, capable of transmitting data at 1.6 terabits per second, is currently the fastest in the world and has become an industry benchmark.
This technology, already adopted by companies like Australia’s Telstra and India’s Vodafone Idea, consumes half the space and power of previous generations while offering ultra-low latency. This is particularly beneficial for datacentre and telecom operators that face stringent sustainability requirements.
Ciena’s advancements have also extended to the servers powering AI. The company recently launched the Vesta 200 6.4T optical engine, which manages the complex electrical-to-optical conversions within datacentres. This innovation is propelling the industry’s transition towards photonics as traditional copper and electrical systems approach their limits. “We do believe that the world is moving from electrons to photons,” Malik said.
APAC datacentre boom
The Asia-Pacific (APAC) region is seeing a staggering scale of infrastructure deployment, driven by hyperscalers establishing massive facilities closer to where most of the world’s data is generated.
“The erstwhile datacentres used to be not very complex. If you were setting up a 30-megawatt datacentre, somebody would have said, ‘That's a huge data centre.’ Today, one gigawatt seems to be the currency,” Malik said.
For a while, people thought we were the derivative layer, because we were getting the benefits of what’s happening on the upper layers in AI, but the world has started believing that we are the mission-critical layer because if the network is down, nothing flows
Amit Malik, Ciena
Growth projections for datacentre buildouts in APAC are being outpaced by reality. While analysts once predicted steady, incremental growth, the pace of development has accelerated dramatically.
“Every time people imagine something, three months later, somebody else comes and announces gigawatt datacentres, which changes the whole figure,” Malik said. He highlighted recent announcements of four to five one-gigawatt facilities in India, along with significant investments in the Thailand-Singapore-Malaysia corridor, Japan, Korea, and Australia.
This regional momentum is clearly reflected in Ciena’s financial performance. During its fiscal Q1 2026 earnings call, Ciena reported a 40% year-over-year growth in orders from India. Additionally, geopolitical issues in the Middle East have spurred the creation of new subsea cable routes, highly benefiting strategic landing points in India and Australia.
AIOps and autonomous networks
Operating networks that move petabytes of data across hundreds of fibre strands requires a new level of management capability. To address this, Ciena is embedding AI operations (AIOps) across its software layers, including its Blue Planet platform, pushing the industry towards fully autonomous networks.
This AI-driven automation helps optimise wavelength selection for high-speed routing and allows for the predictive remediation of network issues before they occur. It also helps mitigate the tech sector’s talent shortage.
“You don’t really have that many trained professionals everywhere,” Malik observed, noting that Ciena's AI tools have significantly improved support efficiency, enabling service teams to resolve complex issues and even manage deployments remotely with fewer resources.
Although training these massive models entirely within national borders remains prohibitively expensive for many countries, localised inferencing is becoming more widespread. This requires heavily encrypted, sovereign networks, Malik noted.
Looking ahead, Ciena is addressing future threats by deploying quantum-safe networking. The company recently demonstrated next-generation quantum-secured communications at the 2026 Optical Fiber Communication Conference.
Despite the massive growth opportunities, the industry is not without its headwinds. The rapid expansion of hyperscalers has outpaced the global supply chain’s capacity, leading to shortages in fibre optics and semiconductor components.
“Do we have enough fibre? "The answer is no, and that is why you are seeing more fibre investments coming up,” Malik said, adding that supply chain bottlenecks for specialised tech components are also a concern.
“Your constraint can be that last golden screw. You might have 99.9% of the equipment ready, but if one screw is missing because the supplier cannot scale, then you’re stuck,” Malik explained.
In this handbook, focused on AI infrastructure trends in the Asia-Pacific region, Computer Weekly looks at how it should be deployed, data and sustainability challenges and the advantages of large language models.
Digital Realty’s CTO Chris Sharp talks up the pace of AI silicon innovation, the growth of inferencing workloads, and why boasting about datacentre megawatts misses the point.
