According to Manish Gulyani, vice president and head of marketing for network infrastructure at Nokia, the acquisition has accelerated the vendor’s position in the rapidly expanding optical and hyperscale connectivity market, particularly as AI-driven infrastructure investments reshape network demand.

“It’s actually been tremendous,” Gulyani says of the integration. “The two teams came together really, really quickly.”

The rationale behind the deal was clear from the outset: strengthen Nokia’s footprint in North America, deepen relationships with hyperscalers, and scale innovation investment in optical technologies as data traffic and AI workloads accelerate. 12 months later, Gulyani says those ambitions are beginning to translate into measurable results.

“When we put the two businesses together, our ambition was to have strong presence in North America, strong presence with hyperscalers, and really become the number one or number two kind of player in the market,” he explains.

Early performance indicators appear to support that strategy. Nokia’s optical division recorded strong growth through 2025, driven largely by hyperscaler demand and new optical architectures emerging around AI workloads.

“If you take a look back at the end of 2025 results for our optical business, it’s been a record year,” says Gulyani. “We grew well north of double digits in the optical business.”

He adds that the momentum culminated in a milestone during the fourth quarter.

“In Q4 of 2025… we came in at number one market share in optical,” he says. “I think it was the first time Nokia was overall number one in optical, which is a tremendous achievement.”

A key driver of this growth has been deeper engagement with hyperscale cloud providers. Nokia now claims design wins across most of the major global players.

“We are now in nine of the 10 hyperscalers, deployed in designs across IP and optical combined,” Gulyani says. “That’s tremendous progress.”

Pluggables and rising optical demand

One of the most visible technology shifts emerging from the combined portfolio is the growing adoption of high-speed optical pluggables, particularly at 800G.

Demand for these modules, which allow operators and hyperscalers to deploy high-capacity optical connectivity in a flexible form factor, is accelerating rapidly as AI infrastructure scales.

“We were super excited about the introduction of pluggable technology, the 800G pluggable,” says Gulyani. “The demand has gone through the roof.”

Nokia has already begun shipping the technology, but Gulyani warns that industry-wide supply constraints could soon emerge as demand continues to accelerate.

“I think the world is going to begin to see supply constraints,” he says. “There’s that much appetite and demand for it.”

The acquisition has also enabled Nokia to consolidate previously separate product portfolios into a unified optical roadmap without cancelling existing platforms.

“We didn’t cancel products,” Gulyani explains. “What we did was optimise our portfolio between the Infinera family and the Nokia family to have a unified portfolio.”

Just as importantly, the integration has brought together previously separate engineering teams and silicon roadmaps.

“We put the R&D teams together because we wanted to be able to do multiple generations of silicon development concurrently,” he says.

By combining Nokia’s digital signal processing (DSP) expertise with Infinera’s photonic integration capabilities, the company believes it can accelerate the pace of innovation in coherent optics.

Hyperscalers reshape network design

Much of the recent demand surge in optical and IP networking is being driven by hyperscalers building large-scale AI infrastructure.

According to Gulyani, the architectural requirements of these deployments are evolving rapidly, particularly as companies shift from single-building data centres to distributed AI campuses.

“Hyperscalers are moving faster than we even think,” he says. “They’re spending more and they’re doing more interesting things than we even envisioned.”

Historically, optical connectivity between data centres was largely focused on linking individual facilities. But AI training clusters are now forcing hyperscalers to design multi-building campuses that function as a single computing environment.

“In the last year or 18 months we’ve started talking about building multi-site AI factories,” he says.

Instead of housing GPUs in a single building, operators are spreading infrastructure across multiple facilities due to space and power constraints.

“They’re building four-building, six-building campuses and connecting them together,” Gulyani explains. “They’re basically treating it like a single data centre.”

This architectural shift is creating a new connectivity requirement that Nokia refers to as “scale-across” networking.

While traditional data centre networks focus on connecting GPUs within racks or rows, scale-across connectivity links entire buildings or campuses together.

“That connectivity across those buildings to make it behave like a single system is driving massive interconnect demand,” Gulyani says.

The result is a surge in both optical capacity requirements and high-performance IP routing platforms.

Nokia has already demonstrated systems designed for this architecture, including platforms capable of supporting hundreds of high-speed ports.

“We showcased a platform with 576 ports of 800G that can be used to connect these locations,” he says.

Such infrastructure not only requires high-capacity routing but also large-scale optical transport to handle the enormous data volumes generated by AI workloads.

What makes a network AI-ready?

As AI becomes the dominant driver of infrastructure investment, network architecture itself is beginning to change.

Gulyani argues that “AI-ready” networking requires optimisation across multiple layers of the network stack, from switching inside data centres to optical transport between facilities.

Within AI data centres, two distinct network environments are emerging: front-end networks that handle user queries and service access, and back-end networks that connect GPU clusters used for model training and inference.

“The front-end networks are more traditional,” he says. “They’re architectures that have been used in cloud networking.”

The back-end networks, however, require entirely different performance characteristics.

“Because of the number of GPUs connecting and the way they share information with each other, you need very high scale but very low latency, very low loss networking,” he explains.

To meet those requirements, Nokia has been optimising its switching platforms to support ultra-low latency and advanced congestion management.

Industry initiatives such as the Ultra Ethernet Consortium are also helping define how Ethernet can evolve to support AI back-end networking.

Meanwhile, optical technology continues to advance to support the scale of these new deployments. Nokia has focused heavily on vertical integration, developing its own DSPs, photonic integrated circuits and packaging capabilities.

“We’ve done full vertical integration inside Nokia,” says Gulyani. “Designing our own DSPs, building our own photonic circuits, packaging the module and testing it ourselves.”

The addition of Infinera’s manufacturing facilities has strengthened that strategy further.

“That was one of the benefits of Infinera,” he says. “They had a fab to build their own optical front ends.”

With AI infrastructure investment continuing to surge globally, Gulyani believes demand for high-capacity optical and IP networking will only accelerate.

For Nokia, the integration of Infinera appears to have arrived at precisely the right moment.

“It’s been one of the really good acquisitions,” he says. “A lot of acquisitions have excitement and then a dip. This one has just gone up and then further up.”

RELATED STORIES

Doubling down on optics: How Nokia’s Infinera acquisition fuels AI driven network scale

Nokia to acquire Infinera for $2.3bn