Justin Hotard is selling Nokia shareholders ninety per cent of an AI story the optical networking margins have not yet earned. The figure is not estimation; it is the change in Nokia’s share price over the first six months of 2026, and it is the figure the company has decided, for the moment, to keep narrating before the numbers get a chance to contradict it. Amanda Lyons, who runs research at Energy Group Capital, said the relevant question is whether Nokia can deliver the margins the market now expects of an AI infrastructure company. The relevant question, more precisely, is whether the company can deliver them on a timeline that does not require the share price to do the work of the share price.
It is true, and the documentation supports it, that Nokia is not a phantom. The Finnish company, founded in 1865 as a wood pulp mill on the banks of the Nokianvirta River, has had a longer industrial life than most of the firms currently being called AI infrastructure plays. It exited the handset business in 2014, when it sold that unit to Microsoft, and it still licenses the Nokia name to a separate phone manufacturer. The mobile network equipment business — cell-tower gear and the software that runs it — still accounts for a little more than half of revenue, although that unit has been in decline as global carriers finish their fifth-generation deployments. The optical networking business, the unit that has done the work of the stock-price move, is a real engineering object: switches that connect servers inside data centres, routers that direct traffic between them, and the lasers, receivers, and chips that haul data at scale through the fibre that runs through the buildings. The metaphor the company offers is the right one. The thousands of miles of fibre in a data centre are the roads. Nokia’s products are the trucks and sorting centres. The trouble is that roads, trucks, and sorting centres do not become a different category of asset because the cargo is now mostly large language models. They are bandwidth infrastructure, and bandwidth infrastructure has been bandwidth infrastructure since long before the current capex wave.
The ninety-percent share-price move is, in the precise sense, a repricing. The business Nokia was running at the start of 2026 was not the business investors are now paying for. The market capitalisation the share price implies is a market capitalisation that requires the optical networking unit to grow at a rate the company has, in its own guidance, said it can nearly double, but on a base whose current margins Lyons, again, has noted the market has not yet seen. Nokia’s own forecast, after the first fiscal quarter, projects eighteen to twenty per cent growth — of a business that has not yet demonstrated the margins the market is paying for. The market is waiting for the earnings to catch up to the story. The earnings are, in their own time, working on the catching up.
The Infinera acquisition is the mechanism that produced the growth. Nokia paid $2.3 billion for Infinera, an American optical networking technology company specialising in the lasers, receivers, and chips that move data at scale, and the deal’s first consequence was a step-change in Nokia’s North American market share. The figure Ian Redpath of Omdia cited in the research note is the figure that does the work: Nokia’s optical network market share in North America went from 6.3 per cent to 27.3 per cent in 2025, the year the deal closed, placing the company firmly in second place behind Ciena, which leads the segment at 50.1 per cent. The acquisition is the reason Nokia is now in the conversation as an optical networking supplier to the hyperscaler buildout.
The Nvidia stake is the second mechanism, and it is a different category of move. In October 2025, Nvidia bought 2.9 per cent of Nokia for $1 billion as part of a broader collaboration agreement. The financial effect on Nokia is, on its face, modest — a billion dollars is not a transformative capital injection for a company of Nokia’s size. The signalling effect is the substance. Its decision to take a stake in a fibre-optics supplier, and to do so publicly, was read by the market as a validation of Nokia’s strategic position. Daryl Schoolar of Recon Analytics, in the same Wall Street Journal coverage, described Nokia’s stock as “basking in the glow of the AI halo.” The halo is a real thing. The halo is also the part of the valuation that does not have an earnings line attached to it.
The honest read of the situation is that Nokia is a competent optical networking equipment maker that has executed a competent acquisition, has won a competent share gain, and has been gifted, by the structure of the current AI capex cycle, a re-rating that is more enthusiastic than the underlying business currently supports. The structural risk is the one the company itself has identified, in the person of Hotard, in the company’s most recent earnings commentary: long lead times on the components Nokia needs, the same components Ciena and the rest of the optical networking market also need, and a competitive environment in which input prices are rising. Hotard said in April that the company is exploring ways to secure supply and control costs. “Long lead times are a risk. If those lead times get unpredictable, then their revenue gets unpredictable,” Redpath of Omdia said. The market that is paying AI-multiple prices for Nokia has not, in the price it is paying, priced that risk.
Nokia’s optical networking business is a real business with real engineering, growing at a real rate, in a market that is structurally real. The market’s repricing of Nokia as an AI infrastructure miracle is a bet that the underlying business will grow into the price the market is currently paying. The bet has been amplified by an acquisition that produced a real share gain, by a stake from the most-watched company in the sector, and by a chief executive who came from Intel’s data centre and AI division and knows how to tell the story the market wants to hear.
The trucks are real. The fibre will stay in the ground. The work is to be done.