Can Nokia Be A Serious Recovery Play?

Taking a new trading position that might turn into an investment

Last week, I took a new position in $NOK. The thesis behind it was more so rooted in the context of a trade after looking at the option flow, but after doing some more diligence over the past week, I think there is deeper potential to the trade.

Now, to start off — I want to preface by saying Nokia, along with many other “picks and shovel” plays in AI can EASILY go down from these levels given how parabolic their run has been. Any bad news around oil, bond yields, Iran, cap-ex slowing down, etc. would lead to the broader market correcting. Understanding some of these trades have to be seen in the context of momentum but also long-term value propositions in case the names experience a correction as the underlying fundamentals, if in tact, would present a buying opportunity on those names.

If the market corrects and semis in particular get hit, NOK will not be immune. If you think a short term correction is coming after the massive move we have gotten and aren’t comfortable with DCA-ing into the name, then it is probably not for you and there may be better opportunities to by.

Why I Got In

Simple thesis:

• Nvidia owns 2.9% of the company, they bought in October 2025 when the stock was around $6.50

• The option flow is incredible signaling that there are larger players that see the name moving into the $20s

• Chart is breaking out from a decade-long consolidated base

• Core AI business is growing 49% YoY and AI data centers are moving towards the cell tower / telecom edge which would imply Nokia having a strong runway of growth

First, let’s talk about the option flow. This is quite incredible to see.

There are 4x the amount of calls vs puts, it doesn’t mean all those calls are bought, but it does mean there is much strong trading activity in the call side. Remember, NOK is not a big name in the market or a name that gets a ton of attention, seeing this much money flow into derivatives for it signals that there may be more institutional players attempting to come in.

Highest Open Interest is for the Jan 2027 $15C:

The options flow isn’t a thesis, but I am seeing it as a signal to assume there are more people thinking this name has potential, which then begs the question…why are people interested?

Enter Nvidia.

From their PR:

Nvidia invested $1B in Nokia for ~2.9% stake, and the two are building an AI-RAN platform: Nokia’s RAN software + Nvidia GPUs/CUDA/Aerial/ARC-Pro hardware to make 5G/6G networks run both wireless traffic and AI workloads on the same infrastructure. Today, telecom networks mostly move data. The pitch is that future networks will also process AI locally for robots, drones, autonomous vehicles, AR/VR, cameras, industrial sites, cities, etc. Nvidia calls this turning 5G into a distributed AI computer.

Nokia brings the telecom stack:

• AirScale radios/baseband

• anyRAN software

• Cloud RAN

• 5G Advanced/6G roadmap

• carrier relationships

They’ve already tested Nokia workloads on Nvidia GPU-accelerated AI-RAN with T-Mobile, Indosat, and SoftBank, and are working with BT, Elisa, NTT DOCOMO, Vodafone, plus Dell/Red Hat/Supermicro/Quanta.

Nvidia brings the accelerated compute layer:

• CUDA

• AI Aerial

• ARC-Pro

• Grace Hopper/Blackwell-class compute

• RTX Pro server hardware

• developer ecosystem.

Right now, almost all of Nvidia’s AI revenue is tied to a handful of gigantic customers including hyperscalers like Microsoft, Amazon, Google, and Meta, giant centralized AI data centers, and training massive foundation models.

That market is enormous, but it also has limits which include eventual growth slowing, hyperscalers build custom chips, AI compute becomes concentrated among fewer buyers, margins compressing over time.

So Nvidia’s next goal is:

“How do we make AI infrastructure as widespread as the internet itself?”

That’s where telecom comes in. This partnership with Nokia is quite frankly Nvidia trying to make telecom another major AI infrastructure market, not just hyperscaler data centers.

Every telecom operator already owns towers, fiber, edge compute locations, power, networking infrastructure, real estate close to users.

There are millions of telecom endpoints globally. Nvidia sees this as a chance to turn telecom networks into a global distributed AI compute layer. Instead of AI only happening in giant Virginia or Texas data centers, AI inference could happen:

• at cell towers

• inside factories

• near autonomous vehicles

• in hospitals

• in smart cities

• at ports and warehouses

• inside retail stores

• near drones and robotics systems

The reason this matters is latency. Some AI applications cannot wait for a round-trip to a hyperscaler cloud. Examples include autonomous driving, industrial robots, military systems, augmented reality, real-time translation, smart traffic systems, security cameras analyzing video live and factory automation.

These need responses in milliseconds.

So instead of:

User → internet → giant cloud → AI response → back to user

you get:

User/device → nearby telecom edge node → AI processed locally

That is the “edge AI” thesis. Many usecases of AI are now coming to the edge, as you may have heard of edge AI in various instances, and the partnership here with Nokia will be one step in Nvidia trying to build a more broadly distributed infrastructure for EdgeAI. Nokia matters because telecom operators trust Nokia already. Nvidia does not have deep telecom relationships by itself. So Nokia becomes the bridge because Nokia provides carrier-grade telecom software/hardware and Nvidia provides AI compute/GPU stack.

Anduril is taking a role with Nokia in these new edge AI usecases as well:

We will see more partnerships like this over time as latency becomes the core element that determines if companies can access the inference they need, at the edge, without needing to communicate with a datacenter.

Jensen

Jensen is bullish.

The biggest thing he has emphasized is that telecom networks are becoming AI infrastructure, not just connectivity infrastructure. He said, “Telecommunications is a critical national infrastructure — the digital nervous system of our economy and security,” and argued that AI-RAN built on CUDA and AI will “revolutionize telecommunications.”

He has repeatedly framed the Nokia partnership as part of a much bigger generational platform shift from traditional 5G toward AI-native 6G networks. Huang described the future network as a system capable of “processing intelligence from the data center all the way to the edge,” which is basically Nvidia’s thesis that telecom towers and edge infrastructure will eventually function like distributed AI compute nodes.

What is especially notable is how directly he praised Nokia itself. Huang said, “Nokia knows telecommunications like no one knows telecommunications,” and explained that Nokia is integrating Nvidia’s ARC and Aerial platforms directly into its base stations so AI and network services can run together on the same infrastructure. He also joked that he wished Nvidia’s investment had been “$2 billion instead of $1 billion,” which investors interpreted as a sign of strong conviction in the partnership.

It seems like Jensen seems to view Nokia as Nvidia’s telecom beachhead. Nvidia already dominates centralized AI data centers, but Nokia gives it deep access to carriers, radio infrastructure, and the future 6G stack. The vision is that Nvidia GPUs, CUDA, and AI software eventually become embedded throughout telecom infrastructure globally.

The Bull Case for Nokia

In my interpretation of seeing where this industry is going, this could move Nokia from “boring 5G equipment vendor” to a key player in AI-native networks, edge AI, 6G, data center networking, and telecom compute. Reuters noted Nokia is using AI/data center demand to offset weaker 5G spending and contract losses.

Today, telecom is mostly viewed as slow-growth, commoditized, low-margin infrastructure. AI-RAN tries to change that model by allowing carriers to monetize AI compute, enterprises to rent edge AI capacity, and telecom networks to become programmable AI platforms.

This is strategically important for Nvidia because it massively expands its TAM. Instead of selling AI infrastructure mainly to 10–20 hyperscale buyers, Nvidia could eventually sell into hundreds of telecom operators, governments, industrial operators, smart city deployments, and sovereign AI infrastructure projects.

Most importantly, Nvidia wants CUDA to become the operating system of all AI infrastructure, not just cloud training clusters. So this partnership is really about expanding AI beyond centralized clouds, creating a massive edge AI market, turning telecom infrastructure into AI infrastructure, and embedding Nvidia deeper into global networking before competitors do. If this works, the telecom tower of the future may partially function like a mini AI data center.

The Bear Case

The bear case is that AI-RAN could take far longer to monetize than investors expect. Telecom operators are historically slow spenders with long upgrade cycles, and many carriers are still trying to earn acceptable returns on their existing 5G investments. While the vision of turning cell towers into AI compute nodes sounds compelling, carriers will need proof that enterprises are actually willing to pay for edge AI services at scale. Without clear ROI, telecom companies may hesitate to spend billions upgrading infrastructure with Nvidia GPUs and new AI-native hardware.

There is also a risk that the market opportunity ends up much smaller than the hype suggests. Many AI workloads can already run efficiently inside centralized hyperscaler data centers, meaning edge AI may only be necessary for a limited set of low-latency applications like robotics, autonomous systems, or industrial automation. Competition could also intensify, with companies like Qualcomm, AMD, Intel, and custom in-house telecom silicon providers all trying to capture parts of the stack. If adoption is slow or margins remain telecom-like rather than software-like, the partnership may end up being strategically important but financially underwhelming for both Nokia and Nvidia.

The bull response to that bear case is that telecom infrastructure transitions always look unnecessary early on, until entirely new applications emerge that suddenly require them at scale. Skeics once argued that cloud computing, GPUs for AI, and even 5G itself lacked clear monetization paths, yet demand eventually exploded as the ecosystem matured. Nvidia and Nokia are not betting that every AI workload moves to the edge tomorrow, they are basically positioning early for a future where robotics, autonomous systems, industrial AI, defense applications, smart cities, and real-time AI agents require ultra-low latency compute closer to users and devices. If you believe in that broader future, the numbers today don’t look amazing, but they can start to over time. If that shift happens, telecom operators could evolve from low-margin bandwidth providers into owners of globally distributed AI infrastructure, while Nvidia embeds CUDA and its AI stack into an entirely new layer of the economy before competitors can establish a foothold.

Numbers

The good news with Nokia is that the financial story is finally starting to change from “declining telecom hardware vendor” to “AI/network infrastructure company.” In 2025, Nokia generated about €2.0B in comparable operating profit with €1.5B in free cash flow, while Q1 2026 operating profit jumped 54% YoY as AI/cloud demand accelerated.

The biggest positive is that AI & Cloud revenue grew 49% in Q1 2026, while optical networking and IP networking demand surged from hyperscalers building AI data centers. Nokia is no longer purely dependent on telecom carrier spending cycles; it is increasingly tied to AI data center networking, fiber, optical transport, sovereign infrastructure, and defense connectivity.

Financially, Nokia actually looks fairly healthy now. Gross margins expanded into the mid-40% range, free cash flow generation improved materially, and the company still carries a strong net cash position around €3.8B. The balance sheet strength matters because telecom equipment companies historically struggled with weak profitability and heavy debt. Nokia also has a valuable patent licensing business that contributes hundreds of millions to operating profit annually, helping stabilize cash flow even during weak telecom cycles. On valuation, the stock still trades at a much lower multiple than AI infrastructure peers because many investors still mentally classify Nokia as a legacy telecom name rather than an AI networking company. That creates the bull thesis: if AI/data center networking becomes a much larger percentage of revenue, Nokia could get a multiple rerating over time.

CEO on Q1 Earnings:

At our Capital Markets Day in November, we outlined our view of the AI supercycle and the market opportunity for Nokia. Since then, demand has accelerated significantly. We now expect the addressable market in AI & Cloud to grow at a 27% CAGR (2025–2028), compared to the 16% we estimated in November. Across the supply chain, demand is accelerating and lead times are extending, reflecting the scale of investment underway.

So demand is coming for the parts of the business that are growing much faster, which you could argue is not fully priced in, but at the same time the legacy business is the core which has the worst growth and could be the reason the stock is not higher today.

The bad news is that the legacy telecom business still matters a lot, and that business remains cyclical, competitive, and lower growth. Mobile infrastructure growth is still weak overall, telecom operators remain cautious on capex, and companies like Ericsson and Chinese vendors continue to pressure pricing. Nokia’s overall operating margins are still far below software or semiconductor companies, meaning even with AI growth, this is not suddenly going to trade like NVIDIA. The company is also in the middle of a major strategic transformation, integrating acquisitions like Infinera while simultaneously repositioning around AI, cloud, and 6G infrastructure. Execution risk is real.

The other major risk is valuation expectations expanding too quickly ahead of actual earnings power. Nokia’s stock has already rallied significantly as investors price in the AI/networking story, with shares hitting multi-year highs after recent earnings. But AI-related revenue is still a relatively small percentage of total company sales today. If telecom spending weakens again, or if AI infrastructure growth slows after the current hyperscaler buildout cycle, the market could quickly revert to valuing Nokia as a low-growth telecom equipment company instead of an AI infrastructure winner. The key question for the next 3–5 years is whether AI/data center networking becomes large enough to structurally change Nokia’s revenue mix and margins.

Conclusion

Nokia is no longer just a legacy telecom equipment company trying to survive the 5G cycle. The company is actively repositioning itself around AI infrastructure, cloud networking, optical systems, edge compute, and eventually AI-native 6G networks. The Nvidia partnership is important because it validates Nokia as a credible player in the next phase of AI infrastructure, where telecom networks may evolve into distributed compute platforms rather than simply pipes for internet traffic.

The bullish case is that Nokia sits at the intersection of several major long-term trends: AI data center expansion, sovereign AI infrastructure, fiber and optical networking demand, edge AI, defense communications, and telecom modernization. Financially, the company is healthier than it has been in years, with improving margins, strong free cash flow, a solid balance sheet, and growing exposure to hyperscaler and AI-related spending. If AI networking becomes a larger percentage of revenue over time, Nokia could eventually receive a valuation rerating away from traditional telecom multiples.

The bear case is that telecom remains a difficult industry structurally. Carrier spending cycles are slow, competition is intense, and AI-RAN monetization is still largely theoretical today. While the vision is compelling, it may take years before telecom operators generate meaningful returns from edge AI infrastructure, and there is no guarantee Nvidia’s telecom strategy becomes as successful as its hyperscaler strategy. Nokia also still carries significant exposure to slower-growth legacy businesses that could weigh on overall performance.

Overall, Nokia looks less like a pure turnaround story and more like a long-duration infrastructure bet on how AI reshapes global networks over the next decade. The stock likely will not behave like a high-growth AI semiconductor company, but if management executes successfully and AI-native networking adoption accelerates, Nokia could evolve from a low-multiple telecom vendor into a more strategically important AI infrastructure platform than the market currently appreciates.

My position continues to remain small and speculative, if more evidence continues to prove the thesis is stronger over time and the stock gives an opportunity at a discounted rate, I would buy more — at this point it’s about my overall conviction level, more deals, potentially a greater dip (like if all semis fall and we go to $11) and more usecases for Edge AI becoming visible which would make it obvious that Nokia could benefit from that super cycle. I would sell not based on a broad semi dip but more so based on lack of execution from Nokia which as of now seems to be fairly strong in an environment that requires extreme discipline to capitalize on the opportunity in front of them.

Thank you for taking the time to read and let me know any ideas you’d like to see me write about!