6G and beyond: what's really coming

I spent twenty-five years inside enterprise data centers — the infrastructure that has to hold when everyone's watching. So when the next generation of mobile networks starts getting hyped, my reflex is the same one I had as a Solution Architect: ignore the marketing, read the spec, and find out what's actually being built.

Here's the short version. 6G is no longer a vision deck. In June 2026, the global standards body 3GPP met in Singapore and locked the timeline. The 6G work item — the formal scope of the new radio — is due for approval in March 2027. The final freeze of the first 6G release (Release 21) is set for March 2029. Commercial networks are expected around 2030.

A lot of the foundations are already decided. During 2025, engineers agreed on the basics of the new air interface — the waveform, modulation, frame structure and channel coding. And tellingly, 5G's channel codes will largely be reused. That one detail says a lot: 6G is far more an evolution of 5G than a revolution.

Strip away the buzzwords and 6G comes down to five things.

More capacity, in new spectrum

The most important new band is the so-called upper mid-band, roughly 7 to 15 GHz. It balances range and bandwidth better than the millimetre waves that 5G promised and mostly failed to deliver at scale. The sub-terahertz frequencies you see in headlines are still research, and will likely arrive later and for niche uses.

AI built in, not bolted on

In 5G, artificial intelligence mostly managed the network from the outside — load balancing, power saving. In 6G, it moves into the signal processing itself. More on that in section seven, because it's the biggest change of all.

The network as a radar

Integrated sensing and communication — ISAC — lets base stations detect movement, objects and people from how radio signals reflect. Useful for traffic, drone safety and smart buildings. Also a serious privacy question, which gets its own section.

Satellites and ground networks as one

Non-terrestrial networks are part of 6G from day one, so coverage extends to sea, wilderness and, for us in Finland, the far north.

Security and energy efficiency

Both are weighted far more heavily than before, including preparation for the cryptographic threat from future quantum computers.

The biggest lesson from 5G is that the transition was made far too complicated. Five migration options were standardized; only two were actually used. Many operators are still in the middle of moving to "real", standalone 5G years after launch.

This time the industry is rallying around one approach: Multi-RAT Spectrum Sharing, or MRSS, which lets 6G run in the same spectrum as 5G. The operator alliance NGMN has been blunt about it — standalone 6G is the goal, MRSS is the baseline, and any alternative has to prove it's worth the extra complexity. At the September 2026 plenary in Madrid, the first performance evaluation of MRSS was delivered, measuring what it costs to run 5G and 6G side by side on the same frequencies.

Sensing through radio isn't new. Wi-Fi already does it: some routers and mesh systems detect motion in a home without cameras, purely from changes in the signal, and the IEEE has standardized it as 802.11bf. The principle works.

But 6G sensing is not simply Wi-Fi scaled up. It's a parallel line of development from the same idea, and the differences matter:

• Scale. Wi-Fi sensing works indoors over tens of metres. 6G base stations cover streets, roads and open areas — vehicles, drones, crowds.
• Precision. Wi-Fi mostly detects that something moved. 6G aims for something closer to radar — distance, speed and direction — using wide bandwidth, large antenna arrays and precise synchronization between sites.
• Ownership. Wi-Fi sensing lives in your own network. 6G sensing would be an operator service. That opens business models — and takes privacy to an entirely different level.

The most likely outcome: Wi-Fi handles sensing indoors, 6G handles the outdoors and wide areas, and the two complement each other.

This is the question I get asked most, so it deserves a straight answer. The research is reassuring — but science never offers absolute certainty, and a lot of what circulates online has no basis at all.

What we know

Mobile network radiation is non-ionizing. It doesn't carry enough energy to break chemical bonds or damage DNA directly, unlike X-rays or UV. The only established biological effect is heating, and exposure limits are set well below the level where that happens. Your biggest personal exposure usually comes from your own phone next to your head, not from masts.

Where the concern comes from

In 2011, the World Health Organization's cancer agency, IARC, classified radio-frequency fields as "possibly carcinogenic" — Group 2B — based on limited evidence. That category means the evidence is weak and uncertain; it shares the group with things like aloe vera extract. Later large studies, including a WHO-commissioned systematic review in 2024, found no link between mobile phone use and brain cancer. IARC's advisors have recommended a fresh evaluation as a high priority for 2025–2029, and as far as I know it hasn't been published yet.

Higher frequency doesn't mean more dangerous

It's actually the opposite. The higher the frequency, the less it penetrates the body. Millimetre waves and any future sub-terahertz bands are absorbed in the outer layers of the skin and don't reach internal organs. The honest gap is that these frequencies have had less long-term study, simply because they weren't used widely before. Physics doesn't predict surprises, but research continues.

Heavy metals and mRNA

A popular claim is that metals or mRNA in the body interact with these frequencies. There's no evidence for it, and the physics explains why. For metal to absorb a radio wave meaningfully, it has to act like an antenna — roughly the size of the wavelength, which is centimetres. Heavy metals in the body exist as dissolved ions, millions of times too small. mRNA vaccines contain no metal or magnetic material at all, and the body breaks the mRNA down within days. What is real: larger metal implants like pacemakers, which is why the long-standing advice is not to keep a phone directly over one.

The radio network equipment market is extremely concentrated. In the first half of 2026, five vendors — Huawei, Ericsson, Nokia, ZTE and Samsung — held about 96 percent of it. Huawei leads globally; outside China, Ericsson leads.

• Huawei is technically very strong and invests heavily in 6G research, but geopolitics largely shuts it out of Western networks, including Finland's.
• Ericsson is Nokia's direct rival and the Western market leader, and is going through a leadership change this autumn.
• Samsung is smaller in radio networks but strong in open networks — and it's the only one that also makes phones and chips.
• Qualcomm dominates phone modems and is one of the most active voices in standardization.
• Nvidia is the newcomer, trying to turn base stations into AI computers.

Nokia's big bet

Nokia has made the boldest strategic move of the group. Nvidia invested one billion dollars in the company, roughly a 2.9 percent stake, around a shared vision of AI-RAN: base stations whose processing runs on Nvidia's GPU platform, with spare capacity sold for AI workloads at the edge of the network. Nokia has called it the final pivot from its own custom chips to merchant silicon, and has run tests with operators including T-Mobile US, Indosat and SoftBank, with Finland's Elisa among its partners.

It's a genuine wager. Critics point out that GPUs are expensive and power-hungry for mobile networks, and Ericsson has chosen a different path. If AI traffic grows as forecast, Nokia could leapfrog its rivals. If not, it could prove costly.

For a country our size, Finland is remarkably well placed: Nokia, the University of Oulu's 6G Flagship — one of the first large 6G research programmes in the world — and serious quantum expertise.

This may be the most important part of the whole story. AI and 6G connect on three levels: AI inside the radio, AI running the network, and the network built to carry AI. In 5G, only the second was really in use. In 6G, all three are design principles.

1. AI in the radio itself

Traditionally, radio is human-designed mathematics: channel estimation, signal detection and error correction all follow carefully derived algorithms. 6G starts replacing parts of that with learned models.

It's already begun. 3GPP studied AI for channel feedback, beam management and positioning, found real performance gains — and also hit a hard problem: making models work across different vendors' equipment. The most interesting example is the two-sided model, where half a neural network runs in your phone to compress channel information and the other half runs in the base station to decompress it. It saves airtime. But if the phone is a Samsung and the base station is a Nokia, how do you train the two halves to fit together?

That's the real tension in 6G. Standards have always meant every device behaves identically and predictably. Neural networks are harder to test and don't always behave deterministically. So the direction is to standardize what goes in and out of the models — not the models themselves — and always keep a conventional fallback.

2. AI running the network

This is closest to everyday reality. The goal is an autonomous network that predicts congestion and failures, powers down parts of base stations at quiet times, fixes faults itself, and works on intent: the operator states the goal — "guarantee five milliseconds of latency for this factory" — and the network decides how.

3. The network built for AI

This explains Nvidia's investment in Nokia. AI devices — glasses, agents, sensors — constantly send images, sound and data upward for analysis, while today's networks are optimized for downloading video. Running AI models next to the base station instead of in a distant data center cuts latency from tens of milliseconds to a few. And the same hardware can run the radio at rush hour and sell capacity for AI the rest of the day. For operators who never found the new revenue 5G promised, AI is the candidate "killer app".

Quantum technology connects to 6G in four ways, at very different levels of maturity.

Quantum-safe encryption — the urgent one

Today's public-key encryption could in principle be broken by a large enough quantum computer. No such machine exists yet, but the threat is already real because of "harvest now, decrypt later": adversaries can store encrypted traffic today and decrypt it years from now. Since 6G networks will run into the 2040s, they must be quantum-safe from the start. The answer is post-quantum cryptography — new algorithms that even a quantum computer can't crack. The first standards were published by the US standards institute NIST in 2024. The challenge is that keys are larger and computation heavier, which matters for small IoT devices. But it's essentially a software upgrade, and it will happen.

Quantum key distribution

QKD uses the laws of quantum physics: any eavesdropping disturbs the quantum state and is detected. Security based on physics rather than mathematics. But it needs fibre or a direct optical link and special hardware. It belongs in backbone networks and government links — not in your phone.

Quantum clocks and sensors — the underrated one

6G needs extremely precise timing, especially for sensing and positioning. Today that often relies on GPS, which can be jammed — as we've seen around the Baltic Sea in recent years. Quantum-based atomic clocks could provide a jamming-resistant time source independent of GPS. For national resilience, that's significant.

Quantum computing — the speculative one

In theory, quantum computers could optimize huge networks better than classical machines. In practice, they're still too small and noisy, and classical AI does the job better for now. The quantum internet belongs to the late 2030s or beyond.

The critics of 6G are mostly right here. But the outcome isn't fixed — it depends on how the technology is designed and regulated.

Start with the baseline: your operator already knows where your phone is, because the phone has to talk to a base station. 5G actually improved one thing — your phone's permanent identifier is now sent encrypted, which made fake base stations, so-called IMSI catchers, much harder to use. Standards can protect privacy, too.

Four new risks

• The network sees people without phones. Sensing detects anyone moving through the area — including someone who left their phone at home or never owned one. Consent-based models don't work when the person detected isn't a customer. Research on Wi-Fi sensing has even shown people can be told apart by the way they walk.
• Positioning gets far sharper. 6G aims for centimetre-level accuracy. "Somewhere in this district" becomes "third table on the left in this café".
• AI is hungry for data. The more is collected, the greater the risk of misuse, leaks — and the quiet expansion of purpose, historically the most common way privacy erodes.
• Data combination. No single source may be a problem. Location, sensing, AI agents, IoT sensors and edge video combined create a picture none would produce alone.

The tools exist

Processing on the device instead of in the cloud. Federated learning, where data stays on devices and only model updates are shared. Differential privacy. Designing sensing to output "twelve pedestrians at this crossing" rather than individual tracks. The problem is that all of these are optional, and usually more expensive than collecting everything.

Regulation is at a turning point

The EU has the world's strongest privacy rules — the GDPR, the AI Act and specific rules for electronic communications. But in November 2025 the European Commission proposed the so-called Digital Omnibus, which would amend the GDPR, the AI Act and cookie rules, including clarifying how personal data can be used to train AI. It's still being negotiated. Supporters call it simplification and competitiveness; critics fear a watering-down of data protection exactly where AI is concerned. The outcome will shape what 6G privacy looks like in Europe.

A useful way to think about device privacy is three layers: the app and operating system layer, the modem layer, and the network layer. The operating system you choose matters enormously for the first, partly for the second — and not at all for the third.

GrapheneOS

The hardened, privacy-focused Android is excellent at the app layer: Google services optional and sandboxed, per-app network and sensor permissions, fast updates, and the option to switch off 2G to reduce fake-base-station attacks. It also isolates the modem from the rest of the phone. What it can't do is see what the modem itself sends — and in 6G, much of the new AI telemetry and data collection will likely live exactly there. After years on Pixel phones only, GrapheneOS announced a long-term partnership with Motorola at MWC 2026, with the first devices expected in 2027.

iOS — and a quiet breakthrough

Strong app-layer privacy, but a closed system you have to trust. The interesting news is at the modem layer. In iOS 26.3, Apple added "Limit Precise Location", which reduces how precisely the operator can locate your phone — down to roughly a neighbourhood rather than a street address — without affecting emergency calls. It only works on devices with Apple's own C1 or C1X modem (iPhone Air, iPhone 16e, the M5 iPad Pro) and needs operator support. But it proves the principle: a maker that controls the modem can protect the layer that operating systems alone can't reach.

Android

"Android" is really many systems. Pixels get fast updates but heavy Google data collection; other makers add their own services and policies. Google introduced a similar location-privacy interface in Android 15, but the modem vendors have to implement it, and most haven't. That's Android's structural weakness here: Google doesn't make the modems.

Windows and Linux — no longer exempt

It's tempting to think this is a phone problem. It won't stay one. 5G laptops already exist, many with eSIM, and when the network is everywhere, computers move into the same position as phones. Many homes already get their internet through a 4G or 5G router, which puts every device behind it on the mobile network.

Windows gives coarse control over the modem — on or off — doesn't own a modem of its own, and keeps its own telemetry running whenever there's a connection. Linux is surprisingly strong: the modem management software is open, so you can see and control what the operating system asks of the modem, such as turning off its built-in GPS. The modem's own firmware is still a black box. Computers do have one real advantage over phones: physical control. Modems are often removable cards, and some privacy-focused laptops have hardware kill switches.

No operating system protects you from network sensing, because it reads your body, not your device. VPNs and Tor hide what you do and who you talk to — not where you are. Network-level privacy is a collective question, which is why I'd like to see an Apple-style privacy flag made mandatory in the 6G standard for every device, not left as one manufacturer's special feature.

This is my home ground, and it's where change is fastest. In short: after 6G, the line between the network and the computer disappears. Computing stops being a central place the network carries data to, and becomes a continuum from your device to the base station to regional and hyperscale data centers.

Energy is the new bottleneck

AI has changed the physics of the data center. A traditional server rack drew roughly five to fifteen kilowatts; AI racks draw a hundred or more, and next-generation plans talk about racks approaching a megawatt. Air cooling can't keep up, so direct liquid and immersion cooling are spreading fast. Data centers get built where the power is, not where the users are.

Waste heat becomes a resource — and here Finland leads the world. Microsoft and Fortum are building a data center region in Espoo, Kirkkonummi and Vihti whose waste heat feeds the district heating network; the first phase is due around the turn of 2026–2027 and construction of the second begins this year. Google's Hamina site cools with seawater. In Finland, a data center can be a heat producer, not just a power consumer.

Five shifts in server technology

• Accelerators over general CPUs. Nvidia leads today, but the big cloud companies build their own chips — Google's TPU, Amazon's Trainium, Microsoft's Maia.
• Memory detached from the processor. CXL lets memory be pooled between servers; in AI, memory bandwidth is often a worse bottleneck than raw compute.
• Light replaces copper. Silicon photonics and co-packaged optics bring optical links right next to the chip — an area Nokia strengthened by acquiring Infinera in 2025.
• Arm challenges x86. Performance per watt is now the metric that matters most.
• Software separates from hardware. Network functions run as software on general hardware, so networks and data centers start to look the same.

Sovereignty and quantum

Geopolitics decides where data lives. The EU increasingly expects Europeans' data to stay in Europe under European law, and "sovereign cloud" — infrastructure run by a European operator rather than a US company — is in demand. On the quantum side, quantum computers won't replace servers; they'll attach to supercomputers as specialist accelerators, the way GPUs once did. Finland is well placed here too, with the LUMI supercomputer in Kajaani and quantum hardware from IQM and VTT.

At Davos in May 2022, Nokia's then-CEO Pekka Lundmark painted a vivid picture: by the time 6G arrives around 2030, the smartphone as we know it would no longer be our most common interface, and a phone call could mean saying who you want to talk to and seeing a holographic image of them appear in front of you. Four years on, it's worth checking that vision against reality.

Three different things called "hologram"

• True holography reproduces the light wavefront so the image is genuinely three-dimensional in the air. Physically very hard and data-heavy. The floating person from the movies is still far off.
• Light-field displays show a different image to each eye and angle, giving convincing 3D without glasses — but only within the screen. Google's Beam 3D video calling works this way.
• Augmented reality glasses draw a 3D model of the other person into your field of view. This is the most likely route into everyday life.

Where 6G comes in — and where AI does

Volumetric video needs many times the data of normal video, and very low latency — delay feels unnatural in conversation, and in AR glasses it causes nausea. Glasses have no room for heavy graphics, so rendering moves to computing next to the base station. That's exactly the infrastructure from the previous sections.

But the bigger shift is this: instead of sending millions of 3D pixels, you send a description — expressions, movement, posture as a few hundred numbers — and AI on the other end rebuilds a realistic 3D image from a model it already has of the person. It's called semantic communication, a hot 6G research topic: the network transmits meaning, not raw bits. It could cut bandwidth dramatically. It also means the person you see is no longer an image but an AI reconstruction — which raises questions about authenticity and deepfakes.

Technology leaders' visions tend to follow a pattern: the direction is right, the timeline too fast. Devices and human habits change more slowly than network technology.

If I read 6G the way I used to read an enterprise architecture — for what's actually load-bearing, not what looks good on the slide — this is what I see.

• 6G is evolution, not revolution. Built on 5G, phased in on the same spectrum, real from around 2030.
• The real change is AI merging with the network. Most of it will happen invisibly, inside the infrastructure.
• The network becomes a computer. Base stations turn into small data centers; the line between telecom and cloud disappears.
• Quantum shows up first as encryption. Everything else is further away than the headlines suggest.
• Privacy is the decisive question. A network that can sense people without devices is a principled change, and no phone setting fixes it. Standards and law will.
• Finland is unusually well placed — Nokia, Oulu, quantum expertise, clean power and cold air. The question is how much of the value stays here.

Every technology generation comes with a story that's bigger than the reality, and a reality that's more interesting than the story. The useful skill — in infrastructure and, I'd argue, in life — is learning to tell the two apart, and then deciding what you'll actually do about it while there's still time to shape the outcome.

Henri Isola

The Code Breaker · Holistic & Business Coach

Twenty-five years as a Solution Architect and Senior Technical Expert in enterprise data centers and cloud migrations — IBM, HP, Sanoma, CGI. Now reading human systems instead. Also read Invisible machines: nanotech, AI and the end of anonymity.