Future Technology in 2026: 10 Innovations That Will Shape the Next Decade

 Future Technology in 2026: 10 Innovations That Will Shape the Next Decade

Future Technology in 2026: 10 Innovations That Will Shape the Next Decade

Introduction

Technology is increasingly advancing in tandem. Progress in one area can strengthen progress in another, creating connections that are becoming harder to ignore.

But how will these innovations reinforce one another in practice?

AI can make robots and autonomous systems more capable. Better computing can accelerate scientific research and drug discovery. Advances in batteries can support wider electrification. Meanwhile, quantum computing is pushing organizations to rethink long-term cybersecurity.

Yet a technical breakthrough is only the beginning. What happens next depends on much more than technical performance. Cost, manufacturing capacity, supply chains, regulation, safety, skills, and real-world reliability can all influence whether an innovation remains in research, moves into early deployment, or reaches a much wider market.

The energy sector offers a clear example of this gap between breakthrough and adoption. The IEA's Energy Technology Perspectives 2026 highlights how supply-chain risks, manufacturing costs, and industrial policies can affect the competitiveness and expansion of clean-energy technologies.

This article examines 10 innovations selected for their combination of broad potential impact, meaningful progress by 2026, and substantial research or investment activity. Each section looks at what is already being used, what is changing, and what still stands in the way of wider adoption.

The final sections bring these developments into a broader picture. How these technologies interact may matter as much as their individual progress. The article therefore examines their convergence, the challenges that could affect adoption, and how their maturity could change by 2036.

Key Takeaways

• AI is becoming more capable and more widely integrated — its reach now extends from software into robotics, scientific research, autonomous systems, and cybersecurity. Reliability, computing costs, data, and governance still decide where it actually gets used.
• Powering that growth takes electricity. The IEA projects global data-centre consumption could roughly double from 2025 to about 950 TWh by 2030; AI-focused facilities are growing faster than the sector overall.
• Practical large-scale quantum computing isn't here yet. Hardware, error correction, and algorithms — in that order — will decide how soon it arrives.
• Clean energy and better batteries have moved past early adoption: the IEA puts the combined global market for clean energy technologies at nearly $1.2 trillion in 2025, after roughly 20% annual growth over the past decade.
• Autonomous vehicles, smart cities, and advanced robotics are heading toward broader deployment. Safety standards, infrastructure, regulation, reliability, and human oversight will decide how fast they scale, and by 2036 that pace could look very different from one country or industry to the next.
• Genomics, biological data, and computational tools keep expanding what researchers can investigate, pushing treatment and drug discovery in biotechnology and precision medicine toward more targeted results. What doesn't move as fast is everything downstream of the lab — clinical validation, cost, manufacturing, regulation.
• Spatial computing and extended reality have found real traction beyond entertainment, in training, design, visualization, and collaboration. Hardware costs, usability, and software ecosystems are the ceiling on how far that spreads.
• The global space economy reached $613 billion in 2024, and the commercial sector accounted for 78% of that growth, per Space Foundation. Keep that pace up, and space-based services won't just support infrastructure by 2036 — they'll be part of it.
• Cybersecurity is shifting on two fronts at once: AI sharpens detection and response, but it hands attackers new tools too. Add the looming security implications of quantum computing, and it's no surprise that 94% of respondents to the World Economic Forum's 2026 survey named AI the top driver of change in the field this year.
• Individual breakthroughs matter less than how they combine. AI, robotics, energy, biotechnology, quantum computing, and connected systems can reinforce each other, producing capabilities no single technology gets to alone.
• Technical breakthroughs won't decide adoption by 2036 on their own. Costs, infrastructure, skills, regulation, safety, supply chains, and public trust will determine which innovations actually go from promising to widely deployed.

1. Artificial Intelligence

Artificial intelligence is becoming a general-purpose technology. Its applications now extend across software, business, research, and consumer products, and in 2026 the frontier has shifted. Progress is increasingly focused on systems that can reason through complex problems, work across different types of information, use digital tools, and support longer workflows.

The result is a transition from isolated AI experiments toward broader integration into the systems people and organizations already use.

How AI Is Evolving in 2026

The capabilities of leading AI models continue to advance rapidly. Stanford's 2026 AI Index reports substantial gains in reasoning, coding, multimodal understanding, and scientific tasks. Some evaluations designed to remain difficult for years are becoming saturated within months, a pace that shortens how long any single benchmark stays useful for measuring progress.

One shift matters more than the rest: AI is moving from producing individual outputs toward completing sequences of tasks. Modern systems can work with text, images, files, software tools, and code within the same workflow, opening the door to more complex processes than simply answering isolated questions.

AI agents represent a further step. Rather than responding to one prompt, an agent pursues a goal through several actions, using tools and responding to intermediate results along the way. Stanford's testing captures how fast this capability has moved: performance on the OSWorld computer-task benchmark rose from roughly 12% to 66.3% in a single year. Leading systems still fail about one in three attempts on structured tasks, though.

Laboratory performance is only part of the picture. A 2026 McKinsey global survey found that 40% of respondents at organizations with more than $1 billion in annual revenue reported scaling AI agents, up from 27% the previous year. Among smaller organizations, the figure sat at 22% and barely moved. Agentic AI, in other words, is starting to shift from experimentation to operational use, mostly at the largest companies.

Why does this distinction matter? Because using AI for assistance is fundamentally different from handing an AI system responsibility for a longer chain of actions. As autonomy increases, so does the need for stronger controls, monitoring, and ways to check whether a system's actions stay within its intended boundaries.

Where AI Is Already Making an Impact

AI now touches a wide range of practical tasks. Businesses use it for content creation, software development, customer support, document analysis, information retrieval, and workflow assistance, while researchers rely on it to process large datasets, identify patterns, and explore potential solutions to scientific problems.

Adoption is no longer confined to a small group of technology companies. According to the OECD, 20.2% of firms in the OECD countries for which data were available reported using AI in 2025, up from 14.2% in 2024 and 8.7% in 2023. That average hides real spread: adoption exceeded 35% in several Nordic countries, including Denmark, Finland, and Sweden, and company size still shapes the picture sharply, with 52.0% of large firms reporting AI use compared with 17.4% of small firms.

Individual use is expanding just as fast. More than one-third of people across OECD countries reported using generative AI tools in 2025; among students aged 16 and over, the figure was closer to three-quarters. Generative AI, in short, is becoming part of ordinary digital activity rather than staying confined to the workplace.

Science is following the same trajectory. AI can help researchers search large bodies of information, analyze complex datasets, generate hypotheses, and prioritize which avenues are worth investigating, reducing, though not replacing, the human expertise and validation those avenues still require.

None of this is confined to one industry, and that is precisely the point. The same underlying capabilities adapt to education, software, research, business operations, creative work, and beyond. This general-purpose character is also why AI resurfaces throughout the rest of this article, shaping several of the other technologies discussed later.

What Still Limits AI Systems

Rapid improvement should not be confused with consistent reliability. An AI model can produce an answer that sounds convincing while containing factual errors, a problem NIST calls confabulation: the confident presentation of erroneous or false information as fact. The stakes rise sharply once AI outputs feed into consequential decisions.

Even the tools used to measure progress have limits. Stanford's 2026 AI Index reports that some widely used benchmarks contain invalid questions, and that several established evaluations are becoming saturated as models improve. One review cited in the report found invalid-question rates ranging from 2% on MMLU Math to 42% on GSM8K. A high benchmark score, in other words, is evidence of capability, not proof of how a system will behave in unfamiliar, real-world conditions.

Infrastructure adds another layer of constraint. More capable AI systems require substantial computing resources, and growing workloads are pushing up demand for data centres and electricity. The IEA projects that global data-centre electricity consumption will roughly double, from 485 TWh in 2025 to 950 TWh in 2030, with AI-focused data centres alone expected to grow about threefold over the same period.

That number matters beyond energy policy. It ties AI's progress directly to the systems that power it: more efficient models and hardware can lower the resources any single task requires, but wider adoption can just as easily offset those savings by pushing up demand overall.

Then there are the less visible challenges: data quality, security, transparency, evaluation, human oversight. These grow more pressing as AI systems shift from generating information to taking action. McKinsey's 2026 research on AI trust points to the same conclusion: governance and controls now need to address not just what AI systems say, but what increasingly autonomous systems can do.

Bigger models and higher benchmark scores won't settle the next stage of AI development on their own. Capability is advancing rapidly. The engineering, infrastructure, evaluation, and governance needed to make that capability dependable at scale are still catching up.

2. Quantum Computing

Quantum computing is developing alongside conventional computing, not in competition with every task that classical computers already handle well. A quantum processor isn't meant to make web browsing, document editing, or ordinary business software suddenly faster. The more interesting question is narrower: what kinds of problems could quantum machines handle differently from today's computers?

The answer lies in areas such as molecular simulation, materials science, optimization, and cryptography. These are problems where the number of possible states or combinations can become extremely large. In some cases, classical computers struggle to model the underlying physics or search through all the possibilities efficiently.

That doesn't mean quantum computing is ready to replace classical systems. Far from it. In 2026, the technology is moving through a difficult middle stage: the basic principles work, commercial interest is growing, and hardware is improving, but reliability and scale remain major obstacles.

The State of Quantum Computing Today

Quantum computers process information through qubits, which behave differently from the bits used by classical computers. A classical bit has a defined value of 0 or 1. A qubit can occupy a quantum superposition of states, while entangled qubits can create correlations that quantum algorithms can exploit.

It is tempting to describe this as a machine trying every possible answer simultaneously. That explanation, however, is misleading. The useful advantage comes from designing quantum operations so that the measurement process makes useful information more likely to emerge. NIST specifically notes that superposition does not provide an efficient brute-force search through every possible solution.

Here is where the engineering problem becomes serious. Quantum states are fragile. Temperature fluctuations, electromagnetic disturbances, and interactions with the surrounding environment can disrupt them. NIST describes today's quantum computers as rudimentary and error-prone, with leading systems containing hundreds of interconnected qubits and experiencing roughly one error per thousand operations.

So, is having more qubits enough? No. Qubit quality matters as much as qubit count. A processor with a larger number of unreliable qubits may be less useful than a smaller system that performs operations accurately and maintains quantum states for longer.

This is why researchers are working toward logical qubits. Instead of treating one physical qubit as perfectly reliable, error-correction techniques distribute quantum information across multiple physical qubits and use additional measurements to detect and manage errors. The goal is to build quantum systems in which reliability improves as the underlying hardware improves.

Commercial interest is growing at the same time. McKinsey's 2026 Quantum Technology Monitor reports that more than 300 global companies are adopting quantum computing. Its analysis also estimates that investment in quantum-technology start-ups reached $12.6 billion in 2025, 6.3 times the 2024 level.

Those numbers show that companies are taking the technology seriously. They don't, however, prove that quantum computers have already achieved a broad practical advantage over classical computing. Investment and adoption can grow before a technology becomes economically useful at scale.

That distinction is important. We can already see progress in hardware, error correction, and commercial experimentation. What remains uncertain is whether these advances will eventually produce reliable quantum calculations that solve valuable problems faster, cheaper, or more effectively than the best classical alternatives.

Where Quantum Computing Could Matter Most

The potential applications become clearer when we look at the kinds of problems quantum computers are being designed to address.

Molecular and materials simulation is one of the strongest examples. Molecules and materials follow quantum-mechanical rules, yet classical computers often have to approximate increasingly complex interactions. A sufficiently capable quantum computer could represent aspects of these systems more naturally, helping researchers investigate molecular structures, chemical reactions, catalysts, and advanced materials. NIST identifies physical-system simulation, including chemistry and materials science, as an important potential application.

Why does that matter beyond physics laboratories? Better molecular modelling could eventually support areas such as drug discovery and materials development. Quantum processors wouldn't replace the classical systems already used by pharmaceutical or materials researchers. Instead, they could become specialized components within larger workflows, handling calculations for which they demonstrate a genuine advantage.

Optimization offers a different challenge. Imagine an airline trying to coordinate aircraft, crews, airport capacity, maintenance, and connecting flights. A manufacturer faces a similar problem when assigning machines, workers, materials, and production schedules. Logistics companies must balance routes, delivery times, vehicles, and capacity.

The number of possible combinations can become enormous. Quantum algorithms are therefore being investigated for selected optimization problems. But there is a crucial test: does the quantum method actually outperform a strong classical method on a realistic problem? A result that looks impressive in a laboratory does not automatically become a useful business solution.

Financial services provide another possible application. Portfolio construction, risk analysis, and other financial calculations can involve many variables and constraints. Quantum computing may eventually contribute to selected parts of these workloads, but theoretical potential isn't the same as demonstrated business value. Financial institutions would need evidence of a measurable improvement in processing time, cost, or solution quality.

Cryptography creates a different situation because the implications are both promising and disruptive. In 1994, Peter Shor showed that a sufficiently powerful fault-tolerant quantum computer could efficiently factor large integers. That could threaten public-key cryptographic systems based on the difficulty of factoring and related mathematical problems.

The machines required to carry out such attacks at meaningful scale don't currently exist. Even so, the possibility has already changed how organizations think about long-term cybersecurity. It is one reason researchers and institutions are developing post-quantum cryptography before large-scale quantum attacks become technically possible.

Across all these examples, the same principle keeps appearing: quantum computing isn't expected to make every workload faster. Its value will come from finding specific problems whose mathematical structure can be exploited by quantum algorithms in ways that classical systems cannot efficiently reproduce.

Why Quantum Computers Aren't Ready Yet

The biggest obstacle is reliability.

A classical computer can execute enormous numbers of operations while maintaining extremely low error rates. Quantum processors work with fragile physical states, so errors can accumulate as a circuit becomes longer. Once those errors become too large, the final result may no longer be useful.

This makes scaling much harder than simply adding qubits. A practical large-scale quantum computer needs high-quality qubits, accurate gates, reliable connections, precise measurement, and an error-correction system capable of protecting information throughout a computation.

Quantum error correction addresses the problem by distributing information across multiple physical qubits and using additional measurements to detect and manage errors. The objective is to create logical qubits whose reliability can improve as the underlying hardware gets better. NIST research on quantum architectures illustrates why useful quantum computing depends on much more than the raw number of physical qubits.

Recent IBM research provides a concrete example of this progress. In a publication dated September 15, 2026, IBM described a continuum from error mitigation and error detection toward increasingly capable error correction. In one demonstration, a post-selected error-correction method encoded 64 logical qubits using 76 physical qubits and produced an approximately 10× improvement in effective gate error rates for a demanding circuit.

What does that improvement mean in practice? It shows that researchers can reduce the effective impact of errors before fully fault-tolerant quantum computers are available. It doesn't mean the engineering problem has been solved.

Error correction itself comes with a cost. Additional physical qubits, measurements, control operations, and computational resources are required to protect quantum information. If the protection overhead becomes too large, the theoretical advantage of the quantum approach can disappear in a real workload.

There is another problem: practical quantum advantage.

A quantum processor can outperform a classical computer on a carefully constructed benchmark without delivering a useful advantage in an industrial workflow. Classical hardware and algorithms continue to improve, too. A quantum method therefore has to compete with increasingly capable classical alternatives, not simply with an older baseline.

NIST notes that some early demonstrations described as quantum advantage have not established genuine practical usefulness, and classical methods have subsequently matched or exceeded quantum approaches for certain tasks. The distinction matters because a laboratory demonstration and a commercially valuable system are not the same thing.

The physical environment adds another layer of complexity. Superconducting qubits, for example, operate at extremely low temperatures. Trapped-ion systems require precise control of individual ions. Other approaches use neutral atoms, photons, silicon-based devices, or different physical architectures. Each brings its own trade-offs involving speed, stability, control, and scalability.

For that reason, the quantum computer of the future probably won't look like a conventional desktop machine. A more realistic model is a specialized quantum processor connected to classical computing infrastructure. Classical systems can prepare inputs, manage the wider workflow, and analyze results, while the quantum processor handles particular calculations for which it has demonstrated an advantage.

Quantum computing has therefore reached an important transition point. We already know that quantum information can be manipulated and measured. The harder challenge is turning those fragile operations into reliable, scalable computation that delivers measurable value on problems that matter outside the laboratory.

3. Green Energy and Next-Generation Batteries

The transition toward cleaner energy isn't just about adding more solar panels and wind turbines. It's about solving a harder problem: how do you balance electricity supply and demand when generation itself keeps changing? Renewable generation is expanding at record levels, and batteries are becoming central to that balancing act. But grid capacity, supply chains, and project delays are holding the pace back.

The Shift Toward Cleaner Power

Renewable energy didn't just grow in 2025, it broke its own record for the 23rd year running. According to the International Energy Agency (IEA), global renewable capacity additions increased by 16% to around 800 GW. Solar PV accounted for more than three-quarters of those additions, while wind represented about 20%.

Solar led the way. Global solar PV additions exceeded 600 GW of installed capacity in 2025 for the first time, and solar electricity generation reached around 2,800 TWh, more than double its output just three years earlier.

That expansion is reshaping the global electricity mix. Renewable electricity generation continued to grow strongly in 2025, with solar PV making the single largest contribution to the increase in global power generation. Clean-energy technologies aren't experimental anymore, they're becoming a major part of the global power system.

So what's the catch? Solar and wind don't generate power on demand, they generate it when the sun shines and the wind blows. That's exactly why storage matters so much right now.

How Better Batteries Are Changing Energy Storage

Battery storage is the fastest-growing technology in the electricity sector today. The IEA reports that 108 GW of new battery storage capacity was deployed globally in 2025, around 40% more than in 2024, enough to exceed the historical annual record for gas-fired power capacity additions. Total installed battery-storage capacity is now eleven times higher than it was in 2021.

Why does this matter so much? Because batteries can respond in seconds when supply or demand shifts. They store surplus electricity when solar and wind generation is high, then release it when demand rises, which is exactly what's needed to make variable renewable power usable around the clock.

The chemistry behind these batteries is shifting too. Lithium-iron-phosphate (LFP) batteries accounted for around 90% of battery-storage deployments in 2025, up from less than half just five years earlier. They're less energy-dense than some other lithium-ion chemistries, but they're cheaper and handle frequent cycling better, which makes them well suited to stationary storage.

Other chemistries are coming up behind them. Solid-state batteries replace the liquid electrolyte in conventional lithium-ion cells with a solid material, while sodium-ion batteries aim to cut reliance on lithium and other constrained materials. Neither is dominant yet, but both could matter more if manufacturers can hit competitive cost and scale.

The takeaway isn't "bigger batteries." It's a more diverse storage system, where different chemistries and technologies each do the job they're best suited for.

What Still Holds Clean Energy Back

Here's the problem: rapid growth in generation and storage doesn't fix the grid it depends on. That's one of the biggest constraints right now.

The IEA reports that more than 2,500 GW of renewable, storage, and large-load projects are currently stalled in grid-connection queues worldwide. Meeting electricity demand through 2030, the agency estimates, would require annual grid investment to rise by roughly 50% from today's level of around USD 400 billion.

That's a real mismatch. Solar farms, wind projects, and battery facilities can often be built faster than the transmission lines needed to connect them. New high-voltage lines typically take 5 to 15 years to plan, permit, and build in advanced economies, compared with just 1 to 5 years for new renewable projects and under 2 years for EV charging infrastructure. The grid, not the generation, is often the slower half of the equation.

Supply chains add another layer of difficulty. Renewable deployment continues to face challenges tied to supply-chain strain, grid-connection delays, financing pressure, and shifting policy, and batteries bring their own set of questions around mineral supply, manufacturing concentration, durability, and recycling.

Recycling will only grow more important as more batteries reach the end of their working lives, though how much it can actually contribute to future mineral supply depends on how quickly retired batteries become available and how efficiently their materials can be recovered.

Put simply: generation capacity is racing ahead, but grids, storage systems, supply chains, and recycling infrastructure all have to keep pace with it. The next decade won't be decided by better solar panels and batteries alone. It'll be decided by whether the rest of the energy system can catch up.

---

4. Autonomous Vehicles

We're watching a technology move from prototype to product faster than most people expect. Autonomous driving isn't stuck in the lab anymore, it's moving into limited but real commercial use. Advances in AI, computing power, sensors, and electric-vehicle technology are making increasingly capable automated-driving systems possible. But as we'll see, the technology isn't developing evenly: driver assistance is already everywhere, while fully autonomous driving remains boxed into specific operating environments.

From Driver Assistance to Driverless Vehicles

Not all automated-driving technology means the same thing, and it helps to be clear about what we're actually talking about. So what's actually on the road today? Today's consumer vehicles commonly offer Level 1 and Level 2 driver-assistance features, such as adaptive cruise control, lane centering, and automated braking. The driver's still responsible for monitoring the road and taking control when needed.

Higher levels of automation change that relationship, and this is where things start to get interesting. At Level 3, the system can drive under defined conditions but may still hand control back to the human. At Level 4, it can drive without human intervention, but only within specific environments and conditions. Level 5 would mean automated driving under any condition a human could handle, but according to the IEA, that level isn't even in sight yet.

Right now, we're seeing the commercial market concentrate around Level 4 robotaxis rather than universally autonomous private cars. The IEA reports that electric driverless taxis are already operating commercially in more than 20 cities worldwide, mainly in China and the United States.

How AI Is Changing Transportation

AI is becoming the core technology behind advanced automated driving, and once we look under the hood, it's clear how much is actually happening at once. A vehicle has to continuously read its surroundings, identify other road users, predict what they'll do next, and decide how to respond. Cameras, lidar, radar, high-performance computing, and increasingly centralized software architectures all feed into that process.

Electric vehicles turn out to matter a lot here. According to the IEA, every commercial robotaxi service currently in operation runs exclusively on battery-electric vehicles. Why EVs specifically? High-voltage batteries, less mechanical complexity, and precise torque control make them a natural fit for automation systems, and the digital-first architecture EV makers pioneered gives them an edge in the constant data collection and software refinement that self-driving demands.

Cost tells us just how fast the economics are shifting. Waymo has cut its sensor count and switched to cheaper EV models, bringing vehicle cost down to around USD 70,000. Baidu went further: it announced in 2025 that manufacturing costs for its Apollo RT6 had fallen to under USD 30,000, thanks to steep drops in lidar prices in China. Tesla's chasing similarly low costs by relying on cameras alone, though it hasn't secured Level 4 approval for that approach yet.

The field is getting more crowded, too, and we're likely to feel that shift soon. Uber, Lyft, and Bolt have all partnered with self-driving developers, though none is running its own autonomous service yet, many of those partnerships are aiming for a 2026 or 2027 launch, which would bring a wave of new players into the market. And robotaxis aren't the only application worth watching: autonomous systems are also being developed for freight, where repetitive routes and controlled environments make automation easier to pull off.

The Remaining Safety and Regulatory Barriers

Here's the real challenge, and it's one we shouldn't gloss over: it's not just building a car that can drive itself. It's making that car handle the unexpected.

Construction zones, unusual road layouts, emergencies, bad weather, pedestrians, cyclists, unpredictable drivers, any of these can trip up an automated system, and this is exactly where we tend to underestimate the difficulty. That's why Level 4 systems stick to clearly defined operational design domains instead of operating everywhere.

Safety validation is still a live issue. In the US, NHTSA keeps collecting and analyzing crash data involving automated-driving and Level 2 driver-assistance systems, though it cautions that the numbers need careful reading: reporting entities have sometimes classified systems inconsistently, and duplicate reports do happen.

New technology brings new risks, too, and we're only starting to reckon with some of them. The IEA points to growing cybersecurity exposure as vehicles become more software-defined, rising semiconductor demand, and supply chains for autonomous-vehicle components that are already concentrated, particularly in China.

Then there's regulation. A driverless service that works in one city can't just be copied into another, it needs fresh testing, fresh approval, and adaptation to local roads. That's a big part of why robotaxi expansion has stayed concentrated in a handful of cities even as costs keep falling.

So where does that leave us? Commercial driverless services are already running, but universal self-driving is still a much harder engineering, safety, and regulatory challenge. Over the next decade, progress won't come from better AI and sensors alone, it'll depend on reliable performance in messy real-world conditions, lower costs, regulatory approval, and whether people actually trust getting in.

5. Smart Cities

A smart city isn't simply a city filled with sensors and connected devices. The underlying idea is using data, digital systems, and automation to make existing infrastructure more responsive. Traffic lights can adjust to real conditions, buildings can manage their own energy use, and electricity networks can react to demand as it happens rather than on a fixed schedule.

The scale involved is what makes this significant. Cities house more than half the world's population and generate around 80% of global GDP, while accounting for roughly two-thirds of global energy consumption and more than 70% of annual carbon emissions. Improvements to urban infrastructure therefore affect a disproportionate share of the planet's resource use.

When Cities Start Thinking in Real Time

The basic components of a smart city aren't new. Connected sensors, cameras, smart meters, cloud platforms, artificial intelligence, and increasingly capable communications networks have existed in some form for years.

What's changed is the degree to which these systems are now connected rather than operated separately. A traffic-management platform can combine congestion data, public transport, road conditions, and signal timing into a single operational picture. An energy-management system can combine building data with weather forecasts and electricity prices. The result is infrastructure that responds to current conditions instead of running on a fixed schedule.

Energy use in buildings is one of the clearest examples. Smart controls can automatically adjust heating, cooling, and lighting according to occupancy and real-time conditions. The IEA estimates that digitalization in buildings could cut total energy use by as much as 10% between 2017 and 2040, assuming limited rebound effects.

Electricity grids represent another major application. AI and digital tools can help network operators forecast demand, monitor equipment, identify problems early, and extract more capacity from existing infrastructure. That matters because many grids are aging and already operating under increasing strain.

What Smart Technology Can Actually Change

The most useful smart-city applications are often the least dramatic. Smart traffic systems adjust signal timing to actual conditions rather than fixed schedules. Smart street lighting dims or brightens based on real usage. Smart parking systems help drivers locate open spaces, while connected public-transport systems provide real-time route and delay information instead of a static timetable.

Electric vehicles add a further layer. Smart, time-of-use charging can shift when EVs draw power away from the evening peak. In an IEA analysis, unmanaged EV charging could account for as much as 4–10% of evening peak demand in major EV markets such as China, the European Union, and the United States, while shifting charging to off-peak hours could more than halve that contribution. Vehicle-to-grid systems could eventually allow some EVs to feed electricity back into the grid at moments of peak need.

Waste management illustrates the same pattern on a smaller scale. Connected bins report fill levels so that collection routes adjust to actual need rather than running fixed loops. Air-quality sensors can map pollution at the level of individual city blocks rather than citywide averages, and water networks can use monitoring to detect leaks before they become costly problems.

None of this requires transforming an entire city at once. In practice, smart-city development tends to happen one system at a time, gradually connected into a broader whole.

What Still Holds Smart Cities Back

The primary obstacle isn't a shortage of sensors. It's the difficulty of connecting technology to existing infrastructure, budgets, regulation, and the people who use the city day to day.

Data presents a particular problem. A smart city can generate enormous volumes of information on traffic, energy use, movement, buildings, and public services, but collecting data doesn't automatically translate into better decisions. So where does all that data actually go? The World Economic Forum has found that less than 1% of IoT data is fully used, a gap driven by fragmented systems, privacy concerns, and weak data governance.

Privacy and cybersecurity risks grow as cities become more connected. A system that monitors traffic flow raises different questions than one that tracks individual people, vehicles, or locations, and connected infrastructure introduces new exposure to disruption, failure, or attack.

Cost is a further constraint. Replacing streetlights, installing sensors, upgrading buildings, modernizing transport networks, and connecting public infrastructure all require substantial investment, and cities with older infrastructure often face the most difficult tradeoffs about what to upgrade first.

Technology also doesn't automatically improve urban life on its own. A poorly designed digital service can add complexity rather than remove it, and a city with thousands of connected devices can still struggle with congestion, housing, pollution, or unreliable public transport.

The underlying value of smart-city technology lies less in the technology itself and more in its capacity to make infrastructure more responsive, measurable, and efficient. Over the next decade, the cities most likely to benefit from digitalization will be those that combine it with sound infrastructure planning, responsible data governance, and practical services, rather than those that simply add more connected devices.

6. Advanced Robotics

Robots aren't limited anymore to the repetitive movements of a factory arm. They are moving into warehouses, hospitals, farms, hotels, laboratories, and other environments where tasks are less predictable and interaction with people matters more. Advances in artificial intelligence, computer vision, sensors, and machine learning are making robots more capable, but the technology is still far from the general-purpose machines often shown in science fiction.

The change is already visible in the numbers. According to the International Federation of Robotics (IFR), 542,000 industrial robots were installed worldwide in 2024, more than twice the number installed a decade earlier. The operational stock reached about 4.66 million industrial robots. At the same time, professional service robots are expanding beyond traditional manufacturing into logistics, cleaning, agriculture, hospitality, and healthcare.

From Factory Arms to Flexible Robots

Industrial robots remain the backbone of the robotics industry. They're particularly effective at tasks that are repetitive, precise, and carried out in controlled environments. Manufacturing therefore remains a major area of deployment, but the technology is becoming more flexible.

Mobile robots are an important part of that shift. In warehouses and factories, autonomous mobile robots can move materials and goods without following fixed routes. They can navigate around obstacles, communicate with other systems, and adjust their movements as conditions change.

The growth of professional service robotics shows how quickly these applications are spreading. IFR recorded almost 200,000 professional service robots sold in 2024, a 9% increase from the previous year. Transportation and logistics represented the largest application group, with 102,900 units, followed by hospitality robots at more than 42,000 units. Professional cleaning robots grew 34% to exceed 25,000 units, while agricultural robots accounted for nearly 19,500 units, a slight decline driven mainly by cultivation and milking applications.

Healthcare stands out as the fastest-growing category by far. IFR recorded around 16,700 medical robots sold in 2024, a 91% increase driven by sharp gains across several categories: rehabilitation and therapy robots rose 106%, surgical robots grew 41%, and robots for diagnostics and laboratory automation surged 610%. Rehabilitation, therapy, surgery, diagnostics, and laboratory automation are all becoming more important as healthcare systems look for ways to handle growing demand and workforce shortages.

The important change isn't simply that there are more robots. It's that robots are being designed to work in environments where the task can change from one moment to the next.

When Robots Start Working Alongside People

A new generation of robots is being built with closer human interaction in mind.

Collaborative robots, or cobots, can operate near human workers and assist with tasks such as handling materials, assembly, inspection, and packaging. Mobile robots can transport components through a workplace instead of requiring workers to move everything manually. In hospitals and other professional environments, robots can take over repetitive transport or support tasks while people remain responsible for decisions and direct interaction.

Artificial intelligence is changing what these machines can do. Computer vision allows robots to identify objects and understand their surroundings. Machine-learning systems can help them adapt to variations in objects, positions, and environments. Better sensors and onboard computing also allow robots to make more decisions locally rather than depending entirely on fixed instructions.

Humanoid robots are receiving particular attention because many workplaces are already designed around the human body. IFR notes that the goal of humanoid robotics is to create more general-purpose machines that can operate in environments built for people. Companies in China, the United States, and Europe are investing heavily in this direction.

But attention shouldn't be confused with widespread deployment. Humanoid robots are still an emerging area, and questions remain about reliability, cost, dexterity, safety, and whether a human-shaped machine is actually the most efficient design for a particular task.

What Still Holds Advanced Robotics Back

The hardest part of robotics is often not making a machine move. It's making that machine behave reliably when the environment does something unexpected.

A factory robot can perform extremely well when its workspace is controlled and the task is clearly defined. A warehouse robot faces more variation. A robot working around people faces even more. Objects can be moved, surfaces can change, people can behave unpredictably, and a seemingly simple task may require several decisions at once.

Safety is therefore a central issue. Robots need to detect people and obstacles accurately, respond appropriately to failures, and operate within clearly defined limits. As robots become more connected, cybersecurity also becomes part of the problem. A robot that communicates with factory systems or cloud services introduces additional points where software failures or security incidents can affect physical operations.

Cost remains another barrier. Robots require hardware, sensors, software, maintenance, integration, and trained personnel. For some businesses, purchasing a robot is only the beginning; the larger challenge can be redesigning workflows so that the robot actually provides useful value. This upfront burden is starting to ease for some adopters: robot-as-a-service models let companies use robotic systems through subscription or rental agreements instead of buying outright, and IFR reported that the global fleet of robots operating under such service models grew by 31% in 2024. The underlying cost of building and maintaining a robotic system hasn't disappeared, though — it's simply shifted toward more flexible payment structures.

The next stage of robotics is therefore unlikely to be defined by humanoid machines alone. Industrial arms, mobile robots, medical systems, agricultural machines, cleaning robots, and collaborative robots are all developing at the same time. The real shift is toward machines that can handle a wider range of tasks while working safely around people.

Over the next decade, progress will depend less on making robots look more human and more on making them reliable, adaptable, affordable, and useful in the real environments where people work and live.

7. Biotechnology and Precision Medicine

Biotechnology is changing medicine in a different way from many of the technologies discussed so far. Instead of simply making existing tools faster or cheaper, it is giving researchers new ways to understand disease and intervene at the biological level. Genomic sequencing, gene editing, cell therapies, artificial intelligence, and increasingly precise diagnostics are moving medicine toward treatments designed around the biology of individual patients.

That shift is already visible in clinical practice. In 2026, the World Health Assembly endorsed a resolution on precision medicine, defining it around the use of clinical, molecular, genomic, and other health data to guide prevention, diagnosis, and treatment while taking ethical and legal safeguards into account. The same resolution also highlighted a major problem: access to these technologies and the data behind them remains uneven across countries and populations.

From Treating Symptoms to Targeting Biology

Traditional medicine often groups patients according to the disease they have. Precision medicine tries to go further by identifying biological differences that can affect how a disease develops or how a patient responds to treatment.

Genomic information is particularly useful here. Sequencing can reveal genetic variants associated with inherited diseases, help classify some cancers, and provide information that can influence treatment decisions. The approach is not limited to genetics alone. Medical records, molecular measurements, imaging, and other forms of health data can also contribute to a more detailed picture of an individual patient.

Gene therapy takes the idea one step further. Instead of treating only the consequences of a genetic disorder, some therapies aim to modify the underlying biological problem.

A major example is Casgevy, which became the first FDA-approved therapy using CRISPR/Cas9 genome editing when it was approved in 2023 for certain patients with sickle cell disease. The patient's own blood stem cells are collected, genetically edited outside the body, and then returned after preparatory treatment. In July 2026, the FDA expanded the approval to patients as young as 2 years old with sickle cell disease or transfusion-dependent beta thalassemia.

This does not mean gene editing can now repair any genetic disease. It shows something more specific: genome editing has moved from an experimental technology into an approved medical treatment for particular conditions.

What AI and Gene Editing Add to the Equation

Artificial intelligence is becoming another layer of this transformation.

Modern biomedical research produces enormous amounts of genomic, molecular, imaging, and clinical data. AI can help researchers identify patterns within those datasets, predict molecular structures, analyze medical images, and search for potential drug targets. In precision medicine, the value comes from combining these capabilities with biological and clinical information rather than treating AI as a replacement for doctors or researchers.

Gene editing is developing at the same time. CRISPR-based techniques can make targeted changes to DNA, while newer approaches are being investigated to improve the precision, delivery, and range of genetic interventions.

But precision matters for another reason: an unintended genetic change can itself create a safety problem. The FDA issued draft guidance in April 2026 recommending methods for comprehensively assessing off-target genome-editing risks using advanced sequencing technologies. The guidance is still a draft, but it shows how safety assessment is becoming an important part of the technology's development.

The field is also moving toward treatments designed for extremely small patient populations. In February 2026, the FDA proposed a framework for developing individualized therapies for ultra-rare diseases where traditional randomized clinical trials may not be practical because there are too few patients. The framework specifically discusses genome-editing and RNA-based therapies.

That could eventually make medicine more responsive to diseases that affect only a handful of people. It also creates difficult questions about manufacturing, evidence, cost, and how regulators can establish safety and effectiveness when the number of patients is very small.

What Still Holds Precision Medicine Back

The science is advancing quickly, but turning a promising biological discovery into an accessible medical treatment is much harder.

Safety is one obvious constraint. Changing cells or DNA can have consequences that are difficult to predict, which is why long-term monitoring and careful evaluation remain essential. The WHO continues to distinguish between somatic genome editing, which affects treated cells and is already being investigated clinically, and heritable genome editing, which could pass genetic changes to future generations. WHO has emphasized the need for strong governance and has warned against moving prematurely toward clinical applications of heritable human genome editing.

Cost is another barrier. Advanced cell and gene therapies can involve highly specialised manufacturing, complex treatment procedures, and long-term follow-up. Even when a therapy is scientifically successful, making it available to patients at scale is a separate challenge.

Then there is the data problem. Precision medicine depends on high-quality health and genomic data, yet populations are not equally represented in biomedical research. The WHO's 2026 resolution specifically warns that underrepresentation can limit the benefits of precision medicine and potentially widen existing health disparities.

Privacy also becomes more complicated when medical systems can analyse increasingly detailed information about a person's biology. Genomic data is not simply another medical record; it can reveal information about relatives and potentially about future health risks.

The next decade will therefore not be defined by gene editing alone. The larger change is the convergence of genomics, biotechnology, AI, diagnostics, and clinical medicine. If these technologies become safer, more affordable, and more widely accessible, treatment could become increasingly tailored to the biology of individual patients rather than based mainly on what works for the average patient.

8. Spatial Computing and Extended Reality

Spatial computing is moving extended reality beyond the idea of simply putting a digital image in front of a user's eyes. The technology connects digital information with the physical environment, allowing people to see, manipulate, or receive computer-generated information in relation to the space around them. Extended reality (XR) includes virtual reality (VR), augmented reality (AR), and mixed reality (MR), but the boundaries between these categories are becoming less important as devices combine several capabilities.

The change is visible in the hardware market. Worldwide smart-eyewear shipments reached about 3.566 million units in the first quarter of 2026, up 130.1% from a year earlier. Audio and audio-capture glasses accounted for about 2.248 million of those shipments, while AR/VR devices reached approximately 1.318 million, an 85.9% year-over-year increase. By the second quarter, worldwide shipments of XR headsets and glasses were still growing, although the annual growth rate had moderated to 35.3%. Display glasses accounted for 14.3% of second-quarter shipments, compared with 11% a year earlier.

These figures reveal an important change in the market. Growth is no longer coming only from conventional VR headsets. Smart glasses, including devices without displays, are becoming a much larger part of the category. That matters because glasses are easier to imagine as an everyday interface than a device that requires users to isolate themselves inside a fully virtual environment.

Bringing Digital Information Into Physical Space

The underlying idea of spatial computing is straightforward: the computer should understand something about the space in which the user is working.

Cameras and sensors can help map surroundings. Computer vision can identify objects and surfaces. Three-dimensional graphics can then place digital information within that environment. The result can be useful even when the user spends most of the time looking at the real world.

Consider a technician inspecting a machine. Instead of moving repeatedly between the equipment and a separate screen, the worker could view instructions or diagnostic information in relation to the component being examined. An engineer could inspect a digital model at realistic scale. A trainee could repeat a procedure inside a simulation before facing the consequences of a mistake in the physical world.

This is where spatial computing differs from simply owning another display. Its potential value comes from connecting information with context.

Training: Evidence Is Growing, but It Is Not Uniform

Training is one of the areas in which XR has accumulated a meaningful body of research. The results, however, are more nuanced than the promotional language surrounding immersive technology sometimes suggests.

A 2026 systematic review examined 38 studies of XR in hygiene education and training. Most studies reported improvements in practical skills, behaviors, or attitudes, while some found limited or no significant gains, particularly for certain measures of knowledge and compliance. The researchers also identified cost, safety, validation, and integration into standard curricula as continuing barriers.

A separate 2026 systematic review focused on whether VR training translates into real-world infection-prevention practice. It found improvements in learner-centered outcomes, but evidence for sustained changes in real-world behavior remained limited and inconsistent.

That distinction is important. A trainee performing better inside a simulation does not automatically mean that the same improvement will appear months later in a real workplace. XR therefore works best when it is treated as part of a broader training system rather than as a replacement for every other form of instruction.

A Real-World Medical Test

Healthcare provides a useful example because the technology has to function under real operational constraints.

A 2026 prospective observational study documented the institutional use of Apple Vision Pro for surgical and procedural visualization at an academic medical center. Thirty-five users employed the system during 174 procedures, producing 227 individual headset uses across minimally invasive surgery, interventional pulmonology, and oculoplastic surgery. No intraoperative complications or adverse events were attributed to the headset. It was removed in six of the 227 uses, or 2.6%, and the researchers observed a learning curve as users gained experience with the system.

The study does not establish that spatial computing is better than conventional surgical displays. It was a single institutional deployment, and conventional monitors remained available as a backup. What it demonstrates is more specific: spatial-computing hardware can be incorporated into real clinical workflows, with measurable operational experience rather than only laboratory testing.

Why Three-Dimensional Work Matters

The strongest applications may emerge in fields where the physical arrangement of objects already matters.

Architecture, engineering, manufacturing, medicine, and maintenance all involve situations in which a two-dimensional screen can be an imperfect representation of the task. A three-dimensional model viewed at the relevant scale can make spatial relationships easier to inspect. Digital instructions can be associated with a particular machine or component instead of appearing as a detached list.

This also explains why spatial computing is closely connected with digital twins and other real-time 3D systems. The device is only one part of the process. The larger system needs accurate models, current data, reliable connectivity, and software capable of translating that information into something useful for the person doing the work.

AI Changes the Interface

Artificial intelligence adds another reason why spatial computing is developing now rather than simply repeating the XR cycle of previous years.

An XR device can provide a view of the environment. AI can help interpret that environment and decide which information is relevant. A user might ask for a translation, identify an object, request instructions, or retrieve information without navigating through a conventional application interface.

The combination changes the role of the device. The glasses or headset do not have to become a complete computer replacement. They can instead become a convenient point through which an AI system observes, interprets, and communicates information about the user's surroundings.

This could eventually make spatial interfaces less dependent on menus and conventional applications. Instead of searching through a series of screens, a worker could ask for information while continuing to interact with the physical environment.

The Obstacles Are Practical

The remaining barriers are therefore not simply about improving graphical quality.

Comfort matters when a device is expected to remain on someone's face for hours. Battery life, weight, field of view, display quality, cost, and reliability all influence whether a technically impressive product becomes something people actually use.

There are software problems as well. Spatial applications require accurate 3D content and environmental mapping, while organizations need ways to integrate them with existing systems. Training research adds another warning: improvements measured in controlled simulations do not necessarily translate into long-term changes in real-world behavior.

Privacy presents a separate challenge. A spatial device may capture images, sounds, movements, objects, and information about people who are not wearing the device themselves. As AI assistants become more deeply integrated with smart glasses, the question is no longer simply what the camera can see. It is also what the system can infer from what it sees.

The Outlook Through 2036

The most consequential future for spatial computing may therefore be less spectacular than the early visions of fully virtual worlds.

A technician could receive instructions while standing beside a machine. An engineer could examine a proposed system before construction. A medical team could share different visual information during a procedure. A trainee could repeat a difficult task in a controlled environment before performing it in reality.

The market is moving toward lighter and more wearable devices, while research is becoming more specific about where immersive technology helps and where the evidence remains incomplete. This suggests that the future of spatial computing will probably be shaped less by immersion alone and more by whether these systems can solve concrete problems.

By 2036, spatial computing could become an ordinary interface between people, AI systems, digital twins, and physical equipment. But that outcome will depend on solving fairly ordinary problems: making devices comfortable enough to wear, affordable enough to deploy, secure enough to trust, and useful enough to justify changing an existing workflow.

That is ultimately the test for spatial computing. The technology does not need to make every interaction three-dimensional. It needs to make the situations in which physical context matters easier to understand and act on.

9. The Commercial Space Economy

Space is no longer a field shaped mainly by government programmes. It has become a broader commercial ecosystem involving launch services, satellite communications, Earth observation, spacecraft manufacturing, and emerging in-orbit services. The change isn't simply about putting more hardware into orbit. It's about building commercial systems that support communications, navigation, data collection, and economic activity on Earth.

The scale of that transformation is substantial. Space Foundation estimated that the global space economy reached $613 billion in 2024, with the commercial sector responsible for 78% of the economy's growth. The estimate uses a different market definition from ESA's European-focused figures, so the two sets of numbers shouldn't be treated as directly interchangeable.

From Government Missions to Commercial Infrastructure

For much of the space age, governments funded the systems that made space useful. Navigation, communications, weather monitoring, and Earth-observation satellites were closely connected to public programmes and national space agencies.

That model has changed. Commercial launch providers now offer reusable rockets and rideshare services, while private satellite operators are building large constellations for communications and other applications.

Launch economics have changed along with the industry. SpaceX's 2026 prospectus cites a historical average launch cost of about $18,500 per kilogram, compared with approximately $2,700 per kilogram for the first version of Falcon 9 in 2010. These are historical comparison points, not current launch prices. SpaceX attributes later reductions in its internal launch costs to engineering improvements, manufacturing efficiencies, economies of scale, and especially more frequent rocket reuse.

Lower access costs helped open the door to a wider commercial market. Starlink is a prominent example, while Eutelsat OneWeb provides another through its low-Earth-orbit connectivity network for land, maritime, and aviation users.

Satellites Are Becoming More Numerous

The growth of satellite constellations is one of the clearest signs of this transition.

In 2025, more than 300 launches took place and more than 4,000 new payloads were placed into orbit. Large constellations and rideshare missions allow many satellites to be deployed during a single launch, making the traditional model of one large spacecraft per mission less dominant.

The shift changes both the economics and the architecture of satellite services. Rather than relying on a small number of very large spacecraft, operators can distribute functions across networks of smaller satellites, add capacity over time, and provide broader or more frequent coverage.

Broadband is a highly visible example. Starlink shows how large low-Earth-orbit constellations can provide connectivity across wide geographic areas. Earth-observation companies are using a similar distributed approach to collect information about the planet at frequent intervals.

Earth Observation Becomes a Data Industry

The value of commercial space increasingly lies in what satellites can measure, process, and communicate rather than in the spacecraft themselves.

Planet offers a clear example. The company currently reports approximately 200 satellites in orbit, with its Dove and SkySat constellations capturing more than 25 terabytes of imagery each day. Planet also notes that its satellite count changes as older spacecraft reach the end of their operational lives and new satellites are launched, so the figure isn't a fixed total.

What can all that imagery actually do? Earth-observation data can support agriculture, infrastructure monitoring, environmental analysis, and disaster response. Planet describes applications that include monitoring agriculture, water resources, land-use change, and environmental conditions. In this model, orbital imagery becomes a continuing data service rather than simply a product produced by a satellite.

The Economics of the Space Industry

Launch services represent only one layer of the commercial space economy. Behind them are spacecraft manufacturing, ground infrastructure, data processing, communications networks, regulatory compliance, and end-of-life spacecraft management.

ESA's 2026 Space Economy Report illustrates the scale of these different layers. It values the upstream market, covering spacecraft manufacturing and launch services, at about €75 billion in 2025. The downstream market, covering services such as satellite communications, Earth observation, and navigation-related applications, was approximately €490 billion.

These figures shouldn't be added to Space Foundation's $613 billion estimate as though they measured exactly the same market. Their definitions and geographic focus differ. Taken together, though, they show how much economic activity now exists beyond the physical construction and launch of spacecraft.

The New Challenge: A Crowded Orbit

Commercial growth in space faces a physical constraint: orbital space is limited.

More satellites and other objects are entering orbit, while debris from defunct spacecraft, rocket bodies, and fragmentation events remains a long-term concern. ESA's 2026 assessment describes Earth's orbital environment as a finite resource and highlights the growing need for better space-traffic coordination and responsible end-of-life disposal.

As of July 2026, space-surveillance networks regularly tracked about 46,950 objects, while the total mass of all objects in Earth orbit was estimated at more than 17,000 tonnes. That mass includes the broader population of tracked and estimated orbital objects, not simply active satellites, making it a measure of the amount of material already placed around Earth.

The long-term concern is the Kessler syndrome, in which collisions create fragments that increase the probability of further collisions. ESA's 2026 modelling shows the runaway effect becoming visible earlier in its long-term projections than in the previous year's assessment. This is a modelled future scenario, not a prediction that a specific collision sequence will occur.

There are signs of progress as well. In 2025, controlled reentries of rocket bodies continued to outnumber uncontrolled ones for the second consecutive year. The transition from a 25-year to a five-year disposal target for relevant low-Earth-orbit activities also reflects a stronger emphasis on shortening orbital lifetimes.

A More Sustainable Commercial Space Economy

The next stage of the industry will involve more than launching additional satellites. Companies and agencies will need to consider how spacecraft are designed, operated, serviced, and eventually removed.

ClearSpace-1 is designed to demonstrate active debris-removal technologies by rendezvousing with and capturing ESA's unprepared PROBA-1 satellite for removal from low-Earth orbit. ESA currently lists the mission with a planned launch in 2029 and describes it as an in-orbit demonstration intended to establish technologies for active debris removal and support a future commercial sector in space.

In-orbit servicing could extend the useful life of spacecraft as well. On 22 September 2026, ESA signed a contract with ClearSpace to advance the Phoenix project, which aims to develop technologies for servicing and extending the operational life of satellites in geostationary orbit. The project is intended to lay foundations for a recurring commercial in-orbit servicing capability in Europe, while its commercial viability will depend on technical reliability, regulation, international coordination, and sustainable business models.

The industry faces broader commercial and strategic questions too. Concentration among a small number of large operators can create dependencies in critical communications infrastructure, while many space technologies have both civilian and defence applications. ESA's 2026 Space Economy Report also highlights the growing importance of defence investment within the space sector.

What Space Could Look Like by 2036

Over the next decade, the most important development may not be a single breakthrough mission but the deeper integration of space services into everyday infrastructure.

Satellite communications could become more closely integrated with terrestrial networks. Earth-observation data could become a routine input for agriculture, logistics, environmental monitoring, and disaster response. Navigation and timing services could support increasingly autonomous systems, while in-orbit servicing could gradually change how spacecraft are maintained.

Sustainability will become just as important as expansion. ESA reported around ten objects being launched into orbit each day in 2025, while more than three intact satellites or rocket bodies reentered the atmosphere each day on average.

ESA has set itself the goal of significantly limiting the production of debris in Earth and lunar orbits from its own future missions, programmes, and activities by 2030 through its Zero Debris Approach. This is an ESA target, not a general 2030 requirement for the entire global space industry.

By 2036, space could be less visible as a separate industry and more noticeable through the services it enables on Earth. Connectivity, Earth-observation data, navigation, and in-orbit servicing could become increasingly integrated into other industries. The commercial opportunity, however, will depend not only on expanding those services but also on keeping orbital environments usable as the number of spacecraft continues to grow.

10. Cybersecurity and Post-Quantum Security

Cybersecurity is entering a period in which artificial intelligence and quantum computing are changing the security landscape at the same time. AI can strengthen detection, analysis, and incident response, but it can also help attackers automate and scale their operations. Quantum computing presents a different kind of challenge: sufficiently capable quantum machines could eventually break some of the public-key cryptography that protects digital communications today.

Very different in nature, both forces are already shaping decisions.

Organizations' risk assessments already reflect the shift. The World Economic Forum's Global Cybersecurity Outlook 2026, based on responses from 800 global leaders, found that 94% of respondents expected AI to be the most significant driver of change in cybersecurity in 2026. A different question produced the 87% figure: that share identified AI-related vulnerabilities as the fastest-growing cyber risk during 2025. The first number concerns the expected influence of AI on the field; the second concerns the perceived growth of a specific category of cyber risk. Neither figure is a measurement of actual cyber incidents worldwide.

AI Changes Both Sides of Cybersecurity

AI is becoming part of the defensive security stack. Security teams can use machine-learning systems to analyze large volumes of logs, identify unusual activity, prioritize alerts, and accelerate parts of incident response. A separate World Economic Forum report published in May 2026 found that 77% of organizations surveyed were using AI in at least some part of their cyber operations. The figure refers to the organizations covered by that survey, not to organizations worldwide.

Attackers gain capabilities from the same technology. AI can support targeted social engineering, automate parts of vulnerability discovery, and increase the speed and scale of malicious activity. Voice cloning is one concrete example: the U.S. Federal Trade Commission has warned that scammers use cloned voices to impersonate family members, business executives, and others in attempts to obtain money. The FTC launched a Voice Cloning Challenge in 2024 to encourage technical approaches for detecting and preventing this type of fraud.

Organizations are targets too. In early 2024, a finance employee at the engineering firm Arup's Hong Kong office was persuaded to make 15 transfers totalling about HK$200 million (roughly US$25 million) after joining a video call with people who appeared to be senior colleagues. Hong Kong police described the participants as deepfake re-creations, and Arup later confirmed that fake voices and images had been used. The case shows how synthetic media can be turned against corporate payment and verification processes, not only against individuals. A familiar face or voice can no longer be treated as proof of identity. For individuals, a hardware security key is one practical way to add phishing-resistant sign-in to important accounts.

AI also creates new security problems inside organizations. Models and AI-enabled applications can introduce additional attack surfaces, while sensitive data may be exposed through poorly designed systems or inadequate access controls. As organizations integrate AI into business processes, securing the AI systems themselves becomes part of cybersecurity rather than a separate technical concern.

The Quantum Security Problem

Quantum computing introduces a longer-term threat to widely used public-key cryptography. Some algorithms that are secure against conventional computers could become vulnerable to sufficiently powerful quantum machines because quantum algorithms can solve certain mathematical problems much more efficiently.

It is not possible to identify a specific date when a quantum computer capable of breaking today's widely deployed public-key cryptography will appear. NIST therefore treats the issue as a migration problem that must be addressed before such a machine can break widely deployed cryptographic systems. Sensitive information collected today could potentially be stored and decrypted later, a risk often described as "harvest now, decrypt later."

The machine may not exist yet, but the exposure may already have begun.

The practical response is under way. NIST finalized three post-quantum cryptography standards in 2024: FIPS 203 for ML-KEM, FIPS 204 for ML-DSA, and FIPS 205 for SLH-DSA. NIST says these standards are ready for implementation and that organizations should begin migrating to quantum-resistant cryptography.

Why Migration Takes Time

Replacing a cryptographic algorithm is not as simple as installing a software update. Organizations first need to identify where vulnerable cryptography is being used across applications, devices, networks, certificates, cloud services, and embedded systems.

Inventory comes first.

Legacy infrastructure can make the process particularly difficult. Industrial equipment, connected devices, specialized hardware, and systems with long operating lifetimes may not be easy to upgrade. Supply chains add another layer of complexity because software and hardware can come from many vendors with different update cycles and cryptographic dependencies.

Timelines are becoming more concrete. NIST's transition planning distinguishes between deprecation and disallowance. A deprecated algorithm may still be used, but its use is discouraged because of growing risk; a disallowed algorithm is no longer permitted. Certain quantum-vulnerable algorithms and security levels are scheduled to be deprecated after 2030, while specified uses are to become disallowed after 2035. The broader objective is to move systems away from vulnerable public-key cryptography before quantum computers can pose a practical threat.

Cryptographic agility will also matter. Systems designed so that cryptographic components can be replaced without rebuilding the entire architecture can adapt more easily as standards evolve or particular algorithms become unsuitable. NIST's migration work emphasizes identifying vulnerable systems and developing practical, interoperable migration paths.

Cybersecurity Beyond Traditional IT

Cybersecurity is also spreading into physical infrastructure. Connected vehicles, industrial control systems, medical devices, energy networks, satellites, and smart buildings increasingly depend on software and communications networks.

That connectivity creates efficiency and automation, but it also increases the consequences of a successful attack. A compromised business application may cause financial or operational disruption; an attack against critical infrastructure can affect physical services and public safety.

When software controls physical systems, the stakes change.

Geopolitical risk is becoming part of the same calculation. In the WEF's 2026 survey, 64% of organizations reported accounting for geopolitically motivated cyberattacks, such as disruption of critical infrastructure or espionage, in their cyber-risk mitigation strategies. The report does not provide a directly comparable 2025 percentage for this specific measure.

Concern about national-level resilience appears to be rising alongside this organizational planning. The share of respondents reporting low confidence in their nation's ability to respond to a major cyber incident increased from 26% in 2025 to 31% in 2026, suggesting that the attention organizations give to geopolitical risk is matched by uncertainty about how well countries could respond to it.

People remain an important part of the equation. The WEF identifies skills shortages and resource constraints as factors contributing to differences in cyber resilience between organizations. Automated tools can process data and accelerate routine tasks, but organizations still need specialists who can configure systems, investigate incidents, assess risk, and make decisions when automated tools produce uncertain results.

What Cybersecurity Could Look Like by 2036

By 2036, cybersecurity may depend less on protecting isolated networks and more on continuously verifying the security of interconnected systems. AI could help security teams analyze threats in real time, automate routine responses, and identify patterns that would be difficult to detect manually.

Post-quantum migration will form part of that longer transition. NIST's timelines for deprecating vulnerable algorithms after 2030 and disallowing specified uses after 2035 provide concrete reference points, although the pace of migration will vary according to the systems involved, their risk levels, and their operational lifetimes.

Security will remain a moving target. New AI capabilities may improve defensive tools while creating new attack techniques, while advances in quantum computing could change the urgency of cryptographic migration. The practical goal is not to predict exactly when these technologies will reach specific milestones. It is to build systems that can adapt as the threat environment changes.

By 2036, cybersecurity may be less about defending a fixed perimeter and more about maintaining trust across constantly changing digital systems. AI-assisted defence, post-quantum cryptography, resilient infrastructure, skilled security teams, and adaptable security practices could become ordinary parts of the technology that businesses and societies rely on every day.

Conclusion

The technologies shaping 2026 are not developing in isolation. Artificial intelligence is becoming more capable and more deeply integrated into software, robotics, scientific research, and cybersecurity. Advances in computing are supporting new approaches to drug discovery and complex scientific problems, while batteries, renewable energy, and new energy systems are influencing how quickly electrification can expand. Space infrastructure is becoming more commercial (more than 4,000 payloads reached orbit in 2025), and connected physical systems are bringing digital technologies into transport, industry, healthcare, and everyday environments.

The coming decade will depend less on whether these technologies can demonstrate impressive capabilities and more on whether they can operate reliably at scale. A system that performs well in a laboratory still has to deal with cost, energy consumption, manufacturing capacity, supply chains, regulation, safety, cybersecurity, and maintenance before it can become widely useful. Capability is only the starting point.

That makes 2036 difficult to describe as a single technological destination. Some innovations may become ordinary infrastructure; others may remain limited to specialized applications. Adoption will also vary by country, industry, and access to investment and technical skills. Progress in the laboratory will not necessarily match progress in deployment.

What is clearer is the growing connection between the technologies themselves. AI can improve autonomous machines and scientific research. Better computing can support AI and biotechnology, but it also raises the stakes for cryptography: NIST's timeline calls for specified quantum-vulnerable uses to be disallowed after 2035. Energy systems influence how much computational infrastructure can expand. Space-based networks can extend connectivity and provide new sources of data. Cybersecurity has to protect all of these systems as they become more interconnected.

The central question is less which technology will advance fastest: it is whether these innovations can be integrated into systems that are affordable, secure, sustainable, and dependable enough to support real-world use.

Many of today's emerging technologies may eventually stop appearing as separate breakthroughs. They could become ordinary parts of the infrastructure behind how people communicate, travel, produce energy, conduct research, manage businesses, and interact with digital systems. None of this is guaranteed. The transition will depend not only on technical progress, but on the decisions made around deployment, governance, investment, and responsible use.

References

1. Stanford Institute for Human-Centered Artificial Intelligence (Stanford HAI). AI Index Report 2026. Stanford University, 2026.

2. McKinsey & Company. The State of AI in 2026. 2026.

3. Organisation for Economic Co-operation and Development (OECD). Artificial Intelligence. OECD, 2026.

4. International Energy Agency (IEA). Energy Technology Perspectives 2026. IEA, 2026.

5. International Energy Agency (IEA). Energy and AI. IEA, 2025.

6. National Institute of Standards and Technology (NIST). Artificial Intelligence Risk Management Framework (AI RMF 1.0). NIST, 2023.

7. National Institute of Standards and Technology (NIST). What Is Post-Quantum Cryptography? NIST. Accessed September 2026.

8. National Institute of Standards and Technology (NIST). Post-Quantum Cryptography. Computer Security Resource Center, updated 2026.

9. National Institute of Standards and Technology (NIST). FIPS 203: Module-Lattice-Based Key-Encapsulation Mechanism Standard. NIST, 2024.

10. National Institute of Standards and Technology (NIST). FIPS 204: Module-Lattice-Based Digital Signature Standard. NIST, 2024.

11. National Institute of Standards and Technology (NIST). FIPS 205: Stateless Hash-Based Digital Signature Standard. NIST, 2024.

12. International Federation of Robotics (IFR). World Robotics 2025 – Industrial Robots. IFR, 2025.

13. International Federation of Robotics (IFR). World Robotics 2025 – Service Robots. IFR, 2025.

14. World Health Organization (WHO). Precision Medicine. WHO, 2026.

15. World Health Organization (WHO). Human Genome Editing: Recommendations. WHO, 2021.

16. U.S. Food and Drug Administration (FDA). CASGEVY. FDA, 2023.

17. U.S. Food and Drug Administration (FDA). Human Gene Therapy Products Incorporating Human Genome Editing: Guidance for Industry. FDA, January 2024.

18. World Economic Forum (WEF). Global Cybersecurity Outlook 2026. World Economic Forum, 2026.

19. World Economic Forum (WEF) and KPMG. Empowering Defenders: AI for Cybersecurity. World Economic Forum, 4 May 2026.

20. Federal Trade Commission (FTC). FTC Submits Comment to FCC on Potential Harmful Effects of AI-Generated Voice Cloning. FTC, 2024.

21. Space Foundation. The Space Report 2025. Space Foundation, 2025.

22. Space Exploration Technologies Corp. (SpaceX). Prospectus. U.S. Securities and Exchange Commission, 11 June 2026, p. 77.

23. European Space Agency (ESA). ESA Space Environment Report 2026. ESA, 2026.

24. European Space Agency (ESA). ESA Space Economy Report 2026. ESA, 2026.

25. European Space Agency (ESA). ClearSpace-1. ESA. Accessed September 2026.

26. European Space Agency (ESA). ESA Supports Development of European Spacecraft for Servicing Geostationary Satellites. ESA, 23 September 2026.

27. European Space Agency (ESA). Mitigating Space Debris Generation. ESA. Accessed September 2026.

28. Eutelsat. OneWeb LEO Constellation. Eutelsat. Accessed September 2026.

29. Starlink. Starlink Business Services. SpaceX/Starlink. Accessed September 2026.

30. Planet Labs PBC. Planet Satellite Imagery and Earth Observation. Planet. Accessed September 2026.

31. Apple. Apple Vision Pro and Spatial Computing. Apple. Accessed September 2026.

32. National Institute of Standards and Technology (NIST). NIST IR 8547: Transition to Post-Quantum Cryptography Standards. Initial Public Draft, 12 November 2024. https://doi.org/10.6028/NIST.IR.8547.ipd

33. Magramo, Kathleen. British Engineering Giant Arup Revealed as $25 Million Deepfake Scam Victim. CNN, 16 May 2024.

34. Federal Trade Commission (FTC). The FTC Voice Cloning Challenge. FTC, 2023–2024.

35. The White House. Executive Order 14412: Securing the Nation Against Advanced Cryptographic Attacks. 22 June 2026.

36. Office of Management and Budget (OMB). M-26-15: Execution of the Migration to Post-Quantum Cryptography. Executive Office of the President, 24 June 2026.