The World in 2035: Ten Technologies That Could Change Everyday Life

A decade is an awkward planning horizon. It is long enough that the technologies which will matter most are already visible in laboratories and early products, and short enough that physics, capital cycles, regulation and human habit will still be recognisable constraints. Nothing in this article requires a scientific miracle. Everything in it requires something harder to arrange: cost reductions, supply chains, evidence, trust and institutional adaptation.

What follows is ten technology families, each with the same structure — what exists today, what could plausibly change by 2035, and what would have to go right or wrong. Treat it as a map of live uncertainties rather than a forecast. Where a claim is speculative, it says so.

How to read a ten-year technology outlook

Two failure modes dominate this genre. The first is straight-line extrapolation: taking a steep recent curve and extending it to absurdity. The second is the opposite error — assuming that because a technology disappointed once, it always will. Both ignore the mechanism that actually governs technology timelines: diffusion.

Diffusion is slow because it is not primarily technical. A capability becomes an everyday reality only after unit costs fall far enough, regulators develop a supervisory approach, insurers price the residual risk, organisations redesign their processes, and enough workers acquire the relevant skills. That chain routinely takes a decade after the demonstration that made headlines.

The useful discipline is therefore to ask three questions of any claim about 2035: is the underlying capability demonstrated or assumed; is the cost trajectory grounded in manufacturing reality; and what non-technical bottleneck decides the timeline?

1. AI agents: software that acts, not just answers

Today. Large language models can plan short sequences of steps, call external tools, write and execute code, and operate browsers under supervision. Reliability degrades as tasks lengthen; small per-step error rates compound quickly across a long chain.

Plausible by 2035. Narrow, well-instrumented agentic workflows become normal in business software: reconciling records, preparing filings, triaging tickets, monitoring systems, assembling first drafts of routine documents. Consumers may delegate bounded errands — comparing options, filling forms, scheduling — with a confirmation step before anything irreversible happens.

What decides it. Not raw model quality alone, but the boring infrastructure of delegation: machine-readable authorisation, spending limits, audit logs, identity for non-human actors, liability allocation and rollback. Governance frameworks such as the NIST AI Risk Management Framework and the EU's regulatory framework for AI matter here as much as model architecture. This remains uncertain: an equally plausible scenario is that agents stay confined to low-stakes tasks because the accountability problem proves harder than the capability problem.

2. Robotics and mobile manipulation

Today. Robots are excellent in structured settings — fixed cells, defined parts, mapped warehouse aisles. General manipulation in cluttered, unpredictable environments remains difficult, and learned control policies are improving but not solved.

Plausible by 2035. Steady expansion into semi-structured commercial environments: logistics, food production, agriculture, inspection, cleaning, some construction tasks. Humanoid form factors may find niches where existing human-shaped infrastructure makes them convenient, but the economic case will be decided per task, not per press release.

What decides it. Dexterity, reliability over thousands of hours, maintenance cost, safety certification and total cost of ownership against the wage it replaces. Household general-purpose robots by 2035 remain speculative.

3. Personalised and preventive medicine

Today. Genomic sequencing is cheap by historical standards. Machine learning supports imaging and pathology workflows in specific approved contexts. Targeted therapies exist in oncology and rare disease. Most care remains reactive.

Plausible by 2035. More systematic use of risk stratification, earlier detection for a handful of cancers, wider deployment of decision support, better matching of patients to therapies, and growing use of continuous data from wearables in chronic disease management.

What decides it. Clinical evidence, reimbursement, data governance and workforce capacity. Laboratory promise routinely outruns clinical adoption — see the health and longevity analysis on this site for the detailed version. This is informational analysis, not medical advice.

4. Energy storage and flexible grids

Today. Lithium-ion costs have fallen substantially over the past decade, making short-duration storage economic in many markets. Long-duration storage is technically diverse and commercially immature.

Plausible by 2035. Storage becomes a normal grid asset rather than a pilot; electric vehicle fleets provide meaningful demand flexibility; sodium-ion and other chemistries reduce reliance on constrained minerals. The International Energy Agency publishes ongoing analysis of how quickly electricity systems are actually changing.

What decides it. Transmission build-out and permitting, not cell chemistry. Grid connection queues and planning consent are the rate-limiting steps in many countries.

5. Advanced computing

Today. Performance gains increasingly come from specialised architectures — accelerators, packaging, memory bandwidth — rather than classical transistor scaling. Quantum computers exist as research instruments with steadily improving error-correction results, not as general-purpose machines.

Plausible by 2035. Continued growth in AI-specific compute, with energy and cooling as first-order design constraints. Quantum advantage on narrow scientific problems is plausible; broadly useful, fault-tolerant quantum computing remains uncertain and should not be assumed in any plan.

What decides it. Fabrication capacity, capital intensity, electricity supply and the concentration of the supply chain in a small number of facilities.

6. Synthetic biology and bio-manufacturing

Today. Engineered organisms produce enzymes, flavours, materials and pharmaceutical precursors at commercial scale. Design tools have improved sharply, including protein structure prediction and design.

Plausible by 2035. More materials, chemicals and food ingredients produced by fermentation where feedstock and energy costs allow; faster iteration in biological research.

What decides it. Scale-up economics — the gap between a working strain and a profitable plant — plus biosafety governance and public acceptance.

7. Autonomous mobility

Today. Driverless ride-hailing operates commercially in a small number of mapped urban areas with defined operating conditions. Highway driver-assistance is widespread and frequently overstated in marketing.

Plausible by 2035. Expansion city by city, with freight corridors and fixed routes leading. Full unsupervised autonomy everywhere, in all weather, is not a safe planning assumption.

What decides it. Per-mile safety validation, insurance, municipal politics and the cost of remote support operations.

8. Space infrastructure

Today. Reusable launch has reduced cost per kilogram to low Earth orbit; large constellations provide broadband and imaging; Earth observation quietly supports agriculture, insurance and climate monitoring. Public lunar programmes are active, including NASA's Artemis campaign and ESA's exploration work.

Plausible by 2035. Cheaper and more frequent launch, richer Earth observation data as an input to ordinary industries, early in-space servicing, and crewed lunar activity in campaign form.

What decides it. Launch cadence, orbital congestion and debris management, and whether demand exists beyond communications and imaging.

9. Human–machine interaction

Today. Voice and text interfaces have improved markedly. Lightweight augmented-reality devices remain niche. Invasive brain–computer interfaces are in small clinical trials with genuine but narrow results for people with severe motor impairment.

Plausible by 2035. Conversation and intent become a normal interface layer over software; ambient capture and summarisation spread through workplaces; assistive neurotechnology advances within medical settings.

What decides it. Battery and optics engineering for wearables; for neural interfaces, surgical risk, long-term implant stability and regulatory evidence. Consumer brain-computer interfaces by 2035 are speculative.

10. Materials, sensing and the quiet layer

The least discussed technologies are often the most consequential. Cheap sensors, better batteries at small scale, improved separation membranes, higher-efficiency power electronics and inexpensive precise positioning enable everything above. A decade of incremental gains in this quiet layer will probably change everyday life more reliably than any single headline breakthrough.

Counterarguments worth taking seriously

The productivity sceptic's case. Powerful capability can coexist with weak measured productivity for a long time. Complementary investment, reorganisation and skills take years, and some AI deployment simply shifts work rather than removing it.

The constraint case. Energy, minerals, skilled labour and fabrication capacity are all finite in the medium term. Several of the scenarios above compete for the same inputs.

The trust case. Synthetic media, opaque automated decisions and a few high-profile failures could slow adoption in exactly the domains where the benefits would be largest — healthcare, public services, finance.

The institutional case. Regulation is not only a brake. Clear liability rules and credible standards often accelerate adoption by making risk priceable. Ambiguity, not strictness, is what stalls deployment.

What would change my mind

Useful indicators to watch, each observable rather than rhetorical:

• Agent reliability on long-horizon tasks measured by independent evaluations rather than vendor demonstrations.
• Robot fleet hours in unstructured environments, and maintenance cost per hour.
• Grid connection queue times and transmission kilometres actually energised.
• Regulatory approvals of AI-supported diagnostics, and evidence of outcome improvement rather than accuracy alone.
• Launch cadence and the first commercially profitable non-communications space service.
• Total electricity demand from data centres versus efficiency gains per unit of computation.

Conclusion

The plausible 2035 is not a single scene. It is a set of overlapping, uneven transitions: capable but supervised software agents; robots that are excellent in some settings and absent from others; medicine that becomes more anticipatory in wealthy systems first; an electricity system rebuilt around variable generation and storage; and space infrastructure so ordinary it becomes invisible.

The point of laying it out this way is not to be right about 2035. It is to be prepared for several versions of it — which is the entire purpose of thinking carefully about the future without pretending to predict it.