KPMG: 2026 Frontiers Report - AI as the Frontier Multiplier
Source: KPMG, 2026 Frontiers Report: AI as the Frontier Multiplier, July 2026
Key takeaways
AI is becoming a structural layer of the economy rather than a standalone technology sector. KPMG argues that improvements in AI increasingly accelerate progress in computing, physical systems, infrastructure, biological production and other domains.
The frontiers are interconnected. Advances in reasoning affect robotics. Computing architecture affects AI scale. AI infrastructure changes energy demand. Better sensing changes physical autonomy. Progress in one domain increasingly changes the economics and trajectory of another.
Competitive advantage shifts towards orchestration. As intelligence becomes more widely available, the differentiator is increasingly how effectively enterprises direct it, combine it with context, govern it and connect it to existing systems and workflows.
Scalable intelligence changes a fundamental constraint of the enterprise. Organisations have historically expanded cognitive capacity by hiring people and developing expertise. As reasoning becomes increasingly scalable, the constraint moves towards defining problems, supplying context, coordinating work and turning intelligence into outcomes.
Context becomes more scarce as intelligence becomes more abundant. KPMG argues that proprietary knowledge, organisational memory, customer relationships, judgement, trusted principles and accountability become more valuable as access to powerful models becomes widespread.
Governance needs to move into the flow of work. Periodic approvals and retrospective controls are poorly matched to increasingly autonomous systems. Identity, provenance, verification, observability and continuous assurance become part of enterprise infrastructure.
AI identity is emerging as a critical control plane. Non human agents increasingly participate in enterprise workflows, but identity systems and governance models have not necessarily evolved at the same pace.
Physical AI takes intelligence beyond the screen. AI increasingly acts through robots, autonomous systems, sensors and industrial equipment, shifting the problem from digital productivity towards physical execution, safety, liability and operational resilience.
Energy and compute are becoming binding constraints on AI scale. Electricity, grids, chips, memory, cooling, land and community acceptance increasingly influence where AI infrastructure can be built and how quickly it can scale.
Compute itself becomes a supply chain issue. Enterprises increasingly need to think not only about power availability but also about whether they can secure enough inference capacity when business processes depend on AI continuously.
Quantum is moving from a distant technology issue into enterprise preparedness. Post quantum cryptography, quantum sensing, hybrid architecture and targeted quantum computing use cases increasingly influence security and technology planning today.
The future computing estate will likely be heterogeneous. Quantum, photonic, neuromorphic, processing in memory and other specialised architectures are developing alongside CPUs and GPUs. The strategic capability becomes matching workloads to the right architecture while preserving flexibility.
Trust becomes an architectural requirement. In a world of autonomous agents and synthetic media, organisations can no longer rely on assumed identity, assumed provenance or statements made by a system. Verification increasingly needs to be built into how systems operate.
Space is becoming part of enterprise infrastructure. Connectivity, navigation, timing, sensing and geospatial intelligence already depend on space based systems, while AI is increasing the value of satellite data and beginning to influence how compute itself could be distributed.
Leaders should treat the six frontiers as one strategic landscape. KPMG’s final argument is that these are not independent trends. The organisations best positioned for the next decade will understand how they interact and make integrated choices across strategy, infrastructure, risk, operating models and investment.
Executive summary
KPMG’s 2026 Frontiers Report examines how several major technology and infrastructure shifts are beginning to interact, with artificial intelligence acting as what the report calls the frontier multiplier.
The report builds on KPMG’s 2025 focus on convergence. The difference in 2026 is that convergence is increasingly being shaped by intelligence embedded across enterprise operations, infrastructure and markets. Advances in reasoning, planning, multimodal understanding and autonomy reinforce one another. Capabilities that were previously experimental are entering production environments, while constraints such as energy, compute, regulation and infrastructure are becoming central to strategy.
KPMG organises this landscape around six frontiers: Path to AGI, Physical AI, Energy and Compute, Quantum and NextGen Computing, Future of Trust and Space. The report also highlights emerging areas such as the bioeconomy and what it calls the Perception Layer.
The central proposition is that AI no longer operates as an isolated sector. Improvements in AI affect robotics and automation. Computing architectures determine which models can scale economically. AI demand changes energy infrastructure. Sensing creates the real world data required for physical autonomy. Verification and identity become more important as AI agents act independently. Space provides connectivity, sensing and potentially new infrastructure for computation.
The first frontier, Path to AGI, is less concerned with predicting a formal arrival date for artificial general intelligence than with what happens as intelligence becomes progressively more general and scalable. KPMG argues that enterprises will feel the consequences before anyone agrees that AGI has arrived.
Reasoning, planning, memory, learning, tool use and increasingly autonomous execution are already converging. As this happens, cognitive capacity becomes less tightly constrained by the number of people an organisation can employ. The strategic question moves from how many people are available to perform cognitive work towards how much reasoning capacity the organisation needs, where it should be deployed and how it should be governed.
That shift increases rather than eliminates the importance of human capability. KPMG argues that human intelligence increasingly defines objectives, values, trade offs and accountability. Intelligence can become abundant while judgement, institutional context and trusted relationships remain scarce.
Physical AI pushes the same transition into the real economy. AI increasingly moves from generating recommendations towards controlling machines that move products, inspect assets, operate equipment and navigate dynamic environments. This raises a different class of enterprise questions because failures can affect safety, equipment, operations and people.
The Energy and Compute frontier introduces another constraint. The ability to scale AI increasingly depends on electricity, grid capacity, data centre infrastructure, cooling, chips, high bandwidth memory, land and permitting. KPMG argues that energy and compute are moving from enabling inputs towards strategic constraints.
Quantum and NextGen Computing extend the infrastructure question further. The future of computing is likely to involve multiple specialised architectures rather than one dominant system. Quantum matters not only because of future computing performance but because existing encryption may need to be replaced years before quantum systems are capable of breaking it.
The Future of Trust frontier addresses what happens when autonomous agents and synthetic media make digital interactions harder to verify. Identity, provenance, cryptographic verification and continuous assurance increasingly become embedded into system architecture rather than applied later through governance.
Finally, KPMG argues that space is becoming another layer of enterprise infrastructure. Most companies may never own satellites, but many already rely indirectly on satellite communications, GPS, timing or geospatial intelligence. AI makes those data streams more valuable and increases the strategic importance of the infrastructure supporting them.
The overall message is that leaders should not manage these developments as six separate technology trends. They increasingly reinforce one another. AI drives compute demand. Compute drives energy demand. New computing architectures respond to energy and workload constraints. Physical AI relies on sensing, edge compute and trust. Space provides sensing and connectivity. The resulting advantage increasingly comes from orchestration across the entire landscape.
Key stats and quotable claims
KPMG cites research suggesting that the length of software tasks frontier agents can complete autonomously has doubled roughly every seven months over the past six years.
KPMG’s Q2 2026 Pulse Survey found the share of organisations orchestrating multiple agents across workflows rose from 9% to 18% in one quarter.
An Okta survey cited in the report found 91% of organisations already using AI agents, while only 10% had a strategy for managing non human identities.
Microsoft reporting cited by KPMG said more than 80% of Fortune 500 companies were already using active AI agents.
Research cited from Cloud Security Alliance and Strata Identity found only 18% of organisations highly confident their existing identity and access management systems could manage agent identities.
KPMG’s 2025 CEO Outlook found 86% of CEOs expect AI agents to operate as embedded team members within the next year.
Venture funding in robotics and Physical AI exceeded $16 billion in 2025.
Approximately 4.7 million industrial robots were operational in 2024.
82% of manufacturers cited a lack of AI ready skills as their top challenge in one manufacturing study referenced by KPMG.
More than one million robots are operating within Amazon facilities.
Global industrial robot installations approached 575,000 in 2025 and are projected to exceed 700,000 annually by 2028.
Global edge AI is projected to reach approximately $386 billion by 2034.
The digital twin market is expected to exceed $140 billion by 2031.
The International Energy Agency projects global data centre electricity demand could roughly double by 2030 to around 945 TWh.
KPMG cites estimates that major technology platforms could spend roughly $600 billion in 2026, around 70% higher than 2025, with a large share directed towards AI compute and next generation data centres.
Forty three publicly traded utilities project approximately $1.1 trillion of capital investment from 2025 to 2029.
KPMG introduces the idea of a “Token Brownout”, where an enterprise may be unable to secure sufficient AI inference capacity for critical workloads even when it has not suffered a traditional computing outage.
Q CTRL and IBM reported a commercially relevant materials science simulation running in approximately two minutes instead of more than 100 hours, around a 3,000 times speed improvement.
Cleveland Clinic, RIKEN and IBM reported a quantum assisted simulation of a 12,635 atom protein.
Deepfake fraud attempts in contact centres reportedly increased by more than 1,300% in 2024 in research cited by KPMG.
The space economy is projected in the report to grow from approximately $630 billion in 2023 to more than $1.8 trillion by 2035.
Bio based plastics currently represent only around 0.5% of global plastics production, biofuels less than 3% of transport fuel and bio based chemicals an estimated 2%, illustrating the gap between biological technology readiness and commercial scale.
Overall summary
The cleanest way to understand the report is that KPMG believes AI is changing from a discrete technology investment into a force that changes the economics and constraints of several adjacent systems at once.
The previous enterprise AI conversation largely focused on model access. Which model should the organisation deploy? Which use cases should it automate? How can employees become more productive?
KPMG pushes the analysis one level further. If reasoning itself becomes increasingly scalable, what becomes scarce instead?
Its answer is context, judgement, trust, governance, infrastructure, energy and the ability to coordinate complex systems.
This is most obvious in the Path to AGI chapter. KPMG is not making the report dependent on a prediction that AGI arrives in a particular year. It argues that the enterprise implications emerge progressively as reasoning, planning, memory, learning and autonomous execution become more capable.
The organisation therefore begins to treat intelligence as another scalable resource. Workforce planning starts to overlap with cognitive capacity planning. Institutional memory becomes infrastructure. Knowledge becomes dynamic context rather than static documentation. Governance becomes continuous.
Physical AI then converts intelligence into action. The risk profile changes because a poor model answer can become a poor machine action. The operating model therefore has to incorporate safety, monitoring, exception management and physical accountability.
Energy and Compute reveals the infrastructure underneath that intelligence. AI cannot scale independently of electricity, transmission, cooling, chips, memory, land and permits. The resulting competition for compute introduces new forms of operational dependency.
Quantum and NextGen Computing show how the computing estate itself is fragmenting into specialised architectures. Instead of assuming every workload sits on one standard infrastructure layer, enterprises increasingly need portfolio and orchestration capabilities across multiple computing technologies.
Future of Trust addresses a similar fragmentation problem in digital interactions. As agents, synthetic content and automated transactions scale, enterprises increasingly need independent evidence of who acted, what data was used, where information originated and whether the system behaved as intended.
Space extends the same argument beyond terrestrial infrastructure. Satellite systems already underpin navigation, communication and sensing. Their growing importance creates both business opportunity and dependency.
Across all of these frontiers, KPMG repeatedly arrives at the same managerial problem: access is not the differentiator. The ability to integrate, govern and orchestrate increasingly interconnected capabilities is.
Deep dive
AI as the frontier multiplier
AI now plays a structural role across the domains shaping the global economy.
AI is becoming embedded within computation, infrastructure, industrial production, biological manufacturing and geopolitical competition. As capabilities improve, progress in one frontier increasingly accelerates developments in others, creating a multiplier effect.
Improvements in reasoning models influence robotics and automation. Advances in computing architectures expand deployment scale. Growth in data centre capacity reshapes energy demand and infrastructure planning. Biological design increasingly depends on computational discovery tools. The pattern is also visible across industrial environments, where advances in machine perception, planning and control enable intelligent automation.
AI functions less as a standalone sector and more as a connective capability embedded within the economy. Understanding these interactions is critical for strategic decision making, where investments within one frontier shape opportunities in another.
KPMG reduces the argument to four points: AI is multiplying across systems, not replacing them. Energy and governance are the real multipliers or constraints. The frontiers are interconnected, not independent. Competitive advantage shifts to orchestration.
What changed from 2025
The 2025 Frontiers Report focused on convergence.
Last year, the challenge was understanding convergence. This year, the challenge is understanding how that convergence is being shaped by intelligence embedded across enterprise operations, infrastructure and markets.
As volatility compounds across the economic, political and technological landscapes, organisations are striving to adapt faster, make decisions sooner and execute change more effectively.
KPMG identifies three changes.
Capabilities are compounding. Advances in reasoning, planning, multimodal understanding and autonomy are no longer progressing in isolation. They are reinforcing one another, accelerating timelines and compressing development cycles.
Execution is moving closer to the real world. Capabilities previously confined to advisory or experimental settings are now operating within production environments. Decisions increasingly translate directly into action across infrastructure, workflows and physical systems.
Constraints are becoming central. Energy availability, infrastructure readiness, compute capacity, regulation and geopolitical exposure are no longer secondary considerations. They are shaping how and where systems can scale.
Path to AGI
When intelligence stops being scarce
Artificial General Intelligence is usually framed as a race against the calendar. Will it arrive in five years, ten years or longer?
KPMG takes a different approach. For two centuries, every institution we have built has assumed that thinking is expensive. Organisations acquire cognitive capacity by hiring people, training them for years, buying firms that already have it or waiting for experience to accumulate.
Cognitive capacity has been one of the hardest resources to scale and the slowest to move. Org charts, budgets, governance, career ladders and the shape of the modern firm are downstream of that constraint. Remove it, even partially, and the assumptions built on top of it start to give way.
AGI is often used to describe systems that can reason, learn, adapt, transfer knowledge and operate effectively across a wide range of unfamiliar domains without task specific retraining. But for enterprise leaders, KPMG argues that the exact definition matters less than the direction of travel.
Frontier AI systems are becoming more capable, more general and more useful across increasingly complex work.
Enterprises do not need to wait for a formal declaration of AGI to feel the consequences. Reasoning, planning, memory, tool use, learning, machine teaching and autonomous long horizon execution are improving quickly and converging into systems that can take on more ambiguous, cross functional work.
The significance is not whether AGI arrives on a precise date or meets a perfect definition. Intelligence is moving from a scarce organisational resource towards something that can increasingly be scaled, aimed and governed as an economic unit.
The economics of scalable intelligence
Organizations have traditionally expanded cognitive capacity by hiring people, developing expertise, bringing in specialist firms or investing in training.
Intelligence has therefore been expensive, unevenly distributed and difficult to scale.
As intelligence becomes more scalable, that equation starts to change. Organizations are no longer limited only by how many experts they employ.
The limiting factor becomes how effectively they define problems, provide context, coordinate work, reimagine the work itself, govern decisions and convert reasoning into outcomes.
This does not diminish the importance of human expertise. Quite the opposite.
Human intelligence increasingly defines objectives, exercises judgment, establishes values, manages trade offs and evaluates outcomes. Scalable intelligence expands the organisation’s ability to reason, explore alternatives, synthesise knowledge and learn across the enterprise.
The economic implications therefore extend beyond productivity. Organisations may rethink how they innovate, allocate capital, organise work, preserve institutional knowledge and compete.
Just as cloud computing changed assumptions about infrastructure, scalable intelligence may change assumptions about how cognitive work is created, distributed and managed.
KPMG’s Q2 2026 Pulse Survey provides an early signal: the share of organisations orchestrating multiple agents across workflows doubled from 9% to 18% from the previous quarter.
The binding constraint stops being whether the organisation has enough cognitive capacity and becomes how well it aims it.
Capabilities that change the enterprise
The significance of AGI is not that it introduces entirely new capabilities. Early versions of many of these capabilities are already visible in today’s frontier models and are converging.
What changes is their breadth, depth, reliability, adaptability and applicability.
The first capability is generalisation. Can a system reason through a problem it has not been explicitly trained or taught to solve? Many consequential enterprise problems have little precedent. The value is not simply retrieving the right historical example but reasoning when precedent is limited.
The second is cross domain reasoning. A more general system can combine knowledge that traditionally sits in separate disciplines. For a global bank, that could mean reasoning simultaneously across risk, tax, treasury, compliance, legal, operations, technology and product strategy. The value is not faster analysis inside one function. It is seeing relationships and trade offs that are difficult to identify from one discipline alone.
The third is transferable learning. Knowledge and insights learned in one context can improve performance somewhere else. Patterns around flow, capacity and variability in manufacturing may inform hospital operations. Financial risk techniques may help supply chain resilience because the underlying problems share uncertainty and interdependence.
This is materially different from automation. Automation repeats a known process. Transferable learning allows insight to move across the enterprise and compound.
The fourth is long horizon reasoning. Rather than optimising isolated activities, more capable systems could evaluate choices over months or years and continually revise recommendations as geopolitical, regulatory, demographic, technological or market conditions change.
The fifth is adaptive intelligence. Systems may increasingly learn, reason about their own reasoning and update their understanding of complex environments as new evidence arrives.
The sixth is invention and discovery. A system that can generalise, reason across disciplines, transfer knowledge and adapt its understanding can generate hypotheses, designs, products, molecules or operating models that may not emerge from one team or function.
Collectively, these capabilities change the role intelligence plays inside the enterprise.
Instead of helping people perform individual tasks, intelligent systems increasingly participate in how organisations learn, adapt, discover, plan and make decisions.
As these capabilities mature, intelligence itself becomes a more scalable resource, reshaping how enterprises create value, manage risk and compete.
Workforce planning becomes cognitive capacity planning
The precise moment at which a system qualifies as AGI is less important than the direction of capability improvement.
Organizations do not need to wait for a formal AGI milestone to experience the consequences. As AI systems become more capable at reasoning, learning, planning, discovery and adaptation, enterprises will need to rethink assumptions built around intelligence being scarce, expensive and difficult to scale.
Workforce planning has traditionally determined how much cognitive work an organization could perform.
As intelligence becomes increasingly scalable, leaders may need to think about cognitive capacity alongside financial capital, physical assets and human talent.
This is particularly important because organisations have historically built expertise through apprenticeship models. If scalable intelligence begins to automate a greater share of entry level cognitive work, the traditional talent pipeline may also need to be redesigned.
The strategic question shifts from “How many people do we need?” towards “How much reasoning capacity do we need, where should it be applied and how should it be governed?”
Knowledge becomes more than documentation
Organizations have always depended on more than what is written down.
Tacit expertise, informal networks, accumulated judgment, customer nuance, project history and context that exists only in people’s heads often matter as much as formal documentation.
As AI systems become more capable, enterprises will increasingly find ways to preserve, externalize and reason over both explicit and tacit knowledge.
Documents, policies, engineering standards, contracts and research will remain important, but they will no longer exist only to be read.
They become part of the living context that informs decisions as they are made.
Institutional memory becomes infrastructure
Most organizations lose valuable context every day.
Decisions disappear into meetings. Rationale gets buried in emails. Teams repeat work because prior learning is hard to find. Knowledge leaves when people leave.
In a world of scalable intelligence, that becomes a structural disadvantage.
Living memory systems that preserve decisions, assumptions, relationships, evidence, lessons learned and organizational knowledge may become as important as today’s data platforms.
Organizations that remember better will reason better.
Institutional memory therefore changes from something the organisation hopes to retain into a potentially strategic infrastructure layer.
Technical debt becomes a strategic constraint
As intelligence becomes more scalable, technical debt becomes more visible and more expensive.
Legacy systems, fragmented data, brittle integrations, undocumented business rules and ageing controls increasingly limit an organisation’s ability to deploy AI safely and at scale.
More capable AI systems may also expose vulnerabilities, control gaps and process weaknesses that were previously masked by human effort.
The competitive advantage will increasingly belong to organisations whose technology, data and trust environments are ready for scalable intelligence, rather than those forced to navigate around legacy constraints.
The work surface itself changes
Email, presentations, meetings, task lists and chat are unlikely to disappear.
But KPMG argues that they may increasingly be complemented by persistent cognitive workspaces where people and intelligent systems explore ideas, evaluate alternatives, retain context and build on previous work rather than recreating it.
This is not simply a better collaboration tool.
It is a different surface for enterprise work, one where human and machine intelligence operate against shared context over time.
Governance becomes continuous
Existing governance models were designed around human decision velocity. They rely on periodic reviews, approvals, audits and controls.
As increasingly capable systems participate in more consequential decisions, governance must operate inside the flow of work.
Policies, guardrails, verification and assurance become embedded within operational processes rather than applied after the fact.
Identity, provenance and observability become foundational.
Organizations need confidence in where knowledge came from, how conclusions were reached, what influenced a recommendation and whether systems remained aligned with enterprise objectives throughout the reasoning process.
Identity becomes the new control plane
Because an agent’s own claim that compliance checks have occurred cannot simply be accepted at face value, identity becomes a foundational layer for governance.
KPMG cites an Okta survey in which 91% of organisations reported already using AI agents while only 10% had a strategy for managing non human identities.
Microsoft reporting cited in the report said more than 80% of Fortune 500 companies were using active AI agents, yet Cloud Security Alliance and Strata Identity research found only 18% of organisations highly confident that their current identity and access management systems could manage agent identities.
The issue therefore extends beyond giving employees access to AI tools.
Enterprises increasingly need to know which human or machine actor performed an action, under whose authority, using what permissions and with what evidence.
Cognitive security becomes a board level issue
Protecting information is no longer sufficient.
Enterprises will also need to protect the integrity of objectives, reasoning, memory, accumulated knowledge and decision pathways.
The challenge exists at several levels.
Operationally, shadow agents may proliferate. Legally, liability for autonomous decisions remains unresolved in many areas. At a broader systemic level, concentration in a small number of AI providers can create cascading risk.
As AI moves from information processing into decision making, the security perimeter increasingly includes what systems believe, remember and are trying to accomplish.
Context becomes the moat
Access to powerful models is unlikely to remain exclusive.
But enterprise context will be harder to copy.
Customer relationships, operating experience, institutional knowledge, decisions, proprietary data, organizational memory, trusted principles, judgment and accountability become more important as intelligence becomes more widely available.
KPMG captures the shift simply:
Intelligence may become abundant. Context does not.
As access to capable models becomes widespread, advantage shifts towards trusted relationships, institutional knowledge, judgment, accountability and the operating context in which decisions are made.
The leadership challenge
None of these shifts require the arrival of a universally accepted definition of AGI.
They emerge as AI systems become increasingly capable and increasingly embedded in how organizations create knowledge, make decisions, preserve context, allocate capital and manage risk.
The strategic challenge for leaders is therefore not predicting exactly when AGI arrives.
It is determining how quickly these capabilities are improving, where they create competitive advantage and how the organisation will adapt as intelligence becomes increasingly scalable.
KPMG summarises the broader transition as follows:
Enterprise today: workforce determines cognitive capacity. Enterprise with scalable intelligence: cognitive capacity becomes elastic.
Knowledge is captured in documents. Knowledge becomes continuously accessible and extensible.
Governance is periodic. Governance becomes continuous.
Institutional memory decays. Institutional memory compounds.
Expertise is scarce. Context is scarce.
Intelligence is expensive. Judgment is expensive.
When intelligence stops being scarce
Each technological era changes what is abundant and, in doing so, changes what is precious.
Industrialization made physical labor more scalable and elevated design, capital and management. Computing made calculation abundant and elevated software. The internet made information access abundant and elevated networks, platforms and attention.
As intelligence becomes more scalable, the prize moves to what abundance cannot easily touch: judgment, trust, context, governance, accountability, leadership and human relationships.
The organizations that succeed will not necessarily be those with access to the most capable models.
They will be those that know how to combine scalable intelligence with deep institutional knowledge, disciplined governance, resilient cognitive security and clear human accountability.
The future of AGI remains uncertain. Its timing, technical path and ultimate capabilities will continue to be debated.
But the strategic direction is already visible.
AI systems are becoming more capable, more general and more embedded in how organisations reason, learn, decide and act.
The question facing leaders is therefore not whether AGI has officially arrived. It is whether their organisations are prepared for a world in which intelligence becomes increasingly scalable and no longer serves as the constraint around which the enterprise was designed.
Physical AI
The AI race has left the screen
The biggest AI story is no longer only about the models running on screens. It is increasingly about intelligence acting in the physical world.
For enterprise leaders, the provocative question shifts from whether AI will reshape operations towards whether the operating model can survive when AI begins to act, not just advise.
When intelligence is embedded into machines that move products, inspect assets and navigate dynamic environments, the consequences become immediate.
Earlier waves of AI changed how people write, search, analyse and decide. Physical AI changes how work gets done in the world.
When intelligence is embedded into machines that move products, inspect assets, operate equipment and navigate dynamic environments, productivity can rise, but so can safety, liability and operational risk.
The transition therefore moves Physical AI from a technology issue into an operating model, governance and capital planning issue.
Why Physical AI is accelerating
Several forces are moving simultaneously.
Capital is returning to the sector. Venture funding in robotics and Physical AI exceeded $16 billion in 2025.
Operating pressure is also increasing. Labour shortages, scarce skills and supply chain disruption are encouraging companies to automate targeted tasks, particularly where reliability and continuity matter.
Open world foundation models, simulation tools and synthetic data are also making Physical AI easier to develop. Organisations do not necessarily need to create every model from scratch.
But a major bottleneck remains: high quality real world training data for robots is hard and expensive to collect, especially around edge cases and dynamic physical interaction.
Simulation and synthetic data help narrow that bottleneck without eliminating it.
Why Physical AI matters
Physical AI raises the stakes of AI adoption.
In software, mistakes can slow work or reduce quality. In physical systems, they can affect safety, uptime, asset performance, reputation and financial results.
This makes Physical AI a matter of operating model, governance and capital planning.
Competitive advantage can shift from basic process efficiency towards systemic autonomy.
Operational risk also needs to be reconsidered because a software failure can increasingly trigger a physical world consequence.
The near term market is broader than humanoid robots
Physical AI should not be conflated with fully general purpose humanoid robots.
The near term market is broader and more practical: autonomous mobile robots, industrial systems, warehouse automation, inspection platforms and domain specific embodied agents operating inside constrained environments.
Humanoid robots may become an important long term frontier, but KPMG expects much of the enterprise value over the next several years to come from narrower systems that perform reliably in defined tasks and settings.
Intelligent task orchestration
AI systems are moving beyond single purpose automation towards handling multiple actions within defined environments.
Progress is strongest where tasks, environments and failure modes can be bounded.
Models increasingly translate language and perception into real world execution.
This does not necessarily imply general purpose robotics. Instead, enterprises can achieve value where autonomy is sufficiently constrained that systems can be monitored, tested and trusted.
Edge based autonomy
Decision making is also moving towards the edge, placing computing power where it is needed.
The global edge AI market is projected to reach around $386 billion by 2034.
The important operating implication is latency.
A machine working in a warehouse, vehicle, factory or remote asset cannot always wait for a cloud round trip before making a decision.
Physical AI therefore brings model capability, local compute, sensing and operational reliability into the same architecture.
Software defined robotics
The breakthrough in Physical AI is increasingly in software rather than hardware alone.
New model architectures make it possible for machines to interpret broader instructions, understand context and translate perception into action.
Hardware remains important, but differentiation increasingly moves towards models, data, orchestration and continuous improvement through software updates rather than complete mechanical redesign.
Simulation first design
Companies are increasingly using virtual models to test physical systems before deployment.
The digital twin market is expected to surpass $140 billion by 2031.
Manufacturing, energy and infrastructure organisations can use digital replicas to optimise designs, predict maintenance requirements and test potential changes before committing capital in the real world.
The boundary between digital decision making and physical execution is therefore narrowing.
Human machine safety becomes an operating requirement
As AI evolves from generating insights to driving real world actions, the risks become more severe.
Failures can cause operational disruption, equipment damage or harm to people.
Yet adoption of Physical AI is accelerating faster than the safeguards required to ensure safety for workers and the public.
Without purpose built safety frameworks, organisations face a growing gap between capability and control.
Key considerations include safe human machine interaction, validating AI driven actions in dynamic environments and establishing accountability for autonomous decisions.
How enterprises are responding
Physical AI is shifting from a technology initiative to an operating model decision.
Organizations are rethinking value chains, governance and infrastructure.
Enterprises are moving from isolated pilots into real operating environments. Edge computing supports local execution. Digital twins validate workflow changes before physical rollout. Organisations are scaling deployment across multiple sites to build operational confidence.
Many near term deployments will be limited less by model capability than by integration, monitoring, safety and governance at scale.
Companies are therefore establishing central governance models, real time performance monitoring and standardised middleware and integration layers.
The issue moves from can the machine do the task? towards can the organisation repeatedly deploy and control the system across real operations?
Energy and Compute
The new constraint: power, chips and AI scale
For decades, companies could expand computing without worrying significantly about power.
Electricity was abundant, predictable and largely grew alongside digital demand.
That relationship is changing.
Computing power and the electricity required to support it are becoming tightly connected, and the availability of power increasingly shapes where and how quickly data centres and digital infrastructure can grow.
Training large AI models requires large numbers of powerful chips operating simultaneously. Once deployed, AI systems can also run continuously, increasing electricity requirements across data centres.
The International Energy Agency projects global data centre electricity demand could roughly double by 2030, reaching around 945 TWh in its base case.
Energy infrastructure expands much more slowly.
Power plants, grid connections and transmission lines require planning, permits, financing and construction. Technology cycles can be measured in months while energy infrastructure often takes years to expand.
Digital growth is therefore increasingly shaped by physical limits that cannot be scaled at software speed.
Electricity becomes a business risk
The rapid growth of AI driven computing is colliding with constrained power systems.
Electricity moves from a predictable operating cost towards a potential business risk.
Access to power, energy prices and infrastructure development timelines are becoming board level considerations, particularly for enterprises whose growth depends heavily on digital infrastructure.
AI scale also depends on specialised hardware.
KPMG points specifically to high bandwidth memory and the manufacturing capacity required to assemble deployable systems.
Energy and hardware may be the most visible constraints, but AI increasingly depends on an interconnected stack of resources.
From power outages to Token Brownouts
While electricity is the most visible constraint, KPMG identifies another emerging issue: resilience of access to compute itself.
As AI becomes embedded into core business processes, enterprises will require continuous access to the models that run those processes.
But as energy and hardware constraints intensify, the market for AI inference could become more volatile.
KPMG calls one possible outcome a Token Brownout.
A Token Brownout is a situation where an enterprise cannot access the AI inference capacity it requires when it needs it.
This may not be a total outage. It may appear as higher latency, lower priority or inability to secure enough capacity for critical workloads.
The strategic question therefore becomes not only whether the data centre has power but whether the enterprise has priority access to the AI compute its operations depend on.
As a response, enterprises may increasingly segment workloads: using cloud platforms for elasticity and frontier model access while deploying smaller or locally hosted models for predictable high volume inference.
The AI scaling stack
KPMG describes four layers of constraint that shape AI deployment.
At the base are power and grid infrastructure, the electricity systems required to run AI.
Another layer is silicon, including machines and foundries capable of producing AI chips.
Then come chips and memory, including the processors and high bandwidth memory on which AI workloads run.
Finally, there is land and community, including sites, water, permitting and local approval.
A failure or shortage at any layer can constrain the layers above it.
This is why AI infrastructure increasingly becomes an enterprise strategy and supply chain issue rather than simply a cloud procurement decision.
Power becomes strategic infrastructure
Capital flows are already reflecting the mismatch between AI demand and infrastructure.
KPMG cites estimates that major technology platforms could spend roughly $600 billion in 2026, up significantly from 2025, much of it directed towards AI compute and next generation data centres.
Utilities are also increasing investment, but their timelines are longer. Forty three publicly traded utilities project approximately $1.1 trillion in capital investment from 2025 to 2029.
Electricity prices have also risen.
The implication is that electricity is no longer merely a stable operating input. It increasingly acts as a structural boundary on enterprise growth, capital velocity and margin durability.
Constraints are changing architecture
The pressure is producing unusual strategic responses.
Companies are exploring space based compute, advanced cooling, neuromorphic and photonic processors, new forms of nuclear power and fusion.
Industry coalitions are forming between hyperscalers, data centre operators, utilities and grid operators.
Technology companies are also moving upstream into energy sourcing, including power purchase agreements, direct investment and co located generation.
The underlying change is important.
When power is only a utility purchase, expansion depends on external timelines and tariffs.
When organisations gain greater control over supply, energy starts to behave more like a strategic asset.
Treat AI inference as a supply chain input
KPMG’s action agenda argues that enterprises should build an energy baseline, integrate power into scenario planning, preserve flexibility, anticipate community and regulatory scrutiny and secure their compute supply chain.
One recommendation is particularly notable: organisations should map their dependency on external AI providers and inference capacity, diversify providers and explore contractual options that give critical workloads stronger priority or quality of service.
AI inference increasingly needs to be managed with the same resilience thinking applied to important physical inputs.
Quantum and NextGen Computing
Quantum moves into enterprise planning
For years, enterprise technology strategy followed a familiar sequence: the technology matured, standards emerged, and then regulators, investors and large companies moved at scale.
Quantum has not followed that sequence.
Institutional momentum arrived early, and in 2026 the technology began to catch up.
Quantum technologies are already delivering commercially relevant results in targeted areas even as fault tolerant systems continue to mature.
Enterprises are therefore no longer preparing only for a distant technology. They are responding to one that has begun to influence security, architecture and investment decisions.
KPMG identifies four areas leaders should understand: post quantum cryptography, quantum computing, quantum sensing and hybrid compute architecture.
Post quantum cryptography
Post quantum cryptography is the effort to replace or upgrade encryption before quantum computers can break some methods widely used today.
In practical terms, it is a security modernisation programme.
The issue matters now because sensitive encrypted information stolen today could potentially be decrypted later when quantum capability matures.
Encryption is also deeply embedded across networks, applications, devices and third party systems, which means migration can take years.
The preparedness window can therefore close before a cryptographically relevant quantum computer exists.
KPMG’s message is not that every organisation should make a large quantum computing investment today.
It is that enterprises with long lived sensitive information, regulated systems or complicated vendor estates need a quantum position now.
Quantum computing becomes commercially relevant in narrow domains
Quantum computing is not better at every form of computation.
Its potential advantage lies in specific classes of problems, including molecular simulation, optimisation and parts of complex financial modelling.
KPMG highlights a May 2026 result from Q CTRL and IBM in which a materials science simulation requiring more than 100 hours on conventional software ran in approximately two minutes on a quantum system, while delivering comparable accuracy.
It also highlights work from Cleveland Clinic, RIKEN and IBM simulating a 12,635 atom protein using quantum centric supercomputing.
The implication is not that enterprises should purchase quantum computers.
It is that quantum has crossed a threshold where targeted hybrid experiments can begin creating insight or advantage in selected domains.
Quantum sensing may arrive earlier
Quantum sensing uses extremely sensitive devices to detect small changes in motion, location, time, gravity or magnetic fields.
Better measurement can improve navigation, diagnostics and monitoring.
KPMG highlights an example from Boeing where a plane completed a flight test using quantum sensors to navigate without GPS for four hours.
The business implication is that some of the first meaningful quantum impacts may emerge from better sensing rather than from general purpose quantum computing.
Hybrid architecture is the likely model
Quantum is not expected to replace classical computing.
Instead, quantum processors, CPUs, GPUs, storage and orchestration software increasingly operate as parts of one technology environment.
KPMG argues that the enterprise question is therefore not whether to buy a quantum computer.
It is whether the organisation’s architecture is flexible enough to integrate quantum capability as useful domains emerge.
Architecture readiness comes before quantum performance.
Delay compresses the transition
The risk of waiting is not necessarily that a competitor suddenly acquires a fully mature quantum computer.
The larger risk is losing the time required to prepare for changes already emerging in security, vendor requirements, architecture and workforce planning.
Delay does not eliminate the work. It compresses it.
Security exposure accumulates because sensitive data can be collected now and targeted later.
Standards or mandates can convert a long transition into a rushed remediation programme.
Vendor and architecture decisions become harder to unwind.
Capability gaps grow as early movers gain experience through pilots, partnerships and reskilling.
Strategic flexibility gradually erodes.
The goal is therefore not to make huge bets too early. It is to build enough understanding and flexibility to respond deliberately later.
NextGen Computing
The one size fits all model begins to break
Quantum is only one specialised architecture among many.
The same forces driving quantum readiness are reshaping the wider computing stack.
Workloads are outgrowing general purpose hardware. Energy and memory are becoming constraints. Interoperability standards are developing. Specialised architectures are advancing in parallel.
Quantum may receive the most attention, but the more immediate shift for many enterprises could come from neuromorphic, photonic, processing in memory and other architectures converging into a heterogeneous computing portfolio.
For decades, enterprise computing relied on the assumption that processing power would continually improve and that organisations could rely on relatively standard systems.
Those assumptions are weakening.
Modern AI, simulation, optimisation and real time workloads have very different performance profiles.
The strategic question is therefore not whether one platform will dominate. It is how to place workloads across a more diverse environment without creating unnecessary complexity.
Hardware alone is no longer enough
For decades, chip improvements drove computing gains relatively directly.
Today, faster hardware alone does not necessarily produce equivalent economic value.
Software, talent, architecture and organisational change need to evolve alongside hardware.
Waiting for a faster chip is therefore not an enterprise computing strategy.
Infrastructure is becoming modular
Infrastructure has historically been delivered as tightly integrated packages.
That simplified purchasing and deployment but created long term lock in.
Emerging standards increasingly allow compute, storage, memory and acceleration technologies to be combined more flexibly.
This enables organisations to change hardware providers, run different types of computing side by side and introduce technologies incrementally rather than through complete replacement.
Without flexible infrastructure, heterogeneous computing becomes difficult to manage at enterprise scale.
Neuromorphic computing
Neuromorphic computing mimics characteristics of neural systems to recognise patterns using very little power.
The potential enterprise opportunity is true edge learning with extreme energy efficiency.
Instead of simply running a pre trained model, a factory robot or autonomous system could potentially learn from its environment locally without constantly transmitting information to the cloud.
This creates possibilities for long duration autonomous systems operating where connectivity or energy is constrained.
Photonic computing
Photonic systems use light rather than electricity to move and process certain forms of data.
The opportunity is particularly relevant to AI because moving information between processors and memory is becoming an important power and performance bottleneck.
Photonics therefore represents a potential path towards larger AI systems without corresponding increases in energy and data transfer costs.
Processing in memory
Processing in memory moves computation inside or closer to memory itself, reducing the need to continuously transfer data between memory and processors.
KPMG argues that this could significantly reduce cost and latency for AI inference.
If sufficiently mature, intelligence could become cheap enough to embed much more widely into devices, industrial systems and sensors.
The architectural implication is potentially profound: AI could move from a centralised expensive resource towards a more pervasive utility.
A portfolio and orchestration challenge
Taken together, next generation architectures represent a break from the one size fits all model.
Over time, enterprise computing is likely to evolve into a heterogeneous portfolio where workloads are strategically matched to the architecture best suited to them.
For leaders, the question is not which single architecture wins.
It is how to create an operating model that can match workloads to architecture, capture cost and performance advantages and preserve flexibility as the landscape evolves.
KPMG therefore frames next generation computing as less of a procurement exercise and more of a portfolio and orchestration challenge.
Future of Trust
Architecting trust in a zero trust world
We are entering a world in which enterprises can no longer assume that what they see is real or that what a system tells them is true.
KPMG argues that two forces are driving this environment: increasingly realistic synthetic media and the growing autonomy of AI agents.
Consider an AI system that reports that it received the appropriate approval for a transaction.
In a zero trust environment, that statement itself is not proof.
Leaders need independent, verifiable evidence that the approval actually occurred.
The same issue applies to digital media, where inability to distinguish authentic from synthetic content creates risk for decisions, reputation and fraud prevention.
The fundamental enterprise question therefore becomes: in a world where you potentially cannot trust anything by default, how do you establish enough trust to transact, decide and operate?
KPMG argues that the answer cannot come from traditional governance overlays alone.
Trust needs to become a core architectural requirement of the enterprise technology stack.
Why trust becomes a board issue
Four forces are converging.
The first is automation. Agentic AI introduces new artificial actors across workflows.
The second is synthetic reality. Generative AI has dramatically expanded the availability of highly realistic synthetic content. KPMG cites data showing deepfake fraud attempts in contact centres increasing by more than 1,300% during 2024.
The third is the distributed ecosystem. Enterprises operate across multi cloud environments, supply chains, platforms and partners where workflows depend on shared infrastructure.
The fourth is regulation and risk. Governance frameworks increasingly require organisations to demonstrate how systems are controlled and evaluated.
These are not entirely new architectural questions. Identity, provenance and verification were already important with IoT.
The difference is that autonomous AI systems can make decisions and generate content, multiplying the verification burden.
Trust failures become systemic
Trust failures can move from isolated incidents into systemic risks across entire value chains.
Financial services need reliable settlement.
Healthcare and life sciences require traceability.
Industrial autonomy requires demonstrable safety and accountability.
The core requirement is reliable coordination across complex networks of partners, platforms and infrastructure.
Organizations that can verify identity, data and system behaviour gain three advantages: reduced systemic risk, lower coordination friction and the ability to operate new trust dependent business models.
From institutional assurance to technical verification
Verification is moving from periodic checks towards continuous validation integrated directly into system operations.
As more activity is executed by software, platforms and autonomous systems, verification increasingly has to occur within the flow of operations.
KPMG identifies four structural shifts.
Identity becomes the primary security boundary. Access and activity are verified through identity rather than network location. Identity extends beyond employees to devices, services and AI agents.
Data and content provenance become important. Systems increasingly track origin and lineage to determine whether information can be trusted.
System assurance becomes continuous. AI models and automated systems require ongoing evaluation of reliability, performance and accountability.
Transactions increasingly contain verification. Cryptographic methods validate actions at the point of execution rather than relying only on reconciliation afterwards.
As trust becomes part of system architecture, control over how identity, transactions and interactions are defined can itself become strategic.
From governance overlay to architectural requirement
For years, trust has often been treated as a governance function: a compliance checklist applied after the system is built.
KPMG argues that this model is shifting.
In a zero trust world, trust must be engineered into systems from the start through verifiable identity, data provenance and cryptographic assurance.
External assurance will continue to matter, particularly in regulated environments.
But as systems operate at greater speed and scale, more verification must happen inside the system itself.
Three domains of technical trust
The first domain is Identity and Authenticity: who and what is real?
Capabilities need to establish which human, machine or service is participating in an interaction and whether the information itself can be trusted.
The second is Cryptographic and Transactional Trust: verifying value in motion.
The objective is to validate transactions and digital assets inside the transaction itself instead of depending entirely on post transaction reconciliation or intermediaries.
The third is System Integrity and Accountability: ensuring systems behave as intended.
Formal AI assurance practices, structured evaluation, adversarial testing and new cryptographic standards increasingly support the ability to operate automated systems at scale.
Trust can therefore no longer sit inside a single function.
It requires coordination across technology, security, data and business leadership.
A roadmap for trust
In the immediate term, KPMG recommends extending identity across employees, systems, devices and AI agents, particularly in high risk workflows.
Organisations should track provenance and content lineage, implement model evaluation and monitoring and identify where cryptographic exposure is highest.
The next stage is to embed verification directly into transactions and core workflows, operate continuous assurance and extend common standards to important partners and platforms.
Longer term, identity, verification and assurance increasingly become common infrastructure services while autonomous systems and shared data environments rely on validation happening automatically inside workflows.
The transition is from fragmented controls towards integrated verification across systems, transactions and ecosystems.
Space
Space becomes enterprise infrastructure
The technology may still be emerging, but the strategic implications are arriving faster than many leadership agendas.
What is changing is not simply launch economics or new space ventures. It is the growing role of space as a foundational layer of enterprise infrastructure, supporting connectivity, timing, sensing, intelligence and resilience.
Space based capabilities already support navigation, communications, supply chains, financial transactions and environmental monitoring.
Most organisations therefore already depend on space infrastructure in some form, even when that dependence is indirect or poorly understood.
The question for leaders is not whether the organisation participates in the space economy.
It is where space intersects with the business.
AI multiplies the value of space
AI accelerates the value of space in two directions.
On Earth, AI can convert satellite data into operational intelligence, compressing analysis that previously took much longer.
In orbit, AI’s increasing demand for compute and energy is encouraging exploration of processing information closer to where it is generated.
The result is a feedback loop.
AI makes space data more valuable while space increasingly becomes part of the infrastructure supporting AI.
Translate space into P&L impact
KPMG argues that the problem is not simply awareness of space but establishing a clear line of sight to enterprise performance.
It proposes four components.
Revenue growth: create new products or premium services using satellite or geospatial data.
Margin expansion: use satellite information to improve operational decisions, predict disruption and optimise resources.
Capital efficiency: consume space capabilities through subscriptions, APIs and commercial providers rather than building proprietary aerospace infrastructure.
Risk mitigation: understand how failures in GPS, satellite communications or Earth observation could disrupt core operations.
This framing converts space from a future technology topic into a measurable enterprise question.
Operating model and governance
A common barrier is ownership.
Who inside a normal enterprise owns space strategy?
KPMG argues that leading organisations do not need large independent space departments. Instead, space can operate as a horizontal capability.
A C suite leader such as a CIO, CTO or Chief Strategy Officer can act as executive sponsor.
A small cross functional centre of excellence can scan providers, run pilots and translate business needs into technology requirements.
Business units then integrate relevant space derived capabilities directly into their workflows.
Formal governance still matters.
Accountability should reside with a senior executive, risk oversight should include dependency on space infrastructure and investment should be evaluated alongside other major enterprise technology infrastructure rather than treated only as innovation spend.
Space joins the technology stack
For technology leaders, the strategic question is how space based capabilities integrate with existing systems.
KPMG argues that this is not about building an isolated space architecture.
Satellite derived data becomes another trusted input for existing cloud, edge, ERP, risk and AI systems.
The architecture spans a physical layer of satellites and launch infrastructure, a data and connectivity layer, an intelligence layer containing AI and edge computing and an application layer where the value becomes tangible inside industries.
Most enterprises will not own satellites.
Integration rather than ownership is the default model.
The challenge is access, reliability, governance, vendor orchestration and architectural discipline.
Space dependency is also enterprise risk
Boards are conditioned to think in failure modes.
The current conversation around space often focuses on opportunity, but resilience increasingly depends on space based systems.
A GNSS or GPS outage could disrupt financial timing, electrical grids and transport.
Satellite communications failure could create operational blackouts for remote industries.
Earth observation interruption could remove intelligence used in insurance, agriculture, supply chains or climate modelling.
Space failure is therefore not necessarily a specialised aerospace risk.
It can become an enterprise level business continuity issue.
Areas KPMG is watching
Bioeconomy: from refining materials to engineering biology
KPMG asks whether the next industrial revolution may be built not only on silicon but on biology.
For most of the industrial age, value came from transforming physical materials. The digital era shifted the frontier to information.
The emerging change is the ability to program biology itself.
Reading DNA has become dramatically cheaper, while AI has improved biological design, including predicting and designing proteins.
KPMG argues that biology is increasingly becoming an engineering substrate.
A living cell can act as a factory, producing fuels, materials, food or medicines from renewable inputs.
Precision fermentation, cultivated proteins and bio based chemicals already exist commercially.
The unresolved problem is often not whether the science works but whether production can scale from the laboratory into commodity economics.
The competition is also moving beyond individual companies.
Governments increasingly treat biomanufacturing as strategic infrastructure, pairing industrial policy with investment.
Why the bioeconomy matters
Resource security is becoming more important as geopolitical disruption exposes dependencies on global supply chains.
Bio based production can potentially localise important inputs.
Companies are reassessing geographically concentrated manufacturing.
Government industrial policy also shapes where production and innovation develop.
The business opportunity can emerge through supply chain resilience, greater cost stability, proprietary materials and alignment with government incentives.
But commercial scale remains much smaller than technological potential.
KPMG notes that bioplastics represent around 0.5% of global plastics production, biofuels less than 3% of transport fuel and bio based chemicals around 2%.
For most enterprises, KPMG therefore recommends awareness and selective preparation rather than immediate large scale capital commitment.
The practical first step is understanding where the organisation is exposed to critical materials that biological alternatives may eventually replace.
The Perception Layer
Another area KPMG is watching is what it calls the Perception Layer.
Advances in sensing, photonics and semiconductors are making the physical world increasingly visible in real time.
The change is from reactive towards real time, estimated towards observed and partial visibility towards continuous awareness.
In logistics, this can mean continuous knowledge of asset location and condition rather than reacting to delays.
In utilities, forecasting demand can evolve towards real time optimisation.
In manufacturing, scheduled maintenance can increasingly give way to recognising failure signals as they emerge.
The shift is not simply towards adding more sensors.
It is towards systems that can continuously observe, interpret and respond to the physical environment.
This matters across the other frontiers.
Advanced AI needs real world data. Physical AI depends on perception. Energy infrastructure needs sensing to optimise systems dynamically. Next generation computing supports continuous processing. Trust increasingly requires evidence of what occurred in the physical world. Space provides persistent observation.
The Perception Layer therefore functions as connective infrastructure between digital intelligence and physical reality.
What’s next?
While the path to AGI remains uncertain, the direction is not.
AI will reshape every frontier that matters to business.
As capability advances, diffuses through the economy and becomes more foundational, leaders will need to rethink strategy, risk, trust and transformation together.
No area of the business will be untouched.
From digital trust and Physical AI to quantum and space, the organisations that move early and adapt as conditions change will be best positioned to shape the next decade rather than simply react to it.
The frontiers in the report are not separate stories.
They are multiplying.
The leaders who treat them as one strategic landscape rather than six parallel initiatives will set the pace for everyone else.
Methodology
KPMG uses a mixed method foresight approach designed to identify signals of disruption and organise them into larger patterns.
The process looks across headlines, product announcements, expert opinion, client sentiment, capital flows and other quantitative and qualitative information.
Signals are evaluated against four dimensions: strategic relevance to long term business goals, ability to disrupt or redefine industries, readiness for adoption given current barriers and enablers and the speed at which the signal is moving from niche towards mainstream.
KPMG then uses these signals to identify broader macro trends and translate the findings into short and longer term recommendations.
The intention is not to predict one fixed future.
It is to identify the relationships between emerging forces early enough that leaders can make better choices about strategy, investment, risk and operating models.
Conclusion
KPMG’s 2026 Frontiers Report is ultimately less about six individual technologies than about the changing structure around enterprise technology.
Its central idea is that AI has become a multiplier.
More capable reasoning makes physical machines more autonomous. More autonomous systems increase the need for continuous verification and identity. Greater AI deployment increases compute demand. Compute creates pressure on energy infrastructure. Those constraints encourage new computing architectures. More sophisticated sensing connects digital intelligence to physical systems. Space extends sensing and infrastructure beyond Earth.
Each frontier changes the others.
That means technology strategy becomes harder to manage as a series of independent investment decisions.
The organisation increasingly needs to understand dependencies across data, compute, energy, infrastructure, identity, governance, people and physical operations.
The Path to AGI chapter provides perhaps the clearest articulation of the economic shift.
If intelligence becomes increasingly scalable, intelligence itself becomes less differentiated.
The scarce resources move elsewhere.
Judgment becomes more important. Context becomes more important. Institutional memory becomes more important. Governance becomes more important. Human accountability becomes more important.
The same principle applies across the other frontiers.
Physical AI is not simply more robotics. It requires an operating model capable of governing intelligent action in the physical world.
Energy and Compute is not simply an infrastructure story. It creates constraints on AI strategy and introduces compute as an operational dependency.
Quantum is not simply a future machine. It changes security migration decisions today.
NextGen Computing is not simply better hardware. It changes how enterprises orchestrate heterogeneous computing environments.
Future of Trust is not simply a governance framework. It moves verification, identity and provenance into the architecture itself.
Space is not simply a new market. It is increasingly part of the invisible infrastructure on which existing industries already depend.
The competitive advantage therefore moves towards the ability to orchestrate these systems.
Enterprises do not necessarily need to own every model, data centre, robot, chip, energy source, quantum computer or satellite.
They need to know what they depend on, which layers matter strategically, where control is necessary, where partnerships are sufficient and how the pieces fit together.
That is why KPMG’s description of AI as the frontier multiplier is useful.
AI does not merely replace existing systems.
It increases the value, speed and interdependence of the systems around it.
The leadership challenge is therefore no longer only how quickly an organisation adopts AI.
It is whether the enterprise can adapt its work, infrastructure, trust architecture, operating model and investment choices as intelligence becomes embedded across all of them.
Key takeaways
AI is shifting from a standalone technology category into a connective capability across the economy. Improvements in intelligence increasingly affect robotics, compute, energy, trust, sensing and infrastructure simultaneously.
Scalable intelligence changes the economics of the enterprise. As reasoning capacity becomes easier to scale, the scarce resources increasingly become context, judgement, institutional knowledge, trust and accountability.
Governance needs to become continuous. Autonomous systems increasingly operate faster than traditional approval cycles, making identity, provenance, observability, verification and embedded controls part of the operating architecture.
Physical AI extends the impact of AI beyond knowledge work. Intelligent systems increasingly act in the real world, raising the importance of safety, liability, monitoring and human oversight.
Energy and compute are becoming strategic constraints. AI growth increasingly depends on power, grid access, memory, chips, cooling, land, permitting and dependable inference capacity.
The future computing environment will be heterogeneous. Quantum, photonics, neuromorphic systems and other specialised architectures will sit alongside conventional computing, making workload allocation and orchestration strategic capabilities.
Trust must be engineered rather than assumed. As synthetic media and autonomous agents scale, identity, provenance and system behaviour need to become independently verifiable.
Space is moving into the enterprise infrastructure stack. Organisations already rely on satellite enabled connectivity, timing and sensing, while AI is making space derived information increasingly valuable.
The six frontiers should not be managed as six separate technology programmes. Their value and risk increasingly come from how they interact.
The simplest way to capture KPMG’s argument is: AI is the multiplier, but orchestration determines who captures the advantage.
