The Age of Pervasive Intelligence

“The next Great Split separates those who control the architecture of intelligence from those who merely consume its output.”

On October 23, Apple will begin selling its first foldable iPhone, the iPhone Duo—a 7.6-inch internal display, advanced processor and redesigned software that represent its most ambitious attempt to consolidate mobile computing into one versatile device. Ten days earlier, on October 13, Meta expands commercial availability of its Ray-Ban Display AI glasses into France, Italy and Germany, moving the opposite way: toward intelligence accessed through vision, voice, gesture and the physical environment.

These are not merely two product launches. Apple is making the personal device more capable. Meta is attempting to make the device less central. Yet the larger revolution lies beyond both: we are entering the Age of Pervasive Intelligence, in which intelligence is no longer confined to a phone, computer or screen, but becomes a distributed capability embedded in devices, vehicles, factories, infrastructure and cities.

Agentic AI provides the decisive acceleration—artificial intelligence moving from answering questions to executing authorized tasks and coordinating interconnected physical and digital systems. The next Great Split will separate the economies that control the architecture of intelligence from those that merely consume its output.

Three regimes have defined modern computing: centralized mainframes; personal and mobile devices; and now AI-Native Pervasive Computing—distributed intelligence, continuous contextual awareness, agentic software and interoperable physical devices.

In the mobile era, consumers learned to operate ever more sophisticated devices. In the pervasive-intelligence era, technology learns to interpret and support human objectives. Instead of navigating an application's interface, the user expresses an intention; an authorized system interprets the objective, coordinates relevant services and presents an outcome.

This relocates economic value. The application becomes less visible; the intelligent orchestrator becomes more important; the user's trusted relationship with the system becomes a strategic asset. The defining transition is from device-centric computing to intelligence-centric economic systems.

Apple's iPhone Duo embodies consolidation: portability with a larger working surface, advanced computation, personal context and an integrated software ecosystem. The objective is to consolidate more digital activity within Apple's controlled hardware, software, payments and services environment—potentially reducing the need for separate tablets or lightweight computers.

Meta moves the other way, making intelligence accessible during ordinary physical activity—contextual assistance, translation, navigation, image interpretation and eventually agentic services—without requiring users to retrieve a phone. Its Neural Band adds another interface, interpreting subtle muscular signals to control functions without touch.

The two models may coexist: Ambient Intelligence for immediate assistance through the day, Immersive Intelligence for complex reasoning, creation and execution. The strategic contest is over who controls the transition between them—who owns identity, remembers context, determines which agent receives an instruction, authorizes transactions and captures the value when a task completes.

Conventional AI generates information, recommendations or content. Agentic AI can plan a sequence of actions, call specialized tools, coordinate services, evaluate results and execute tasks within defined permissions. An assistant explains how to organize a journey; an agentic system compares options, verifies schedules, assembles an itinerary and completes authorized bookings. The opportunity is not faster information—it is the reduction of the coordination costs required to turn information into action.

The evidence points toward acceleration, but unevenly. McKinsey's 2026 survey finds 40% of large enterprises scaling AI agents, up from 27%, yet only 37% report a positive EBIT impact and just 6% qualify as high performers. Gartner forecasts that 80% of tangible agentic returns in 2028 will come from domain-specific agents, not general-purpose systems. Adoption is accelerating faster than measured financial returns.

A reinforcing loop could emerge: interface friction declines; the value of each interaction rises as systems execute multistep workflows; specialized agents extend intelligence into logistics, manufacturing, energy and mobility; and successful deployment funds further diffusion. But the flywheel is not automatic—security failures, expensive inference and unreliable execution can interrupt it. The decisive metric becomes cost per successfully completed economic task. The productivity discontinuity arrives when AI stops improving individual tasks and begins redesigning the system connecting them.

In a multinational industrial firm, procurement, engineering, finance, maintenance, scheduling and logistics today function through separate systems, coordinated by meetings, approvals and reports. An agentic architecture can connect these functions directly: a demand change adjusts production planning, which informs procurement, evaluated against supplier availability, financing and logistics—while managers retain responsibility for objectives, risk limits and authorization.

The result is a new enterprise velocity—the rate at which an organization converts information into coordinated economic action. Three financial consequences follow: working-capital requirements may fall; asset utilization may rise; and incremental growth may require less organizational complexity, improving return on invested capital. But agent errors can propagate, poorly defined permissions create exposure, and software failures in physical infrastructure carry safety consequences. Winners will combine agentic capability with rigorous oversight and resilience.

The Apple–Meta confrontation is the consumer-facing expression of a larger struggle between the United States and China over the foundations through which pervasive intelligence operates.

Architecture One — Platform-Centric Intelligence. The American ecosystem excels in advanced computing, frontier research, operating systems, cloud, enterprise software and global consumer platforms. Nvidia, Apple, Meta, Microsoft, Google and Amazon control strategic layers of the stack, concentrating value around high-performance computing, proprietary technology and integrated platforms—reinforced by extraordinary access to capital.

Architecture Two — Industrially Embedded Intelligence. China's manufacturing scale, electronics supply chains, EV and robotics sectors and connected-device production create opportunities to embed AI into physical products, aided by open-weight models and alternative computing architectures. Advantage arises from manufacturing efficiency, affordable hardware and large-scale commercialization.

Neither distinction is absolute, and the ecosystems remain commercially interconnected. But the next divide may be less about which country produces the most sophisticated model than which system combines intelligence, physical deployment and sustainable capital returns most effectively.

Governments increasingly recognize that control over AI infrastructure shapes military capability, industrial competitiveness and economic sovereignty. Advanced semiconductor restrictions are one manifestation; AI models, identity technologies, cloud infrastructure and device standards could become others.

But the world need not divide into two closed coalitions. A more plausible outcome is asymmetric interdependence: India pursuing digital sovereignty while buying American computing; Gulf economies investing in U.S.-aligned AI while trading extensively with Asian manufacturers. The important variable is the degree of dependence at each layer of the stack. The next balance of power will depend not only on who invents intelligence, but on who controls its deployment.

Europe cannot credibly duplicate every layer of the American or Chinese ecosystems. Its strengths lie elsewhere: industrial engineering, precision manufacturing, transport and logistics, electrical equipment, aerospace, automotive and regulated infrastructure—industries rich in installed capital and ideal for specialized Agentic AI.

Europe could pursue a third architecture: Trusted Industrial Intelligence—embedding secure, interoperable and financially productive AI across its physical economy, from factory scheduling and predictive maintenance to energy coordination and port utilization. Yet obstacles are substantial: fragmented capital markets, uneven digital adoption, expensive energy, complex permitting and weak links between research and deployment. America can lead in platform intelligence, China in industrial diffusion; Europe must compete to lead in trusted industrial intelligence.

The smartphone era organized an economy around applications, app stores, search and attention. Agentic AI challenges that: an intelligent orchestrator may become the principal point through which consumers access commercial services, leaving the underlying applications necessary but less visible.

Three consequences follow. First, distribution power moves toward orchestration—the next gatekeeper controls the interface through which services are recommended and executed. Second, software development becomes agent-assisted; with 32% of organizations declining a purchase they could build internally, undifferentiated vendors face pressure while those offering proprietary data or trusted execution prove more resilient. Third, monetization shifts toward completed outcomes—transaction fees, subscriptions and charges linked to verified execution. The Attention Economy monetized engagement; the Execution Economy will monetize results.

The first phase of the AI Supercycle has been dominated by investment in data centers, advanced semiconductors, accelerated computing, energy and networking. The cycle is extraordinary, but returns remain uneven. Pervasive intelligence opens a second phase in which intelligence is deployed across consumer activity, enterprise workflows and physical infrastructure—and investors must distinguish the capital required to produce intelligence from the value generated when it is deployed.

A railway's AI infrastructure creates value only when it improves traffic management, reduces maintenance, raises availability or optimizes energy. The key distinction becomes companies that depend on continuous new investment to grow versus those capable of generating progressively greater output from installed capital. If agents reduce coordination costs and improve utilization, they raise the productivity of existing capital—bridging the gap between the buildout's capital demands and the need for sustainable returns. But that is not automatic: models must be reliable, execution secure, savings measurable, and total operating cost below value created. The next phase will reward the conversion of intelligence into productive cash flow, not the accumulation of computing capacity alone.

One — The Integrated Intelligence Platform. A few large companies retain control over operating systems, AI orchestration, identity and distribution; wearables operate within proprietary ecosystems. The principal risk is concentration of power.

Two — The Fragmented Intelligence World. The U.S. and China develop separate architectures across semiconductors, operating systems, cloud and governance. Third countries face harder choices; investment follows geopolitical alignment; industrial efficiency suffers from duplication.

Three — The Pervasive Intelligence Economy. Intelligence embeds across devices, industry and infrastructure through interoperable, permissioned services; enterprises reorganize around execution; the largest returns come from productive deployment across existing and new assets. This offers the greatest potential for broad productivity—and requires the most demanding combination of interoperability, financial discipline and institutional trust.

Five Strategic Takeaways

For investors, boards and policymakers

• Redefine the investment thesis. The frontier is no longer personal computing but Pervasive Intelligence across consumer behavior, enterprise operations and physical infrastructure.
• Treat Agentic AI as the decisive accelerator. Measure completed tasks, cost efficiency, cash conversion and incremental return on invested capital.
• Position for the next advantage. Models alone are insufficient; identity, trusted orchestration, semiconductors, interoperable services, energy and industrial deployment determine value.
• Prepare for a complex U.S.–China divide. Manage technological dependencies without assuming the world splits into two closed blocs.
• Make Industrial AI Europe's renewal. Combine sovereignty with strengths in manufacturing, mobility, engineering and infrastructure—targeting measurable productivity, not expenditure for its own sake.

The first computing revolution placed intelligence inside institutions; the second placed it in the hands of individuals; the third is embedding it throughout the physical economy. The next Great Split will be between those who control intelligence, those who deploy it productively, and those who become dependent on both.