{"slug": "ai-in-everyday-devices-tokenised-intelligence-at-the-edge", "title": "AI in everyday devices: tokenised intelligence at the edge", "summary": "MIPS, a client of Business Reporter, reports that artificial intelligence is moving from centralized cloud environments into everyday devices through edge AI, custom silicon, and embedded computing, with AI agents and tokens becoming fundamental to device operation. The shift addresses limitations of cloud-centric AI such as latency, bandwidth, privacy, and energy inefficiency, requiring new compute architectures for edge devices to handle continuous token flows with deterministic performance and ultra-low latency.", "body_md": "# AI in everyday devices: tokenised intelligence at the edge\n\nTHE ARTICLES ON THESE PAGES ARE PRODUCED BY BUSINESS REPORTER, WHICH TAKES SOLE RESPONSIBILITY FOR THE CONTENTS\n\n- Bookmark\n\n*MIPS is a Business Reporter client*\n\nArtificial intelligence is reshaping computing by moving intelligence from centralised cloud environments directly into everyday devices. Advances in edge AI, custom silicon and embedded computing are enabling a new generation of intelligent systems that are distributed, context-aware and increasingly autonomous.\n\nAt the heart of this transformation are AI agents operating locally within devices and tokens serving as the fundamental units of AI computation and communication. Together, these technologies are redefining how devices process information, interact with users and respond to the physical world, laying the foundation for a future where embedded AI becomes as ubiquitous as operating systems are today.\n\n**Agents and tokens are coming to a device near you**\n\nFor decades, computing has been defined by a simple model: applications run on operating systems, data is processed in centralised clouds or relatively fixed-function endpoints and intelligence is delivered as a service. That model is now shifting, thanks to rapid advances in artificial intelligence, edge computing and heterogeneous silicon, all of which are driving a new paradigm where intelligence is distributed, contextual and increasingly autonomous.\n\nAt the centre of this shift are two concepts that will define the next generation of computing: agents and tokens.\n\nAgents are rapidly evolving from cloud-based assistants into embedded, always-on systems that live directly inside devices. Tokens are becoming the new currency of computation, flowing not just through data centres but across endpoints, sensors and embedded systems.\n\nTogether, agents and tokens are transforming how devices operate, communicate and interact with the physical world. In the coming decade, they will become as common in edge devices as operating systems and microcontrollers are today.\n\n**From cloud-centric AI to edge-native intelligence**\n\nToday’s AI systems largely depend on centralised inference, where massive models are hosted in hyperscale data centres. But as applications expand into robotics, autonomous systems, industrial automation and consumer devices, this approach is encountering fundamental limitations: latency, bandwidth constraints, privacy concerns and energy inefficiency.\n\nThe next phase of AI demands intelligence at the edge, where decisions are made locally, in real time and in direct response to physical environments.\n\nThis is where agents come in. Instead of relying on constant cloud connectivity, edge-native agents can interpret sensor data, maintain local context and take immediate action. Whether it’s a robotic arm adjusting to a changing production line, a smart appliance optimising energy usage or an autonomous drone navigating unpredictable conditions, agents are becoming the software layer that bridges gap between perception and action.\n\nBut to enable this shift, devices need more than software innovation. They require a fundamentally new compute architecture.\n\n**Tokens: the new currency of edge computation**\n\nIn large language models and modern AI systems, tokens represent discrete units of meaning, words, data points or symbolic elements processed during inference. As AI moves to the edge, tokens are no longer just abstract inputs to a model; they become a measure of real-time computational workload distributed across constrained devices.\n\nIn this new environment, every interaction including sensor readings, user inputs and environmental changes can be broken into tokenised streams that must be processed efficiently and securely. That means future edge devices must be capable of handling continuous token flows with deterministic performance, ultra-low latency and strict power budgets.\n\nUnfortunately, this shift places unprecedented demands on today’s silicon architecture.\n\n**Why traditional architectures are not enough**\n\nLegacy processor architectures were not designed for continuous AI inference at the edge. General-purpose CPUs struggle with power efficiency at scale, while GPUs and accelerators are often too power-hungry or too centralised for embedded environments.\n\nWhat’s emerging instead is a need for purpose-built, scalable compute platforms that can support heterogeneous workloads: real-time control, signal processing and AI inference all within the same system.\n\nThis is where the MIPS by GF’s vision becomes critical.\n\n**MIPS: enabling the physical AI era**\n\nAs a modern, RISC-V-native compute IP provider, MIPS is focused on enabling what it calls the “Physical AI” era, where intelligence is embedded directly into machines that sense, think, act and communicate in the real time.\n\nIn this new paradigm, MIPS is not simply providing processor cores. It is delivering a foundation for deterministic, safety-critical and edge-optimised computing designed specifically for agent-driven systems.\n\nThrough its RISC-V-based architecture and scalable processor IP, MIPS enables developers to build systems where agents can operate reliably at the edge, managing tokenised AI workloads without dependency on centralised compute.\n\nThis is particularly important for applications such as robotics, autonomous machines, industrial automation and intelligent infrastructure, where timing, reliability and safety are non-negotiable.\n\nWorking in alignment with the broader semiconductor ecosystem, including advanced manufacturing capabilities from GlobalFoundries, MIPS is helping bridge the gap between software-defined intelligence and physically grounded systems.\n\n**Agents in everyday devices**\n\nAs agent-based computing matures, its reach will extend far beyond enterprise or industrial systems. Everyday devices, from home appliances to personal electronics to connected vehicles, will increasingly behave as autonomous agents.\n\nA thermostat will not just respond to temperature changes but anticipate occupancy patterns. A camera will not just capture images but interpret context and trigger actions. A wearable will not just track health metrics but proactively co-ordinate with other systems to optimise outcomes in real time.\n\nIn each case, agents will rely on continuous token processing to interpret the world, reason about state, and act accordingly.\n\n**The infrastructure challenge**\n\nTo support this shift, the industry must solve three fundamental challenges:\n\n- Deterministic performance at the edge: ensuring real-time response under strict latency constraints\n- Energy efficiency at scale: enabling continuous AI inference without draining power budgets or resources\n- Security and reliability: maintaining trust in systems that operate autonomously in physical environments\n\nMIPS’ approach addresses these challenges by focusing on modular, scalable RISC-V compute IP designed specifically for embedded AI workloads. This enables system designers to tailor compute architectures to specific agent behaviours and token processing requirements.\n\n**The future: distributed intelligence everywhere**\n\nThe transition to agent-driven, tokenised computing is already underway. We are entering a world where intelligence is no longer centralised in the cloud but instead distributed across billions of devices. The question for developers, system architects and silicon designers is no longer if this shift will happen, but how fast they can adapt.\n\nMIPS is helping lead this transformation by delivering the foundational compute IP required for the Physical AI era.\n\n**To learn more about how MIPS is enabling agents and tokens in everyday devices, and how RISC-V-based architectures can power the next generation of intelligent systems, visit **[mips.com](https://mips.com/).", "url": "https://wpnews.pro/news/ai-in-everyday-devices-tokenised-intelligence-at-the-edge", "canonical_source": "https://www.independent.co.uk/news/business/business-reporter/ai-devices-intelligence-agents-tokens-b3036705.html", "published_at": "2026-08-26 07:00:00+00:00", "updated_at": "2026-08-26 07:12:57.528707+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-infrastructure"], "entities": ["MIPS", "Business Reporter"], "alternates": {"html": "https://wpnews.pro/news/ai-in-everyday-devices-tokenised-intelligence-at-the-edge", "markdown": "https://wpnews.pro/news/ai-in-everyday-devices-tokenised-intelligence-at-the-edge.md", "text": "https://wpnews.pro/news/ai-in-everyday-devices-tokenised-intelligence-at-the-edge.txt", "jsonld": "https://wpnews.pro/news/ai-in-everyday-devices-tokenised-intelligence-at-the-edge.jsonld"}}