Electronics Guide

The Age of Intelligent Connected Devices

The period from 2015 to the present has witnessed the convergence of two transformative technology trends: the Internet of Things (IoT), connecting billions of devices into networked systems, and artificial intelligence, enabling machines to learn, reason, and act with increasing autonomy. Together, these technologies have fundamentally reshaped the electronics industry and daily life, creating smart homes, autonomous vehicles, intelligent assistants, and industrial systems that adapt and optimize themselves in ways previously relegated to science fiction.

The Internet of Things emerged from decades of work on embedded systems, wireless networking, and sensor technology, but achieved critical mass as component costs fell, connectivity became ubiquitous, and cloud platforms provided the infrastructure to manage billions of devices. Simultaneously, breakthroughs in deep learning enabled by improved algorithms, vast data availability, and accelerated computing hardware transformed AI from a niche research area into a practical technology deployed across virtually every domain.

The intersection of IoT and AI created capabilities neither could achieve alone. Networked sensors generated torrents of data that AI systems processed to extract insights and automate decisions. AI running on edge devices enabled local intelligence without constant cloud connectivity. Voice assistants combining natural language AI with IoT control changed how people interacted with their environments. This convergence continues accelerating, with implications for every sector of the economy and nearly every aspect of human life.

Articles in This Category

The Internet of Things Expansion

The proliferation of connected devices accelerated dramatically during this period, although growth proved slower than the most optimistic early forecasts, which circulated around 2010 and promised fifty billion connected devices by 2020. Industry analysts estimate that active IoT connections surpassed non-IoT connections for the first time around 2020, reaching roughly twelve billion, and that the installed base grew to approximately eighteen billion in 2024 and roughly twenty billion by the end of 2025. Counts vary widely across analysts because definitions of an IoT device differ. These devices ranged from simple sensors reporting environmental conditions to sophisticated systems like connected vehicles and industrial robots. The resulting network of intelligent devices created both unprecedented opportunities and significant challenges around security, privacy, and data management.

Consumer IoT applications brought smart technology into homes at unprecedented scale. Smart speakers, thermostats, lighting systems, security cameras, and appliances formed ecosystems managed through smartphone apps and voice commands. The convenience of automation and remote control drove adoption, while concerns about privacy, security vulnerabilities, and technology complexity created friction. Market leaders including Amazon, Google, and Apple competed for platform dominance while thousands of smaller companies developed specialized devices.

Industrial IoT transformed manufacturing, logistics, energy, and agriculture through connected sensors and intelligent systems. Factories deployed thousands of sensors monitoring equipment health, production quality, and operational efficiency. Supply chains gained visibility into shipment locations and conditions. Agricultural operations used IoT systems for precision farming, optimizing inputs based on soil conditions, weather, and crop health. These industrial applications demonstrated IoT's potential for improving productivity and reducing waste across sectors.

The connectivity layer diversified to match these very different workloads. Wi-Fi, Bluetooth Low Energy, Zigbee, and the low-power mesh protocol Thread served homes and buildings, where mains power or frequent recharging is available. Low-power wide-area networks served metering, asset tracking, and agriculture, trading bandwidth for range and battery life measured in years: LoRaWAN and Sigfox operated in unlicensed spectrum, while 3GPP standardized the licensed-spectrum alternatives NB-IoT and LTE-M in Release 13 in 2016. Operators then began retiring 2G and 3G networks, stranding an installed base of older machine-to-machine equipment and forcing costly migrations. Battery chemistry, duty cycling, and energy harvesting became as important to device design as the radio itself.

Security proved the weakest link. The Mirai botnet of 2016 conscripted hundreds of thousands of cameras and home routers that shipped with default credentials, then used them to mount distributed denial-of-service attacks large enough to disrupt major internet services. Regulators eventually responded with baseline requirements rather than guidance. The United Kingdom's Product Security and Telecommunications Infrastructure regime, in force since April 2024, bans universal default passwords and requires manufacturers to publish how long a product will receive security updates. The European Union's Cyber Resilience Act imposes mandatory cybersecurity requirements on products with digital elements, phasing in vulnerability reporting duties in 2026 ahead of full application in December 2027. In the United States, the Federal Communications Commission established the voluntary Cyber Trust Mark label for consumer connected devices. Long product lifetimes remain the central difficulty: a thermostat or an industrial sensor may outlast the company that wrote its firmware.

Artificial Intelligence Integration

The integration of AI into consumer electronics proceeded rapidly following breakthroughs in deep learning. Smartphones incorporated neural processing units that enabled on-device AI for photography enhancement, voice recognition, and predictive features. Voice assistants including Amazon Alexa, Google Assistant, and Apple Siri became household presences, fundamentally changing how people accessed information and controlled their environments. These AI capabilities, unimaginable in previous generations of electronics, became expected features in mainstream devices.

The development of AI hardware reflected the technology's importance. NVIDIA's GPUs, originally designed for graphics, became essential infrastructure for training neural networks once frameworks made their parallel arithmetic easy to reach. Custom accelerators followed. In 2016 Google disclosed its Tensor Processing Unit, an inference chip already running in its data centers, and Amazon, Microsoft, and a wave of startups later designed silicon of their own. Mobile processors gained dedicated neural engines, with Apple's A11 Bionic and Huawei's Kirin 970 both shipping in 2017. Numeric precision fell in step with the hardware, from 32-bit floating point toward 16-bit, 8-bit, and in the most aggressive designs 4-bit formats, because lower precision cuts both arithmetic cost and memory traffic.

Edge AI emerged as a crucial paradigm, bringing AI processing to devices rather than requiring cloud connectivity for every inference. This approach addressed latency requirements for applications like autonomous driving, preserved privacy by keeping sensitive data local, and enabled AI functionality without reliable internet connectivity. Quantization, pruning, and knowledge distillation shrank models until useful networks ran on microcontrollers with a few hundred kilobytes of memory, a practice that acquired the name TinyML. Keyword spotting, anomaly detection on vibration signals, and person detection became routine on parts costing a few dollars. The balance between edge and cloud AI continues evolving as device capabilities grow and new applications emerge.

The latter part of the era was defined by generative AI built on the transformer architecture introduced in 2017. The public release of large language models, most prominently ChatGPT in late 2022, brought conversational AI to a mass audience and triggered a surge in demand for accelerator hardware to train and serve ever-larger models. This demand reshaped the semiconductor industry: high-bandwidth memory, advanced packaging, and data-center GPUs became strategic resources, and access to leading-edge fabrication turned into an instrument of trade policy as governments restricted exports of the most capable accelerators. Energy became the other binding constraint. The International Energy Agency estimated that data centers consumed roughly 415 terawatt-hours in 2024, about 1.5 percent of global electricity, and projected that figure to roughly double by 2030 as accelerated servers proliferate. Rack power densities that once sat near ten kilowatts climbed past one hundred, pushing operators from air cooling toward liquid cooling and making power delivery, thermal design, and siting first-order engineering problems.

Smart Home Evolution

Smart home technology matured from early-adopter curiosity to mainstream product category during this period. Voice-controlled speakers from Amazon, Google, and Apple served as control centers for expanding ecosystems of connected devices. Smart lighting, thermostats, locks, cameras, and appliances could be controlled remotely, automated based on schedules or conditions, and integrated into routines that simplified daily life. The smart home demonstrated both the potential and limitations of consumer IoT.

Interoperability challenges fragmented the smart home market as competing ecosystems often failed to work together seamlessly. Matter, an application-layer standard published by the Connectivity Standards Alliance in 2022, aimed to address these challenges by defining common device types and letting one product work with several platforms over Wi-Fi, Ethernet, and Thread. Apple, Google, Amazon, and Samsung shipped support across their hubs and speakers, and successive revisions widened the standard's coverage from lights, plugs, and thermostats to appliances, energy management, and, in 2025, cameras and video doorbells. Adoption nonetheless proceeded unevenly. Early implementations exposed commissioning bugs, vendors continued to reserve their most distinctive features for their own applications, and older accessories on Zigbee or proprietary radios still required bridges.

Security and privacy concerns accompanied smart home adoption. Connected cameras and voice assistants raised questions about surveillance in private spaces. Device vulnerabilities created potential entry points for network attacks. Data collected by smart home devices provided detailed portraits of inhabitants' behaviors and routines. These concerns prompted both regulatory attention and industry efforts to improve device security and data practices.

Autonomous Systems and Robotics

Autonomous vehicles represented the most ambitious application of AI to physical electronics, combining computer vision, sensor fusion, and decision-making algorithms in safety-critical systems. Companies including Waymo, Cruise, and Tesla deployed vehicles with varying levels of autonomy on public roads, and their engineering choices diverged sharply: Waymo built redundant sensor suites combining lidar, radar, and cameras with detailed prior maps and remote assistance, whereas Tesla pursued a camera-centered approach intended to scale across a consumer fleet. Progress proved uneven. Waymo grew its driverless ride-hailing service from roughly fifty thousand paid trips per week in 2024 to about half a million per week across some ten metropolitan areas of the United States by mid-2026, whereas General Motors grounded Cruise's driverless operations after a 2023 pedestrian collision and ended funding for the robotaxi program in December 2024. Fully autonomous vehicles thus remained more geographically constrained than early predictions suggested, expanding city by city as each new operating area demanded validation, mapping, and regulatory approval, even as advanced driver assistance systems incorporating AI became standard features in new vehicles.

Robotics and automation advanced substantially with AI integration. Mobile fulfillment robots, exemplified by the technology Amazon acquired with Kiva Systems in 2012 and later branded as Amazon Robotics, transformed warehouse logistics, while companies such as Boston Dynamics advanced legged and mobile manipulation platforms. Collaborative robots, or cobots, worked alongside humans in manufacturing environments. Autonomous delivery robots and drones demonstrated potential for last-mile logistics. These developments pointed toward increased automation across numerous domains.

Consumer robotics saw both successes and failures. Robot vacuum cleaners achieved mainstream adoption, with AI enabling improved navigation and cleaning patterns. Social and companion robots attracted interest but struggled to find sustainable markets. The gap between consumer expectations, often shaped by science fiction, and practical robotic capabilities remained significant, though progress continued steadily.

Displays and Human-Machine Interfaces

Displays advanced as quickly as the silicon behind them. Organic light-emitting diode panels displaced liquid crystal displays across premium smartphones and spread into laptops, tablets, and televisions, aided by manufacturing scale in South Korea and China and by hybrid approaches such as quantum-dot color conversion layered over blue OLED emitters. Foldable phones, commercialized from 2019, proved that ultra-thin glass and multi-link hinges could survive years of folding, though price and crease visibility kept them a premium niche. MicroLED, long promised as the successor that combines inorganic brightness and lifetime with emissive contrast, remained difficult to manufacture, because transferring millions of microscopic dies onto a backplane at acceptable yield is an unforgiving process. The technology therefore stayed largely confined to modular large-format displays and prototypes.

Immersive interfaces matured more slowly than their advocates expected. Standalone virtual reality headsets brought inside-out tracking and untethered use to consumers, and Apple entered the category in 2024 with a mixed-reality headset built around micro-OLED panels of very high pixel density. Yet headsets remained a specialist product, while camera-and-audio smart glasses without displays found broader acceptance by asking less of the wearer. Voice interfaces, gesture recognition using millimeter-wave radar or time-of-flight sensors, and always-listening wake-word engines running on ultra-low-power cores pushed interaction toward the ambient computing ideal, in which the interface recedes into the environment. Research interfaces went further still. Academic groups decoded intended speech from cortical signals in people who had lost the ability to speak, and companies including Neuralink began investigational human implants of high-channel-count electrode arrays in 2024, though such systems remain experimental rather than clinical products.

Quantum and Neuromorphic Computing

Alongside conventional silicon, two unconventional computing paradigms moved from laboratory curiosity toward engineering discipline. Quantum computers became accessible over the cloud, with IBM offering public access to superconducting processors from 2016 and Amazon and Microsoft later brokering access to multiple hardware technologies. In 2019 Google reported that its 53-qubit Sycamore processor completed a random-circuit sampling task far faster than the best classical methods then known, a claim promptly contested as classical simulation algorithms improved. The deeper problem was never qubit count but noise. Quantum error correction spreads one logical qubit across many physical qubits, and only below a certain physical error rate does adding qubits reduce logical errors rather than multiply them. Google's Willow processor, announced in 2024 with 105 qubits, demonstrated that behavior on hardware: as the surface-code distance grew, the logical error rate fell rather than rose.

That result validated the theory without delivering a useful machine. Factoring a 2048-bit RSA key with Shor's algorithm would require on the order of millions of physical qubits at current error rates, along with cryogenic infrastructure and control electronics that scale with them. Anticipating that eventual capability, the United States National Institute of Standards and Technology published its first post-quantum cryptographic standards in 2024, and migration of long-lived systems began well ahead of any cryptographically relevant quantum computer. Practical near-term work concentrated instead on control electronics, cryogenic CMOS, error-mitigation techniques, and competing qubit technologies including trapped ions, neutral atoms, and photonics.

Neuromorphic computing took the opposite approach, borrowing not from quantum mechanics but from biology. Spiking architectures colocate memory with computation and communicate through sparse, event-driven pulses, so power is consumed only when something changes. Intel's Loihi 2 research processors underpinned Hala Point, a system delivered to Sandia National Laboratories in 2024 that packs 1,152 chips supporting up to 1.15 billion neurons within roughly 2,600 watts. IBM's NorthPole chip pursued a related idea from a different direction, keeping network weights on chip to eliminate the off-chip memory traffic that dominates the energy budget of conventional inference. Event cameras, which report per-pixel brightness changes instead of frames, provide naturally matched sensors. Adoption remains limited by an immature software ecosystem and by the relentless efficiency gains of conventional accelerators, but the underlying argument that sparse, event-driven processing wastes less energy keeps the field active.

Looking Forward

The convergence of IoT and AI continues reshaping electronics with no plateau in sight. Connectivity keeps advancing: 5G networks matured through successive 3GPP releases, and work on 6G moved from study items in Release 20 toward Release 21, which is expected to carry the first 6G specifications around 2028 and to feed the International Telecommunication Union's IMT-2030 process ahead of commercial systems near 2030. AI capabilities expand as algorithms improve and hardware accelerates, while the constraints shift from transistor density toward power delivery, memory bandwidth, packaging, and the supply of the advanced fabrication capacity concentrated in a handful of facilities. The boundaries between computing, communication, and sensing continue blurring as electronic devices become increasingly intelligent and connected.

Several tensions will shape the next phase. Intelligence must be partitioned between devices, network edges, and data centers according to latency, privacy, and energy budgets rather than habit. Security and update obligations now extend across product lifetimes measured in decades, a discipline the consumer electronics industry has never had to practice at scale. Sustainability pressures cut in both directions, since connected sensing reduces waste in buildings, grids, and farms while the devices themselves add to electronic waste and the models they feed add to electricity demand.

Understanding this ongoing transformation provides essential context for electronics professionals, students, and anyone seeking to navigate an increasingly technological world. The patterns established during this era, the companies and ecosystems that emerged, and the technical and societal challenges encountered will influence electronics development for decades to come. This category explores the key developments, technologies, and implications of the IoT and AI era.