The Internet of Things did not arrive with a single invention. It accumulated, over several decades, from ideas in supply chain logistics, embedded computing, wireless networking and cloud services that happened to converge. Understanding that sequence explains a lot about the field today: why there are so many competing radio standards, why security was an afterthought for so long, and why the term itself is both widely used and loosely defined.
This guide picks out the milestones that changed how people built connected devices, grouped roughly by decade. It focuses on events with lasting technical consequences rather than product launches, and keeps to dates that are well documented.
Before the name: networked machines
The earliest experiments were about curiosity more than business. In the mid-1970s, members of the computer science department at Carnegie Mellon University wired the departmental Coca-Cola vending machine to their PDP-10 so that anyone who could reach it over the ARPANET could check from their desks whether it had cold bottles in stock. In 1991, a research group at the University of Cambridge pointed a camera at a shared coffee pot and streamed the image to their terminals, later to the web, to avoid wasted trips. Both capture the basic motivation of IoT: a remote check on a physical state that would otherwise require walking there.
In industry, machines had been exchanging data for years over serial fieldbuses such as Modbus, and remote monitoring over cellular links was already a business under the name machine-to-machine communication. What was missing was a shared network layer, cheap radios, and software to make sense of the data.
1999: a name and a protocol
The phrase “Internet of Things” is credited to Kevin Ashton, who used it in 1999 while working on supply chain technology at Procter & Gamble. His argument was that computers depended entirely on humans for data entry, and that giving objects a way to identify themselves, through RFID tags, would let computers observe the world directly. The Auto-ID Center at MIT, founded the same year, developed the idea into a standard architecture for tagging goods.
In the same year, two engineers at IBM and Arcom wrote the first version of MQTT to carry data from oil pipeline SCADA systems over expensive satellite links. It was designed to be small and to tolerate disconnection, and it would become the most widely used IoT messaging protocol two decades later. The MQTT guide explains why its design held up.
The 2000s: low-power radios and open hardware
The first decade of the century produced the radio foundations. The IEEE 802.15.4 standard, first published in 2003, defined a low-rate, low-power physical layer for short-range mesh networks. Zigbee built on it to create a complete stack for home and building automation, and the Zigbee guide covers how. Later in the decade the IETF published 6LoWPAN, starting with RFC 4944 in 2007, which compressed IPv6 packets enough to run over the same radios, setting up the IP-based approach that Thread would adopt.
Bluetooth, already common in phones, gained a low-energy mode with the Bluetooth 4.0 specification in 2010 that made coin-cell peripherals practical. Bluetooth Low Energy became the default link for wearables.
Hardware changed too. The Arduino board, born in 2005, made microcontroller programming accessible to designers and students who had never written embedded code. Open hardware and low-cost development boards meant that prototyping a connected device no longer required an electronics team.
The 2010s: cloud, consumer products and scale
The second decade turned experiments into an industry. Three developments dominate.
Cloud platforms. Public cloud providers added managed services for device connectivity in the middle of the decade: an MQTT broker, a device registry, identity and rule engines bundled as an IoT platform. This removed the need for every company to build its own back end and made it realistic to connect millions of devices. The IoT platforms guide describes what these services do.
Consumer smart home. The Nest learning thermostat, launched early in the decade, showed that a connected appliance could be a mainstream product, and its acquisition by Google a few years later signaled how seriously large technology companies took the category. Voice assistants followed, turning the smart home into a platform competition among a handful of ecosystems.
Low-power wide-area networks. LoRaWAN, standardized by the LoRa Alliance from 2015, and the cellular variants NB-IoT and LTE-M, standardized by 3GPP in Release 13 in 2016, filled the gap between short-range mesh and expensive cellular: kilometers of range on a battery that lasts years. This made city-scale and agricultural deployments economical.
The decade also brought the term Industry 4.0, first used publicly at the Hannover trade fair in 2011 and later adopted as a project of the German government’s high-tech strategy, and OPC UA emerged as the common data model for connecting factory equipment to IT systems. CoAP was published as an IETF standard in 2014, and MQTT was formally standardized by OASIS the same year. The Raspberry Pi, released in 2012, gave gateways and edge prototypes a cheap Linux platform.
2016: Mirai and the security reckoning
In October 2016, the Mirai botnet, which at its peak had infected around 600,000 devices, mostly cameras, video recorders and routers running with default credentials, was used to launch a DDoS attack against a major DNS provider. Large parts of the web became unreachable for hours. It worked because a generation of devices had shipped with known passwords, no update mechanism and open management ports.
Mirai changed the conversation. Security guidance that had existed for years started to be written into standards and then into law: ETSI EN 303 645 for consumer devices, the UK’s PSTI Act, the EU Cyber Resilience Act and the U.S. Cyber Trust Mark. The regulation guide covers what these require.
The 2020s: interoperability and intelligence at the edge
The current decade has been about fixing two long-standing problems: fragmentation and the cost of sending everything to the cloud.
Matter, published by the Connectivity Standards Alliance in October 2022 with backing from the major smart home ecosystems, defines a common application layer over IP so that a device certified once works with any compatible controller. It runs over Wi-Fi, Ethernet and Thread, and the Matter guide explains its scope and limits.
At the same time, processors capable of running neural networks have become cheap enough to put in cameras, sensors and gateways. Edge AI and TinyML move interpretation of data onto the device, reducing bandwidth, latency and privacy exposure. The edge AI guide and TinyML guide cover this shift.
What the history teaches
Three lessons stand out. First, protocols and radios outlive products: MQTT and 802.15.4 are decades old and still central. Choosing open, standardized building blocks protects a deployment against vendor lock-in. Second, security costs nothing to ignore until it costs everything; Mirai is the clearest demonstration. Third, each wave of IoT growth followed a drop in the cost of one component: tags, radios, boards, cloud back ends, and now inference hardware. Watching which cost is falling next is a reasonable way to predict where the field goes.
Where to go next
- What is the Internet of Things? gives the definition that this history converges on.
- IoT connectivity overview compares the radios introduced along the way.
- IoT security fundamentals covers the defenses that Mirai made unavoidable.
- IoT security regulation summarizes the laws that followed.
Topics: history, milestones, fundamentals