Chapters

16 IoT Evolution: Technology Cycles and Convergence

applications
iot
systems
evolution
Listen
Uses your device's voice. Tap ▶ beside any paragraph to hear just that one.
  • Podcast audio coming soon
  • Character narration audio coming soon
Lecture
Video lecture coming soon. The slides are the lecture for now.
Ask AI
Pick a prompt. We copy it for you and open an AI service; paste it there. Nothing is sent to IoTClass.

16.1 From a Cheaper Sensor to a Useful Service

A tenfold gain can change which system design is practical, but a large device count does not explain why a design works. This chapter follows computing costs, connected entities and processor efficiency to explain how large-scale IoT became practical. It builds on IoT Evolution: Systems Foundations in the applications module.

Keep the chapter’s smart thermostat in view. One thermostat can save energy; thousands of connected thermostats may support city-wide demand response. The design question is what makes that larger service useful and affordable. A cheaper sensor helps, but so do communication, useful data, operating software and a service people continue to need.

Try it

Match technology trends with their roles in making connected systems practical at scale.

Match computing concepts →

The problem is separating an enabling technology from evidence of a viable service. We will first examine the classroom “10x cycle” comparison, then ask what changed when networks connected different entities, and finally explain how integration and power efficiency widened compute-placement choices. Along the way, distinguish illustrative prices, observed counts, forecasts and business outcomes: each answers a different question.

KEY CONCEPTS

  • Technology cycles

    Lower costs and useful functions expand where computing fits.

  • Service value

    Connected devices need a recurring benefit and sustainable economics.

  • Convergence

    Sensors, networks, compute and software work together.

  • Performance per watt

    Power limits shape where useful processing can run.

16.2 Learning Objectives

  • Apply the 10x technology cycle pattern to IoT examples.
  • Explain convergence through cost, compute, and network drivers.

16.3 Read Technology Cycles as Comparisons

A technology cycle is the rise and spread of a computing approach that enables new uses and helps support the next approach. The “10x” label is a teaching pattern for large changes, not a measured growth rule or an investment forecast. The mainframe, PC, mobile and embedded eras overlap; none erased the infrastructure that preceded it.

The schematic device sequence moves from roughly 103 mainframes in the 1960s to 106 PCs in the 1980s and 109 smartphones in the 2010s. These are classroom orders of magnitude, including a 1,000× progression between example nodes, rather than measured era totals. The illustrative price chain is $1M+ for a room-sized mainframe, $2,000 for a desktop PC, $200 for a phone and $2 for a sensor component. It is neither an inflation-adjusted comparison nor a comparison of equivalent systems.

Actual counts require named categories and dates. Mobile-cellular subscriptions reached 738 million worldwide in 2000; smartphones were only a subset. ITU reported that subscription measure, which differs from unique people or devices. The chapter’s IoT outlook is about 39 billion active connected IoT devices in 2030 and more than 50 billion by 2035. Those are forecasts of connected nodes and gateways, excluding unconnected sensors; the $2 component example does not describe their average installed price.

The historical comparison below retains the different categories so that the change in computing access stays visible. Its prices and unsourced counts are schematic examples, not measured market totals. In particular, a subscription count cannot be substituted for a PC installed base. Mainframe time-sharing let multiple users share central computing resources; PCs brought local software to individual desktops.

EraPeriodScale and categoryIllustrative priceCapability added
Mainframe1960sHistorical count not established here$1M+Centralized enterprise computing
Minicomputer1970sHistorical count not established here$100KDepartmental computing
Personal computer1980s–90s~100 million, schematic PC scale$2–5KIndividual computing power
Desktop Internet1990s–2000s~1 billion, schematic Internet-era scale$500–1KGlobal information access
Mobile cellular2010sAbout 5B worldwide subscriptions in 2010$200–500Mobile Internet is a separate measure
Internet of Things2020s–30s~20–40 billion, illustrative connected-device range$1–100Computing in physical objects

The 2010 mobile-cellular figure comes from ITU. Keep that dated observation separate from the future IoT figures and from the table’s teaching ranges.

Adoption S-curves progress from mainframes through PCs and mobile computing to IoT. Each era enables the next as cost and physical size fall and deployment scale rises.
Figure 16.1: Technology evolution cycles from mainframes through PCs and mobile devices to IoT sensors show how falling cost, shrinking form factor, and expanding capability moved computing toward embedded physical processes.

The overlapping-curve model is useful even without treating its vertical scale as a measurement. Each era rises and then flattens, while its components, infrastructure and services help enable the next. Mainframe computing helped support the PC era, and mobile integration helped make embedded designs practical. For a design review, identify the enabling relationship rather than assuming that the next curve must repeat a fixed numerical jump. Lower cost and smaller form factors expand possibilities; they do not set a guaranteed adoption rate.

16.4 The Two Drivers of Each Technology Cycle

Two drivers explain the design opportunity. First, components can become cheaper across generations: the examples run from a $1M computing unit through a $2,000 PC and $200 smartphone to a $1–10 sensor. A claimed 10–100× reduction per cycle still needs equivalent capability and dates. Second, functionality changes: PCs enabled individual productivity, the Internet global communication, mobile devices always-on access, and IoT ambient intelligence in physical environments.

The form-factor progression is qualitative: room, desktop, pocket, then embedded components. Calculation expanded into documents and Internet access, then apps and cameras, and finally sensing and control inside objects. Neither size nor functionality has a defensible universal 10× multiplier. For the thermostat, smaller and cheaper electronics create options near the physical process; useful services determine which options are worth operating.

Try it

Compare the illustrative computing eras, prices and deployment scales.

Explore technology cycle economics

16.5 Turn More Connections into Recurring Value

Lower component cost can extend a one-time product sale into an ongoing service relationship. Data analytics, predictive insights and ecosystem integration may contribute value over years of operation. Their share depends on the application and business model. There is no universal 5–9× lifetime revenue or gross-profit outcome: device price, service cost and retention determine the result.

For the thermostat, ask what recurring benefit the data supports. One unit can help save energy. A coordinated network of thousands may enable city-wide demand response, potentially worth millions, but that value needs an application-specific case. A connected device without useful insights or action can remain a gadget with Wi-Fi. Hardware and sustainable data-driven services must work together.

Network effects occur when adding participants makes a network more useful to others. This is the platform-economics opportunity in IoT, not proof that value grows exponentially with every device. Coordinated thermostats illustrate why a group may support a service that one device cannot. The number connected still does not establish who pays, what service is delivered or whether customers renew.

The same distinction applies when a team looks for an inflection point in a market. Falling endpoint cost can make a previously impractical deployment feasible, while a new service can give that deployment a purpose. Both drivers need to be present in the explanation. Cheaper thermostats with no useful additional service establish only the cost change. More connected thermostats with no account of the benefit establish scale, but leave the value claim unresolved.

A simple revenue example makes the boundary concrete. At $5 per month for 120 months, with no churn, gross service revenue is $600. Compared with $50 in one-time hardware revenue, that is a 12× gross-revenue comparison. It excludes costs and does not establish profit or customer value. Cross-era scale changes are similarly useful for scenario planning, not a numeric 10× investment rule.

Long-lived wireless sensors can support wide-area monitoring. A claim combining sub-$1 hardware with ten-year battery life, however, needs a device datasheet and a field duty cycle. Component price alone says little about the cost of installing, connecting and operating the service. The useful question is which recurring outcome the data enables at a sustainable cost.

16.6 Keep Market Evidence Attached to Its Measure

Market growth can suggest an opportunity, but it cannot supply the thermostat’s service case. IoT Analytics expected active IoT connections to grow 14% during 2025 to 21.1 billion by year end. Here, active connections refer to connected IoT nodes and gateways, rather than all sensors or physical objects. This dated expectation must retain its forecast status; it is not an observed end-year count.

Other projections describe different quantities. McKinsey Global Institute projected $11 trillion in potential economic value by 2030, while Statista projected 40+ billion IoT devices by 2034. Potential economic value, device forecasts and annual market spending are different measures. A spending estimate needs a category, geography and method before it can be compared with another market total.

Adoption also depends on population and technology. Agriculture has no single all-farms IoT adoption rate. As a separate, defined example, USDA ERS reported guidance autosteering on 70% of large-scale US crop-producing farms in 2023. That figure describes a particular practice and farm population; it cannot stand for agriculture as a whole. Manufacturing predictive maintenance, healthcare remote patient monitoring and utility smart-grid adoption likewise need a named survey, population and question wording before a percentage becomes useful.

Business benefits need equally clear boundaries. DOE/PNNL modeled approximately 29% potential commercial-building energy savings from properly tuned controls. Realized savings depend on the building. Reduced equipment downtime, manufacturing cost, irrigation water use and supply-chain inefficiency are possible benefits, but a transferable percentage needs a named study, baseline and scope.

Company examples show why counts and outcomes must remain separate. Amazon reported its one-millionth deployed robot in 2025. Separately, Amazon’s 2022 account said employees supported by its technologies picked, stowed or packed about five billion packages in 2021. The two statements have different dates and describe different work; they do not establish packages per robot.

Deere’s connected-machine progress and Philips connected-care products offer qualitative examples, pending dated machine-count definitions and outcome studies. Google estimated that Nest thermostats had saved users $14B in energy costs and 200B kWh cumulatively since 2011, as of April 2026. Those are estimated cumulative savings, not an installed-device count. For our thermostat case, the lesson is to preserve what each example measures before transferring it into a forecast or business model.

Check knowledge

Check technology-cycle comparisons, recurring value and the boundaries of forecasts.

Check technology cycles and value

16.7 Follow What Each Connectivity Phase Added

The cost story explains why more endpoints became affordable. The connectivity story explains what they could do together. Five overlapping phases connect different entities and carry earlier capabilities forward. Their date ranges are conceptual guides rather than clean replacement points, which helps explain both IoT’s diverse stack and its continuing interoperability challenges.

Network, 1969–1990. Connecting two or more computers enabled direct host communication. ARPANET began with four nodes in 1969 in academic and research settings. Devices were expensive, specialized and operated by trained technicians. Packet-switched networking became the dominant paradigm, and TCP/IP standardization in 1983 helped establish the shared networking foundation.

The Internet, 1991–2006. Tim Berners-Lee’s World Wide Web became publicly available in 1991. HTTP and HTML standardized information exchange, connecting people to information and services through globally connected computers. E-commerce, email and search engines supported new economic models. ITU estimated about 315 million Internet users at the end of 2000; that count does not imply that every user accessed the Internet through a wired desktop.

Mobile Internet, 2007–2012. The iPhone in 2007 and the Android ecosystem demonstrated demand for always-on access beyond desktops. 3G and 4G networks supported mobile apps, while app stores introduced a new software distribution model. Mobile access became increasingly important around 2012, although any claim that it overtook desktop usage must specify the measured traffic or activity.

A phone combined processors, graphics, signal processing, location, camera, motion sensors, short-range radios, cellular service, storage and a user account. Apps composed those capabilities into maps, field notes, mobile payments, ride sharing and StoryCorps-style audio storytelling. This was a shift from voice-centered telephony toward feature-centered systems. An IoT team could start from the user job and choose whether location, camera, accelerometer, microphone, network reach or cloud storage supplied the smallest useful evidence path.

Social and cloud integration, 2004–2015. Facebook in 2004, Twitter in 2006, Amazon Web Services (AWS) in 2006 and Azure in 2010 helped connect people to people and services. Mobile access, social identity and cloud computing created digital representations of human social behavior. Cloud application programming interfaces, or APIs, let devices store and process data remotely without each team building dedicated infrastructure.

Internet of Things, 2010–present. Everyday objects became connected to other things and to people. Sensing, reasoning and actuation made physical environments part of the service. The thermostat can now participate in a feedback loop: observations inform a later action. It inherits Internet protocols, mobile-network options and cloud platforms; its distinguishing addition is the physical sensing-and-actuation path.

Five overlapping phases add computer networking, information access, mobile access, social and cloud services, and connected things.
The five connectivity phases carry earlier infrastructure forward. IoT adds physical sensing and actuation to protocols, mobile networks and cloud services.

Read the five phases in conceptual order, noting that their dates overlap. Network, 1969–1990, connects computers through direct host-to-host communication, anchored by ARPANET in 1969. The Internet, 1991–2006, connects people to information, with the World Wide Web as its anchor. Mobile Internet, 2007–2012, adds access beyond the desktop, illustrated by the iPhone in 2007. Social + Cloud Integration, 2004–2015, connects people and services through Facebook and AWS. Cloud APIs are interfaces for using remote services. Internet of Things, 2010–present, connects things to things and people. Its added capability is autonomous M2M, or machine-to-machine communication.

16.8 What Changed Between Phases

Read the phase figure by asking which entity became connected and what capability was added. Fixed computer-to-computer networks began at a scale of dozens; desktop human-to-content access expanded to millions; mobile human-to-cloud access and social services reached billions. IoT’s machine-to-machine interactions extend to tens of billions of things. These are broad scale descriptions, not a consistent census across categories.

The timeline also includes Kevin Ashton’s naming of “IoT” in 1999 in the context of radio-frequency identification (RFID) supply chains and early machine-to-machine systems. Cisco’s 2010 estimate of 12.5B Internet-connected devices included phones and PCs, alongside early smart-home and industrial applications. It must not be relabeled as 12.5B IoT-only devices. Industry 4.0 developments around 2014 included AI/ML integration, edge computing and early 5G research. The 2020+ period brought broader attention to edge AI, digital twins and autonomous systems; “20B+ deployed” needs a dated category, such as the active-connection outlook above, rather than a claim about every year since 2020.

16.9 Stop at the Product Stage the Outcome Needs

Connectivity phases describe infrastructure. Product stages explain how a particular object uses it. A smart product uses embedded intelligence locally. A microcontroller unit, or MCU, can sense or control without a network. A connected product adds communication so another system can observe or command it. A full IoT service joins sensing, data feedback and automation so observations can improve the next action.

A local smart product gains communication, then a physical feedback loop with sensing, data and automation.
A smart product senses or controls locally. A connected product adds remote communication. An IoT service uses data feedback and automation to improve later actions.

Read left to right through Smart product, Connected product and Full IoT product. For the thermostat, each stage answers a different need. The smart product uses embedded intelligence to sense or control locally, without a network. The connected product keeps those local skills and adds communication for remote observation or commands. The full IoT product carries earlier infrastructure forward and adds sensing, actuation and data feedback. Actuation means taking physical action, which helps close the feedback path. Observations can then improve the next action through automation. Connection alone does not establish that service. The Design Rule is to stop at the stage required, since each step adds dependencies and operating cost.

Read the product figure from left to right: local capability carries forward when communication is added, and earlier networking infrastructure carries forward into the feedback loop. Stop at the stage required by the outcome. That is the practical meaning of technology convergence here: combine the capabilities the service needs, rather than treating the largest device network as the goal.

16.10 Explain the Power-Efficiency Pivot

16.11 Two Laws Behind Computing

Those product choices also depend on what small processors can do. Moore’s Law describes the long-running increase in transistors integrated on a chip. Gordon Moore’s 1965 observation began this account; the familiar approximately two-year doubling cadence is its later formulation. Transistor counts grew dramatically over more than 50 years, but counts alone do not guarantee clock speed, battery life or a useful IoT service.

Intel’s 1971 4004 had 2,300 transistors. The 1995 Pentium Pro had 5.5 million, about 2,391× that count. The 2005 Pentium D 800 had 230 million, about 100,000× the 4004 count. Apple’s 2020 M1 had 16 billion, about 6.96 million× the baseline. These comparisons span different chip classes; they illustrate integration rather than equivalent workload performance.

Try it

Explore a two-year doubling model starting from the 1971 Intel 4004 baseline.

Explore transistor doubling

The separate model starts at 2,300 transistors in 1971 and varies elapsed time from 0 to 50 years, initially 49. It applies one doubling per completed two-year interval: count = 2,300 × 2floor(years/2). This is an illustrative model, not a reconstruction of each processor’s history. Comparing it with the dated chips keeps exponential arithmetic separate from measured integration.

16.12 Dennard scaling and the mid-2000s power wall

Dennard scaling describes ideal constant-field scaling: transistor dimensions and supply voltage shrink together. Individual transistors switch faster and use less energy, allowing greater density without proportional growth in power density. Real process scaling increasingly departed from this ideal as voltage reduction slowed and leakage, interconnect and cooling constraints grew.

Across the 90 nm and 65 nm generations in the mid-2000s, power density, leakage and thermal limits constrained practical frequency growth. More transistors did not make unlimited clock increases affordable. Industry responses included multicore processors, more specialized logic and greater emphasis on performance per watt: useful processing for each unit of power.

This distinction explains why transistor doubling and processor speed cannot be treated as the same law. Integration continued to add possibilities, while energy and heat changed which designs were practical.

Power, leakage and thermal limits slow clock growth at the 90 nm and 65 nm generations, encouraging efficiency, multicore processors and specialized logic.
The mid-2000s power wall slowed clock-frequency scaling. Efficiency, multicore designs and specialized logic widened placement options alongside other IoT enablers.

Read the power-efficiency figure from the earlier reliance on higher single-core frequency through the 90 nm and 65 nm pivot. Central servers often remained economically attractive in the earlier period. Multicore designs, efficient cores and specialized logic later supported more placement options across device, edge and cloud. The change widened choices; it did not dictate that all computation move toward the device.

This is why clock growth and deployment growth need separate explanations. Earlier systems could often pursue more work through faster central processors. Later systems combined efficiency, multiple cores and specialized logic, while cheaper endpoints expanded what could be placed near a physical process. Cloud processing and human workflows still mattered. A placement decision therefore needs the service requirements as well as the component trend: integration makes more designs possible, and the intended outcome determines which is useful.

The pivot was one contributor to practical IoT. Continued integration, lower-cost microcontrollers, radios and sensors, networking, software platforms and deployment demand worked together to widen workloads near physical processes. Return to the thermostat: choose the combination that supports its service, then explain which driver made that combination practical. Neither a transistor curve nor a connection forecast supplies that explanation alone.

Check knowledge

Review connectivity phases, product stages and the limits of transistor-count comparisons.

Check convergence and compute choices

Summary

Computing spread as costs, form factors and useful functions changed. The cycle comparisons illustrate that opportunity, while dated counts and forecasts require their own definitions. A sustainable IoT service also needs recurring value, operating economics and evidence that customers benefit.

Connectivity phases built on earlier infrastructure, and product stages added communication and physical feedback to local intelligence. Integration and the mid-2000s power-efficiency pivot widened compute-placement options alongside better sensors, radios, networks and software. The smart thermostat brings those threads together: choose the capabilities needed for the outcome, then test the service case.

Explain which cost, capability and connectivity changes make the service practical. Keep schematic comparisons, measured deployments, forecasts and business outcomes distinct.

Hi, I’m Ada. What do you need?

Find help

Let’s find what you need

Find a lesson or return to your module guide.

Give feedback

Help improve this chapter

Tell us what helped or what needs fixing.

Tips

Try a small next step

A few ways to get more from this chapter.

Ask AI

Continue the conversation

Take the conversation further with your preferred assistant.

AI help is available in supported modules when enabled in Settings.

No live person answers on independent study. Feedback does not promise a personal reply.

For welfare or safety concerns, follow your school’s procedures outside IoTClass.