Electronics Guide

The Age of Ubiquitous Connectivity

The decade from 2005 to 2015 reordered personal electronics more thoroughly than any comparable period since the arrival of the personal computer. The smartphone grew from a business-oriented curiosity into the primary computing device for billions of people. Cloud computing changed how software was written, deployed, and paid for. New form factors, including tablets and wearable devices, filled niches that desktops and laptops had never served well. Annual smartphone shipments passed one billion units in 2013, outnumbering feature phones for the first time, and reached roughly 1.43 billion in 2015.

The era completed trends that had been building through the digital convergence era that preceded it. Semiconductor process scaling carried mainstream logic from the 65-nanometer node to 14- and 16-nanometer FinFET parts, which put desktop-class arithmetic into a device powered by a battery the size of a playing card. Cellular networks moved to all-IP designs with an order-of-magnitude gain in throughput. Projected capacitive touch sensing, refined over years of research, finally reached the responsiveness and cost that mainstream products required. No single one of these advances produced the smartphone. Their convergence did.

The period also widened access to computing and information. A capable computer became affordable to anyone who could carry a modest monthly service payment, and in much of the world the phone was the first computer its owner had ever used. Cloud services removed the need for local infrastructure, so a small team could serve millions of users without buying a rack. Social platforms connected billions of people and, in doing so, reshaped how information circulated and how attention was bought and sold. Electronics stopped being one industry among many and became infrastructure on which other industries depended.

Articles in This Category

The Smartphone Revolution

Apple announced the iPhone in January 2007 and shipped it that June. The combination mattered more than any single feature: a capacitive multi-touch display with no stylus and no physical keyboard, a browser that rendered ordinary desktop pages, and a hardware and software stack designed together. The App Store followed in July 2008 with roughly five hundred applications, and it converted the phone from a finished product into a platform. Google and the Open Handset Alliance announced Android in November 2007, and the first Android handset, the T-Mobile G1 (also sold as the HTC Dream), reached buyers in October 2008. Android's open-source licensing let dozens of manufacturers build on a common platform, and the resulting duopoly drove a decade of rapid iteration in silicon, cameras, displays, and services.

Smartphones then absorbed the functions of a shelf full of single-purpose devices. Compact cameras, portable music players, satellite navigation units, voice recorders, camcorders, and handheld game systems each lost their reason to exist as separate purchases. Camera quality drove much of that displacement: backside-illuminated CMOS image sensors reached handsets around 2010, dedicated image signal processors handled noise reduction and exposure fusion, and multi-frame computational techniques compensated for the small optics that a thin body allows. Micro-electromechanical sensors did the same for motion. The first iPhone carried an accelerometer, the iPhone 3GS of 2009 added a magnetometer, and the iPhone 4 of 2010 added a three-axis gyroscope, giving applications a continuous sense of orientation and movement that no desktop computer had.

The consolidation was ruinous for incumbents that misread it. Symbian led the smartphone market until Android passed it around the end of 2010. BlackBerry's subscriber base peaked near eighty million in 2012 and then fell away as corporate buyers accepted consumer devices on their networks. Microsoft agreed in September 2013 to acquire Nokia's devices and services business for about 5.44 billion euros, roughly 7.2 billion dollars, and closed the deal in April 2014; the Windows Phone platform never achieved competitive scale, and Microsoft wrote the phone business down in 2015. Component makers, contract manufacturers, and carriers all reorganized around the two surviving platforms, and the industry learned that owning a hardware franchise counts for little without an ecosystem of developers behind it.

Silicon, Power, and Packaging

The binding constraint on a smartphone was never transistor count. It was energy. Lithium-ion cells improved by only a few percent per year, and a pocket device dissipating more than two or three watts becomes uncomfortable to hold, so every capability had to be paid for out of a nearly fixed budget. The response was integration. A mobile system on chip drew together application processors, a graphics unit, an image signal processor, video encode and decode blocks, a display pipeline, and security hardware on one die, with the DRAM stacked directly above it in a package-on-package assembly that saved board area and shortened the highest-frequency traces.

Arm's licensing model supplied the instruction set and, for many vendors, the cores as well, while Apple, Qualcomm, and Samsung invested in their own microarchitectures. Apple's A7, introduced in the iPhone 5s in September 2013, brought the 64-bit Armv8-A architecture to phones ahead of the rest of the industry and forced competitors to accelerate their own transitions. Arm had announced big.LITTLE in 2011, pairing high-performance cores with high-efficiency cores on one die so that background work never woke the expensive hardware; the same heterogeneous approach later migrated into laptop and desktop processors. Fabrication kept pace: FinFET transistors entered volume production at Intel's 22-nanometer node in 2012 and reached mobile silicon with the 14- and 16-nanometer generations of 2015, restoring the electrostatic control that planar devices were losing.

Scaling nevertheless stopped delivering proportional power savings, and designers confronted what became known as dark silicon: not every block on a die can switch at once within the thermal budget. Fixed-function accelerators, aggressive clock gating and power gating, and multiple voltage and frequency domains became standard practice, because a hardwired video decoder consumes a fraction of the energy that a general-purpose core would need for the same work. The rest of the system received similar attention. Displays and their backlights dominated the power budget, which favored active-matrix OLED panels that emit only where the image is bright. Power management integrated circuits grew to dozens of regulated rails. Envelope tracking, which reached handsets around 2013, varied the supply voltage of radio-frequency power amplifiers in step with the signal envelope and cut their waste heat substantially. Thermal design became a matter of spreading heat through the chassis, since no phone has a fan.

Networks That Made Mobility Possible

A pocket computer is only as useful as the link behind it. Third-generation networks were already deployed at the start of the era, and successive High Speed Packet Access releases raised their throughput through the late 2000s, but the decisive change was Long Term Evolution. TeliaSonera opened the first commercial LTE networks in Stockholm and Oslo on December 14, 2009, and Verizon launched LTE in the United States a year later. LTE abandoned the circuit-switched inheritance of earlier generations for an all-IP core, used orthogonal frequency-division multiple access with multiple-input multiple-output antennas, and cut latency enough to make interactive applications usable over cellular. LTE-Advanced, specified in 3GPP Release 10 in 2011, added carrier aggregation so that operators could bond fragmented spectrum holdings into wider effective channels.

Wi-Fi carried the larger share of the traffic. The 802.11n amendment, ratified in 2009, brought MIMO and channel bonding to consumer equipment, and 802.11ac, completed in 2013, added wider channels and multi-user MIMO in the 5-gigahertz band. Operators came to depend on this offload, because the economics of licensed spectrum could not absorb the growth in video alone. Networks densified with small cells and distributed antenna systems, backhaul migrated from copper to fiber, and spectrum auctions became major fiscal events. Tariffs shifted with the traffic: unlimited data plans gave way to metered tiers as data displaced voice and messaging as the operators' principal revenue.

Short-range radio changed as well. Bluetooth Low Energy, introduced as part of the Bluetooth 4.0 specification in 2010, gave sensors, beacons, and accessories a duty-cycled radio that could run for months on a coin cell. That single change made fitness trackers, wireless heart-rate straps, and smartwatch companions practical, and it supplied the connectivity layer on which the following decade's consumer Internet of Things was built.

Cloud Computing Emergence

Amazon Web Services demonstrated that computing infrastructure could be rented by the hour rather than purchased by the rack. Simple Storage Service launched in March 2006, Elastic Compute Cloud entered limited public beta that August, and EC2 dropped its beta label in October 2008. The consequences were larger than the idea appeared. A startup could serve millions of users without owning a server. An enterprise could provision for actual demand instead of anticipated peaks. Capacity became an operating expense drawn from a business unit's budget rather than a capital expense negotiated a year in advance. Google previewed App Engine in 2008, and Microsoft announced Azure the same year and opened it commercially in February 2010.

Virtualization made the model possible. Hypervisors partitioned a physical machine into isolated virtual ones, allowing many tenants to share hardware safely; Xen underpinned the early EC2 fleet, KVM merged into the Linux kernel in 2007, and VMware dominated the enterprise data center. Later in the era the unit of deployment shrank again. Docker, released in 2013, packaged applications into containers that started in milliseconds and shared a host kernel, and Kubernetes, announced by Google in 2014 and reaching version 1.0 in 2015, supplied the scheduler and control plane needed to run containers across large fleets. OpenStack, launched in 2010, offered the same primitives to organizations that wanted a private cloud.

Data centers became an engineering discipline in their own right. The Green Grid published the power usage effectiveness metric in 2007, and hyperscale operators drove annualized values close to 1.1 through free cooling, higher inlet temperatures, and careful containment of hot and cold aisles. Facebook opened the Open Compute Project in 2011 and published server, rack, and power distribution designs that stripped out components a fleet operator does not need. Content delivery networks pushed static assets and video to the network edge, and providers organized capacity into regions and availability zones so that a single failure domain would not take down a service. Netflix, which began streaming in January 2007, started moving its infrastructure to AWS in 2008 and became the standard example of a business rebuilt on rented capacity.

The model's advantages came with a structural cost. Workloads concentrated in a handful of providers and a handful of large facilities, so a regional outage or a misconfigured control plane could disrupt thousands of unrelated services at once. Data residency, vendor lock-in, and the difficulty of estimating a bill that varies with traffic became standing concerns for anyone deploying at scale.

Social Media and Web 2.0

Platforms built on user-generated content matured into the dominant channels for information. YouTube, founded in 2005, was acquired by Google in 2006 for about 1.65 billion dollars in stock and made video publishing free to anyone with a camera. Facebook opened registration to the general public in September 2006, listed publicly in May 2012, and announced one billion monthly users in October 2012. Twitter, launched in 2006, established short public messages as a real-time medium for news and coordination. Instagram, released for the iPhone in October 2010, was acquired by Facebook in 2012 for roughly one billion dollars, and WhatsApp followed in 2014 for approximately nineteen billion dollars. Each acquisition confirmed that attention on mobile devices had become the scarce asset.

Serving that traffic forced advances in systems engineering. The H.264 codec and adaptive bitrate streaming made video practical over links of unpredictable quality. Relational databases were sharded past the point of comfort and then supplemented with distributed stores that traded strict consistency for availability and horizontal scale, including the Hadoop ecosystem derived from Google's published MapReduce and file system designs. Caching tiers, asynchronous work queues, and ranked feeds computed from behavioral signals became standard architecture. The advertising model that funded these services also determined their engineering priorities, since the systems were optimized for engagement and targeting rather than for the user's stated preference.

Privacy moved from a specialist concern to a public one. Behavioral tracking across sites, location histories, and social graph data accumulated faster than any regulatory framework could address. The disclosures that began in June 2013 about government access to communications pushed the industry toward encryption by default: major services adopted HTTPS across their properties, transport security improved, and Let's Encrypt entered public beta in 2015 to make certificates free and automatic. The debates opened in this decade over content moderation, algorithmic ranking, and platform accountability remain unresolved.

Mobile Commerce and Financial Inclusion

Phones became trusted instruments for moving money. Safaricom launched M-Pesa in Kenya in March 2007, using a SIM toolkit application that ran on ordinary feature phones and a network of small retail agents who handled cash in and cash out. The service required no smartphone, no bank account, and no reliable data connection, and it reached roughly twenty-five million users in Kenya by 2015, with annual transaction values amounting to a substantial fraction of national output. Variants spread across sub-Saharan Africa and South Asia and brought remittances, wages, and savings within reach of populations that formal banking had never served. The lesson was that the constraint had been distribution rather than technology.

In wealthier markets the path ran through contactless payment hardware. Google Wallet arrived in 2011, and Apple Pay launched in the United States in October 2014 built on near-field communication, a hardware secure element, and tokenization, so that a merchant received a device-specific token rather than the card number itself. Android 4.4 introduced host card emulation in 2013, allowing credentials to be held in the cloud instead of in a carrier-controlled secure element. Biometric authentication made the transaction fast enough to be worth the effort: Apple's Touch ID, introduced with the iPhone 5s in September 2013, paired a capacitive fingerprint sensor with an isolated coprocessor that kept the template off the main application processor. Hardware-backed key storage, of which Arm TrustZone is the most widely deployed example, gave applications somewhere to put secrets that a compromised operating system could not read.

Tablets, Wearables, and New Form Factors

The decade produced one durable new category and several instructive failures. Netbooks arrived first, beginning with the Asus Eee PC in 2007, and sold in volume on price alone before cramped keyboards, weak processors, and better ultraportables ended the boom. Amazon's Kindle, launched in November 2007, made a stronger case for a specialized device: an electrophoretic display consumes power only when the image changes, so an e-reader runs for weeks, remains legible in direct sunlight, and delivered books over a cellular link at no charge to the reader.

Apple announced the iPad in January 2010 and shipped it that April, and it established the tablet as a real category rather than the perennial failed experiment it had been. Android tablets followed, and the market grew quickly to a peak of roughly 230 million units shipped in 2014 before declining. Two forces pressed on it from either side. Phones grew larger, a trend the 5.3-inch Samsung Galaxy Note of 2011 initiated against widespread skepticism, and tablets proved to have replacement cycles closer to those of laptops than of phones. The category settled into education, point-of-sale, aviation and field service, and media consumption rather than displacing the personal computer outright.

Wearables took longer to find their footing. Fitbit, founded in 2007, shipped its first clip-on tracker in 2009; Pebble raised more than ten million dollars on Kickstarter in 2012 and demonstrated genuine demand for a wrist-worn display; Google announced Android Wear in March 2014; and the Apple Watch reached buyers in April 2015. The engineering recipe was consistent: MEMS accelerometers for activity, optical photoplethysmography for heart rate, Bluetooth Low Energy for the link to a phone, an always-on low-power sensor hub so the main processor could sleep, and a battery measured in a few hundred milliamp-hours. Health monitoring pushed these products toward the regulatory boundary between wellness accessories and medical devices, a line that later generations would cross deliberately.

Two experiments defined the era's limits. Google Glass shipped to developers as the Explorer Edition in 2013 and put a small display and camera on the face; Google halted consumer sales in January 2015 after social objections to always-available recording and the absence of a compelling everyday use. Oculus, funded by a Kickstarter campaign in 2012, was acquired by Facebook in 2014 for about two billion dollars, though consumer headsets did not ship until the following era. Both efforts showed that a wearable must justify its social cost as well as its power budget.

A Pivotal Decade

The platforms established, ecosystems created, and habits formed between 2005 and 2015 still govern the electronics industry. Two mobile operating systems, a handful of cloud providers, and a small number of leading-edge foundries carry an outsized share of the world's computing, and that concentration shapes everything from component sourcing to trade policy. The engineering principles the decade settled on proved equally durable: integrate aggressively, treat energy rather than area as the scarce resource, specialize the hardware for the work that matters, and move whatever can tolerate the latency to a remote data center.

The unresolved questions carried forward as well. Privacy, moderation, and the economics of attention were all posed in this decade and answered only partially. Electronic waste grew with a replacement cycle measured in two or three years. Security obligations began to extend across product lifetimes that no consumer electronics industry had previously had to support. The connected sensors, on-device inference, and always-available assistants of the Internet of Things and AI era that followed are direct descendants of the mobile silicon, the low-power radios, and the rented data centers assembled here.