The First 'Smart' Connected Device Was a Coke Machine Networked by Lazy Grad Students

The Internet of Things, the sprawling network of tens of billions of connected devices that now surrounds us, had a humble and slightly ridiculous beginning. It started with a Coke machine, a few impatient grad students, and the world's first IoT hack. There was no product roadmap, no venture funding, no market research. Just thirsty computer scientists who did not want to make a wasted trip down three flights of stairs.

In 1982, in the Computer Science Department at Carnegie Mellon University, the nearest Coca-Cola machine sat in the third-floor lounge of Wean Hall. For students working late on Unix workstations and PDP-10 mainframes, the machine was a constant temptation and a constant source of frustration. Sometimes it was empty. Sometimes the bottles had been freshly restocked and had not yet chilled. Either way, you only found out after the walk.

“The first IoT device wasn't built for efficiency or profit. It was built because grad students were lazy, and that's the best kind of innovation.”

The students had something most people in 1982 did not: full access to ARPANET, the packet-switched network that the Department of Defense had funded since 1969 and that would later evolve into the internet. CMU was one of the earliest ARPANET nodes. They also had soldering irons, microcontrollers, and the kind of problem-solving instinct that considers 'I will just rewire the beverage cooler' a reasonable engineering plan.

The credit for the hack generally goes to David Nichols, Mike Kazar, John Zsarnay, and Ivor Durham. They installed microswitches inside the Coke machine to detect whether each of its six columns had bottles, and they wired a small board that monitored how long each bottle had been in the column. Bottles that had been there longer than three hours were considered properly cold. The board was connected to CMU's departmental computer, which exposed the data over the network via a simple 'finger' query. Students anywhere on ARPANET could type 'finger coke@cmua' from their terminal and see, in seconds, how many bottles were in each column and whether they were cold enough to drink.

“Every smart thermostat, every connected fridge, every wearable device owes its existence to students who didn't want to walk to an empty Coke machine.”

That was the world's first internet-connected appliance. Not a product. Not a startup. Just a research tool that solved a very specific human problem: do not walk to the machine unless the Coke is there and it is cold.

News of the hack spread through the ARPANET community. The original technical write-up, still hosted on Carnegie Mellon's Computer Science Department servers, has been cited for decades as the first working instance of what we now call the Internet of Things. The machine itself was maintained, replaced, and re-networked for years. The tradition survived long enough that a similar hack, on a different CMU Coke machine, was still operational into the 2000s.

The pattern the CMU hack established, physical object plus sensor plus network plus remote query, is exactly the pattern every IoT device uses today. The difference is scale, not structure.

The line from Wean Hall to today is remarkably continuous. In 1990, John Romkey and Simon Hackett connected a Sunbeam toaster to the internet for Interop. In 1993, Cambridge University's Trojan Room coffee pot webcam became famous as the first web-accessible hardware. In 1999, Kevin Ashton, then at Procter & Gamble and later at MIT's Auto-ID Center, coined the phrase 'Internet of Things' in a presentation about RFID-tagged supply chains. The Auto-ID Center went on to develop EPC standards that underpin modern RFID systems.

The 2000s and 2010s built the rest of the stack. Cheap microcontrollers like the Arduino (2005) and Raspberry Pi (2012) put networked computing within reach of hobbyists. Nest launched the learning thermostat in 2011. Amazon released the Echo in 2014. Philips Hue, smart locks, connected cameras, and industrial sensors followed. By the early 2020s, Cisco and IoT Analytics were estimating installed IoT devices in the tens of billions, a figure projected to pass 40 billion before the end of this decade.

The dominant motivations have scaled with the technology. Industrial IoT now tracks factory equipment, shipping containers, and farm soil moisture. Medical IoT manages glucose monitors and pacemakers. Smart cities deploy sensor networks for traffic, air quality, and lighting. Billions of dollars of infrastructure now depend on the simple primitive CMU proved in 1982: any physical thing can be a data source if you give it a sensor and a network address.

But the motivation at the root is disarmingly ordinary. The first IoT device was not built to optimize supply chains or reinvent healthcare. It was built because four grad students did not want to walk to an empty Coke machine. They saved themselves a few wasted trips per week. They also, without knowing it, drew the first sketch of an industry.

There is a lesson in that. Revolutionary platforms rarely announce themselves as revolutions. They usually show up as a clever fix for a small, slightly lazy problem, and only later do we notice that the fix contained, in miniature, the shape of the next forty years.

(Sources: Carnegie Mellon School of Computer Science archival page, 'The 'Only' Coke Machine on the Internet,' cs.cmu.edu; Teresa Carpenter, 'The Original Coke Machine,' CMU Magazine; John Romkey, 'Toast of the IoT: The 1990 Interop Toaster,' IEEE Consumer Electronics Magazine, 2017; Kevin Ashton, 'That 'Internet of Things' Thing,' RFID Journal, 2009; IoT Analytics 'State of IoT' reports, 2022-2024.)