In November 1982, on the floor of Comdex in Las Vegas, a booth few expected anyone to notice began drawing a crowd. The machine was a Victor 9000 with a screen of 800 by 400 pixels, and on it ran a technical drawing program, the kind an engineer or architect would use, with precise lines snapping to precise coordinates. The program was called AutoCAD. The price tag read $1,000.

Fifty feet in any direction, the established names of computer-aided design held court: Computervision, Intergraph, CADAM. Their systems demanded dedicated graphics terminals, banks of computers, climate-controlled rooms, and trained operators. A single CAD workstation cost a quarter of a million dollars. AutoCAD cost $1,000, ran on a personal computer that fit on a desk, and delivered roughly 80 percent of what those quarter-million-dollar systems could do.

The judges at Comdex named it one of the hits of the show. The first revenue copy sold to Jamal Munchie of Mom’s Computing in Solano, California. The company behind AutoCAD had no venture capital, no corporate parent, no board with industry pedigree. Sixteen programmers had pooled $59,000 in a living room in Mill Valley, California.

Twelve thousand lines of source code was all they brought to Las Vegas. That $1,000 program would grow into a company valued at $67 billion at its peak. Its proprietary file format would be embedded in over two billion engineering drawings worldwide. Virtually every building, bridge, highway, interchange, and airport terminal designed in the last four decades runs through software this company built or acquired.

The company was called Autodesk, and the story of how it traveled from that Comdex booth to the invisible infrastructure of the built world is one of the strangest survival stories in American technology. To understand how sixteen programmers in a living room ended up designing the modern world, you have to go back five years before that Comdex booth, to a man most people in technology have never heard of. In 1977, a programmer named Mike Riddle was working on a Computervision CAD system at Marathon Steel in Arizona, detailing structural steel for the Palo Verde nuclear power plant near Phoenix. He knew exactly what computer-aided design could do.

He also knew exactly what it cost. A single Computervision system ran a quarter of a million dollars and required a dedicated graphics terminal, a minicomputer the size of a filing cabinet, and an operator trained on that specific hardware. This was not a tool for a twenty-person engineering firm in a strip mall. This was a tool for Boeing, for General Motors, for companies large enough to write a check for $250,000 and call it infrastructure.

Riddle saw the gap between what CAD could do and who could afford it, and he decided to close it himself. In his spare time, he began writing a program he called Interact on an S-100 microcomputer, a hobbyist machine that could sit on a workbench. The processor was a Texas Instruments TMS9900 mounted on a main board built by a small California hardware company called Marinchip Systems. The constraints were brutal.

Memory arrived in small boards, and Riddle had to write his code in stages as larger capacity became available, building the program the way you build a house when the lumber arrives one truckload at a time. His first commercial customer was not a hobbyist. It was Atlantic Richfield, one of the largest oil companies in America, which used Interact to plan deep-water dives on offshore oil rigs. The IBM PC did not yet exist.

Apple had just shipped the Apple II, and Mike Riddle had already built the first CAD system to run on mainstream microcomputer hardware. Almost nobody in the industry knew it existed. Meanwhile, on the other side of the emerging personal computer world, a man named John Walker was building the hardware that Riddle’s program ran on. Walker was born in 1949 in Baltimore, Maryland.

His family expected him to study medicine. He chose electrical engineering instead, earned his degree from Case Western Reserve University, then migrated to California and fell into the orbit of the microcomputer revolution as it was just beginning to gather speed. With a partner named Dan Drake, Walker co-founded Marinchip Systems. Walker was a programmer who happened to run a hardware company.

When he saw Interact running on a Marinchip board, he understood immediately what it meant. This was professional engineering design running on a personal computer, a real tool for real work. He became a software dealer for Riddle’s program, but selling someone else’s software was not what Walker had in mind for long. In early 1982, Walker gathered Dan Drake, Mike Riddle, and roughly a dozen other programmers at his home in Mill Valley, California.

The plan he laid out that evening was not to start a CAD company. Walker proposed a cooperative of programmers: pool money, form a company, and collectively develop and market whatever promising software the group could produce. Not a single product, but a portfolio. Each founder contributed a minimum of $3,000.

Sixteen of them pulled $59,000 total. They originally called themselves Marin Software Partners. On April 26, 1982, they incorporated under a new name: Autodesk. Among the group’s various software projects, the strongest candidate was obvious.

Riddle’s Interact was a working, proven CAD program with a paying customer in the oil industry. They rewrote it in C for the IBM PC. On March 19, 1982, they showed it at the Sixth West Coast Computer Faire under the name MicroCAD. Another company claimed that name almost immediately, so they renamed it AutoCAD-80 and reserved a $1,200 booth at Comdex.

The first fiscal year, ending January 1983, produced revenue of $14,733 and a loss of $9,465. It would be the only unprofitable year in the company’s entire history. By April 1983, four hundred copies had sold. By the end of the calendar year, the count passed one thousand, grossing more than one million dollars.

The pricing was deliberate: $1,000 for AutoCAD, $500 for the AD-1 dimensioning add-on package. These were not small sums for a two-person surveying firm or a young architect starting a practice. But set against the quarter-million-dollar systems they replaced, they were a fraction so small it redrew the boundaries of who could afford professional design. Autodesk was not selling software.

Autodesk was demolishing a price barrier that had kept computer-aided design locked inside corporations large enough to build a room for it. Tied inside that pricing structure and ticking was a knot that would take a decade to cut. Mike Riddle had licensed Interact to Autodesk under a non-exclusive agreement carrying a 10 percent royalty on AutoCAD sales. By the fiscal year ending January 1985, Autodesk was paying Riddle nearly $600,000 annually in royalties.

But the company had structured its dimensioning add-ons as separate products rather than enhancements to AutoCAD, which reduced the royalty base. Riddle noticed. In peak years, he earned over one million dollars, but he also watched the company reclassify features in ways that minimized what it owed him. In August 1991, Riddle and his wife filed suit.

In early 1992, they settled for $11. 9 million, a lump sum waiving all future royalties. Most of his original Interact commands still function in the current version of AutoCAD more than four decades later. The foundation was his.

The monument would belong to someone else. By fiscal 1985, the trajectory was unmistakable: 25,000 units sold, $27 million in revenue, $6 million in profit. In June 1985, Autodesk went public on the NASDAQ, selling one million shares at $11 each and raising just over $10 million. Not one dollar of venture capital had ever entered the company.

For the original stockholders who had contributed $3,000 each three years earlier, the IPO delivered $165 for every dollar invested. The stock rose to $21 by year’s end. Then the company stumbled into a lesson it would not forget. AutoCAD version 2.

5, released in 1986, shipped with a hardware lock, a physical device the size of a cigarette pack that plugged into the COM1 serial port. Without the lock connected, the software refused to run. Users revolted. Third-party companies began selling bypass tools.

Autodesk pulled the lock before the end of 1986, distributing AutoCAD version 2. 52 without it to every registered customer, and became a founding member of the Business Software Alliance antipiracy campaign instead. The lesson was permanent: treating customers as thieves is the fastest way to teach them to find someone who does not. In mid-1986, with 255 employees and over $40 million in annual sales, John Walker resigned as chairman and president.

He did not leave. He demoted himself to programmer. But a company doubling its revenue every year could not run on collaborative anarchy forever. Alvar Green, a manager whose background was in business rather than programming, became CEO.

Walker stayed on, writing software and watching the company he had founded evolve into something he recognized less with each passing quarter. The real pivot was a product decision that would define everything that came after. In May 1985, AutoCAD version 2. 1 shipped with the program’s first three-dimensional capabilities, expanding the code base from 12,000 lines at launch to over 100,000 lines of C.

Embedded inside that version was something even more consequential: AutoLISP, a programming language built directly into AutoCAD. AutoLISP gave users direct access to AutoCAD’s drawing database. Every architect, civil engineer, and surveyor could write custom commands, automate repetitive workflows, and build specialized tools on top of AutoCAD’s foundation. AutoCAD stopped being a drawing program.

It became a platform. The ecosystem detonated. By 1985, over 200 third-party applications were under development. By 1987, 400 vendor products filled the Autodesk applications catalog.

By 1988, the count exceeded 700 applications. AutoCAD was not the most technically advanced CAD program on the market, but it had become the de facto standard. The logic was self-reinforcing: the more applications that ran on AutoCAD, the more customers chose AutoCAD, and the more customers who chose it, the more developers built for it. Fiscal 1987 brought $54 million in revenue.

Fiscal 1988, $79 million. By 1989, Autodesk held 60 percent of the personal computer design software market. Fiscal 1990 brought $179 million, fiscal 1991 $238 million, with software sold in 60 countries and published in 12 languages. And in fiscal 1991, AutoCAD represented 88 percent of Autodesk’s total revenue.

One product, one file format, one way of drawing. The strength and the vulnerability were the same figure. The forces gathering outside the company were about to make that terrifyingly clear. The same collapse in the price of computing power that had allowed Autodesk to put CAD on a desktop was now making far more powerful design tools possible on cheaper hardware.

Parametric Technology Corporation shipped Pro/ENGINEER. SolidWorks followed. These were not drafting programs. They were programs that understood geometry as objects, programs that could update every dimension in a design when a single parameter changed.

They made AutoCAD look like a very fast pencil. At the same time, DOS was giving way to Windows, and AutoCAD’s command-line heritage was showing its age. In April 1991, Walker wrote Information Letter 14, a 44-page internal memorandum that read like an autopsy performed on a patient still breathing. AutoCAD was under-resourced for both development and marketing.

The user interface was not competitive. Windows 3. 0 deserved far more attention. The product, Walker wrote, was becoming dangerously antiquated.

The market agreed. Autodesk’s stock hit $61 per share in June 1991. Eight months later, amid deepening uncertainty about the company’s direction, the stock fell to $23. 50.

In April 1992, the board brought in a professional. Carol Bartz arrived at Autodesk on April 14, 1992, as president, chairman, and chief executive officer. She held a computer science degree from the University of Wisconsin and had served as vice president at Sun Microsystems. She was one of only two women leading a major American high-technology company at the time.

The engineers who had built Autodesk on programmer culture viewed professional management with suspicion. Bartz did not flinch. She set three goals: build a billion-dollar company by 1999, reduce the dangerous dependence on AutoCAD revenue, and expand into new markets starting with manufacturing and entertainment. Within her first year, she imposed traditional management structure on the programmer anarchy: five distinct business units, profit-and-loss accountability, clear reporting lines.

Revenue under Carol Bartz grew from $300 million to $1. 5 billion. She overshot her own target by half a billion dollars. By 1998, an estimated two billion DWG files existed in the world.

Not two billion copies of AutoCAD, but two billion individual engineering drawings stored in Autodesk’s proprietary format, referenced by other drawings, embedded in project archives, filed with municipalities. The DWG had become the closest thing the built world had to a universal language. AutoCAD was never the fastest CAD program, nor the most elegant, nor the most technically advanced at any given moment. What it was from the day it shipped was the most available.

The DWG file format guaranteed backward compatibility. A drawing made in version one could be opened and modified decades later. Firms built libraries of drawings over years, over entire careers, and every drawing bound them deeper into the ecosystem. The switching cost was never the price of a new license.

It was the weight of every drawing a firm had ever made. The second product that carries the weight of Autodesk’s ambition was never built inside Autodesk at all. In 1990, a developer named Gary Yost and a programmer named Tom Hudson delivered 3D Studio for DOS, published by Autodesk at a list price of $2,995. Four DOS versions shipped between 1990 and 1994.

Then the Yost Group rewrote the entire application for Windows NT. The result, codenamed Jaguar, shipped in April 1996 as 3D Studio MAX through Autodesk’s Kinetix division. 3D Studio MAX put film-quality 3D rendering on a desktop PC: Titanic in 1997, Avatar in 2009, Inception, Life of Pi, Hugo, Iron Man 2. For sixteen consecutive years, every Academy Award-winning visual effects film used Autodesk tools.

When Autodesk acquired Discreet Logic, a Montreal-based visual effects firm, for $410 million in 1999, and then Alias Systems Corporation, the company behind Maya, for $182 million in October 2005, the position was cemented. Autodesk was no longer just the company that helped architects draw floor plans. It was the company that visualized the built world and the imagined one. Here, at the summit of a golden age that had lasted more than a decade, the seed of the next crisis was already visible.

The business model that had made Autodesk a billion-dollar company, selling perpetual software licenses through a network of authorized dealers, was a model built for a world where software shipped in boxes. The world was moving toward subscriptions, toward the cloud, toward software delivered as a continuous service. For fourteen years, Carol Bartz had assembled the machine. On May 1, 2006, she handed it to Carl Bass, a man who wanted to take it apart and rebuild it in a different shape.

Bass had co-founded Ithaca Software in 1986; Autodesk acquired it in 1993, and Bass joined the company as chief architect for AutoCAD. In 1995, Carol Bartz forced him out. Five months later, Autodesk’s engineers demanded his return. He left again in 1999 to found Buzzsaw, an online construction project management service, which Autodesk acquired in 2001.

Bass rose to chief operating officer by 2004 and took the top job in 2006. He pushed Autodesk into the maker movement, into 3D printing, into consumer design tools. He launched Fusion 360, a cloud-based CAD platform, in September 2013. In January 2016, Autodesk ended perpetual licenses entirely, subscription only.

The strategic logic was sound, but the execution was brutal. Fiscal year 2017 revenue dropped 18. 9 percent, from $2. 5 billion to $2.

03 billion. The stock cratered. Customers who had paid $3,750 for a perpetual AutoCAD license were now paying annual subscriptions for software they rented. Bass resigned on February 8, 2017, with the subscription transition only half finished.

The person who walked through the rest of the fire was Andrew Anagnost. He had arrived at Autodesk as a product manager in 1997 and rose through every layer of the organization. He was the internal architect of the subscription transition that Bass had set in motion. When Bass resigned, Anagnost served as interim CEO before being permanently appointed president and chief executive officer in June 2017.

He finished what Bass had started. The revenue recovery reads like a textbook case study in surviving a self-inflicted wound: fiscal 2018, $2. 06 billion, then $2. 57 billion, $3.

27 billion, $3. 79 billion, $4. 39 billion, $5. 01 billion, $5.

5 billion, $6. 13 billion, and fiscal year 2026, $7. 21 billion. The subscription model that had cratered revenue by 19 percent now powered annual growth of 17 percent.

The market capitalization peaked at $67. 14 billion on December 31, 2020. Sixteen programmers had pooled $59,000. Forty years later, the company was worth more than the entire mainframe CAD industry had been in the year AutoCAD shipped.

Anagnost poured the revenue into expansion. Revit Technologies, acquired back in 2002 for $133 million, had already transformed architecture by introducing building information modeling, designing a building not as a set of flat drawings but as a three-dimensional database of every wall, pipe, wire, and window. The acquisitions accelerated: PlanGrid for $875 million in 2018, BuildingConnected for $275 million, Innovyze for $1. 6 billion in 2026.

Autodesk entered the Fortune 500. The workforce reached 14,300 employees. John Walker died on February 2, 2024, in Neuchâtel, Switzerland. He was 74 years old, the cause injuries from a fall at home.

He had moved to Switzerland in 1991. He left behind Fourmilab, a personal website filled with the kinds of projects only a restless polymath builds: a hardware random number generator, an interactive planetarium, a book called The Hacker’s Diet. He also left behind the Autodesk File, a meticulously honest internal history of the company’s first decade, published because Walker believed the documented record was always more interesting than the polished legend. When he departed Autodesk in 1994, the company ranked as the sixth largest personal computer software company in the world.

Mike Riddle settled for $11. 9 million in 1992 and moved on. He went on to write EasyCAD and FastCAD, programs that won technical benchmarks and lost the market. Most of his original Interact commands still function in modern AutoCAD four decades and 31 major versions later.

Return to that Comdex booth. November 1982. A Victor 9000, 800 by 400 pixels. A program called AutoCAD.

A price tag that reads $1,000. In the next aisle, Computervision is demonstrating the same capability for a quarter of a million dollars. One of these products will become the invisible infrastructure of the built world. The other will be acquired, merged, and dissolved until not even its name survives in common use.

Sixteen programmers pooled $59,000, carrying no venture capital, no board of directors with Fortune 500 resumes, and no five-year strategic plan, just a single insight planted in 12,000 lines of code: that professional design did not need to cost a quarter of a million, and that the personal computer was about to prove it. That $1,000 program now generates over $7 billion a year. Its file format is the silent standard underlying the physical infrastructure of the modern world. They did not invent the blueprint.

They killed the drafting table.