The Night They Faked a Computer to Launch an Empire

The teletype clattered in the quiet MITS office, its paper tape reader chewing through the thin strip with a sound like dry leaves. Paul Allen sat at the console, fingers hovering over the keys. Ed Roberts, the tall Texan who ran Micro Instrumentation and Telemetry Systems, watched from the doorway. This was the moment. Allen typed PRINT 2+2 and hit return. The machine paused, lights flickering on its front panel, then spat out a single digit: 4. It worked. The interpreter they had written for a computer they had never touched had just passed its first test on the real hardware.

The Bluff That Started It All

The story began five weeks earlier, in December 1974, when the January issue of Popular Electronics hit newsstands. Its cover featured the Altair 8800, a kit computer built around Intel's new 8080 microprocessor. To most people it looked like a box of switches and blinking lights. To Bill Gates and Paul Allen, it looked like the future arriving ahead of schedule.

They were 19 and 21, respectively. Gates was a Harvard sophomore who spent more time in the computer center than in class. Allen had already dropped out to work at Honeywell in Boston. They had been programming together since high school at Seattle's Lakeside School, where a Teletype terminal connected to a GE time-sharing mainframe had given them a head start most professionals lacked. They knew the Altair needed software — specifically a BASIC interpreter — to be useful to hobbyists. They also knew MITS didn't have one.

So they called Ed Roberts and told him they did.

It was, by any reasonable measure, a fabrication. They had no Altair. They had no 8080 chip. They had never written a BASIC interpreter for a microcomputer. But Roberts, eager to make his machine competitive, invited them to Albuquerque for a demonstration in March. The deadline was immovable. Gates later called it "extreme confidence in their ability to deliver." Allen called it "a slight exaggeration."

Writing Code for a Ghost Machine

Back in Cambridge, the pair turned Harvard's PDP-10 mainframe into a surrogate Altair. Allen wrote a software simulator that mimicked the 8080's instruction set, cycle by cycle. Gates then wrote the BASIC interpreter on top of that simulator, line by line, optimizing furiously to squeeze it into the Altair's 4KB of memory. They worked around the clock, fueled by pizza and the particular adrenaline of a bluff that had grown too large to walk back.

The constraints were brutal. Every byte mattered. They stripped comments, reused code fragments, and exploited quirks of the 8080 architecture they had only read about in manuals. Gates handled the high-level logic; Allen focused on the interface between the interpreter and the simulated hardware. They debugged by staring at octal dumps, imagining how the real machine would respond.

When the code was finally complete, they punched it onto paper tape — the only medium the Altair could read. Allen boarded a flight to New Mexico with the tape in his carry-on. Gates stayed behind, unable to face the possibility of failure. "I was pretty nervous," Allen recalled later. "If it didn't work, we'd have to explain why we'd wasted his time."

The Demo That Changed the Industry

The MITS office smelled of solder and ozone. Roberts had an Altair 8800 set up on a table, its front panel a constellation of red LEDs. Allen loaded the tape. The reader clicked and whirred. Then the prompt appeared: a simple dot. He typed the test case. The machine answered.

Roberts didn't cheer. He nodded, the way a man nods when a bet pays off. "We'll take it," he said.

The deal they signed gave MITS a license to distribute Altair BASIC. Microsoft — still hyphenated as Micro-Soft — would receive royalties on each copy sold. It wasn't much money, but it was the first time anyone had paid them for software as a product rather than a service. The idea that code itself had commercial value, independent of the machine it ran on, was still radical. Most hobbyists believed software should be free, passed hand to hand on paper tape and cassette. Gates would famously rail against this in an open letter the following year, but the principle was already established in that Albuquerque room.

From Albuquerque to the World

The Altair BASIC deal forced a choice. Gates left Harvard. Allen quit Honeywell. They moved to Albuquerque, setting up shop in a cramped office near the MITS factory. The team grew: a few programmers, a secretary, a bookkeeper. They ported BASIC to other processors — the 6502, the Z80 — and wrote FORTRAN and COBOL compilers. They fought with Roberts over exclusivity, insisting on the right to license their languages to other hardware makers. That insistence, born of a conviction that software would outlast any single machine, became the architectural principle of their business.

By 1979, the microcomputer market had outgrown Albuquerque. The talent pool was in Seattle. They moved home to Bellevue, a suburb across the lake from their high school. The company dropped the hyphen. Within two years, a call from IBM would turn their BASIC business into an operating system empire. But that empire rested on the foundation laid in a month of sleepless nights in Cambridge, and a single keystroke in a New Mexico office that returned the number 4.

Allen died in 2018. Gates stepped back from daily operations years earlier. The Altair 8800 sits in museums now, a curiosity from the dawn of personal computing. But the logic of that first deal — software as the product, hardware as the platform — still governs the industry they built. Every app store, every cloud service, every subscription license traces its lineage to the moment two young men bet everything on code they hadn't written yet, for a machine they'd never seen, and won.

This is one episode in a much longer story. For the full account of the founding of Microsoft, read “Microsoft” by Arthur Sanchez on MixCache.com.