TimelineAltair 8800
Altair 8800
A $439 box of switches that started the personal computer industry
The Altair 8800 arrived as a mail-order kit from a small Albuquerque firm with a metal case, a grid of toggle switches and a row of red lights. It had no keyboard, no screen and 256 bytes of memory. It also produced more orders in one month than the company had expected to fill in a year, and it pulled a programming language, a bus standard and a software company into existence around it.
A cover story before there was a computer
By 1974 Micro Instrumentation and Telemetry Systems was a small electronics firm in Albuquerque, New Mexico, with a broken business. MITS had started in 1969 selling instruments for model rocket hobbyists and moved on to calculator kits, until Texas Instruments began selling complete calculators for less than half the price of anyone else's. Ed Roberts, the company's founder, was carrying a quarter-million-dollar debt and needed a product nobody else was making.
His answer was a personal computer kit built around the Intel 8080, a microprocessor introduced in April 1974 at a list price of $360. Roberts negotiated the 8080 down to $75 in bulk and began development in the summer of 1974. The machine was not finished when Popular Electronics agreed to put it on the cover of its January 1975 issue. The magazine photographed an empty case with switches and lights on the front panel, and editor Les Solomon chose the name, by one account after his twelve-year-old daughter suggested Altair, by another because the staff wanted to name it after a star. The issue reached newsstands on 19 December 1974, the date the kit became officially available for sale.
What $439 bought
The kit sold for $439, or $621 fully assembled. Inside the two-piece case was a backplane of 100-pin edge connectors with the basic machine spread across five cards, one holding the 8080 running at 2 MHz and another holding memory. The standard memory board was sold as a "1024 word" card but shipped populated with 256 bytes, not kilobytes. The front panel, modelled on the Data General Nova minicomputer, carried toggle switches and red LEDs.
That panel was the whole user interface. Programming meant setting switches to the binary pattern of an instruction, pressing DEPOSIT NEXT to write it into the next memory address, and repeating that for every opcode. The output was the pattern of lights. Anything more useful required extra cards bought separately: a 4K memory board at $264 in kit form, a serial interface at $124, and a Model 33 ASR Teletype at $1,500.
The bus that outlived the machine
Roberts had originally planned a machine whose parts would not fit on one board. The redesign made the motherboard an interconnect between plug-in cards, joined by the 100-pin edge connector he could buy cheaply. The result was a bus that carried the pins of the Intel 8080 out onto the backplane, unrefined in places, with unidirectional data buses and power lines of differing voltages side by side.
Crude or not, it worked and it was public: when rival firms built their own computers in 1976, most simply copied the Altair card interface so that existing expansion boards would fit. Cromemco's Roger Melen encouraged the habit of calling it the S-100 bus rather than the Altair bus, and the standard was later adopted formally as IEEE-696.
Microsoft begins with a letter and a phone call
Roberts received a letter from a company called Traf-O-Data asking whether MITS wanted to buy a BASIC language for the new machine. When he called the number he reached a private house where nobody had heard of such a thing. The letter had come from Bill Gates and Paul Allen, and they had no BASIC to sell yet.
They wrote one anyway, on a DEC PDP-10 mainframe using a software simulator, because neither could get near an Altair. Allen then flew to Albuquerque with the finished interpreter on paper tape. On its first run the program printed READY; Allen typed PRINT 2+2 and the machine answered 4. Gates joined him, and the partnership they formed, then spelled Micro-Soft, became Microsoft. Altair BASIC, announced in July 1975, cost $60 in its 4K version and $75 in its 8K version when bought with the machine.
Demand, and the gap it opened
Roberts had told his banker he might sell 800 computers. In February 1975 MITS took 1,000 orders, and the promised 60-day delivery slipped by months. The company claimed 2,500 machines delivered by the end of May and more than 5,000 by August. Staff went from under 20 in January to 90 by October. Roughly 25,000 units were sold in total.
The Altair itself was close to a break-even sale, and that was the point: the profit came from memory and interface cards. Those cards were also its weakness: MITS's 4K dynamic memory board was unreliable and slow to ship, so buyers built their own supply chain. Robert Marsh's Processor Technology sold a plug-compatible 4K static board for $255, Don Tarbell's cassette interface arrived in February 1976 at $150 assembled, and Cromemco's Bytesaver ROM put a boot program into the machine at power up. The Altair created the add-on industry, then watched it supply the parts MITS could not.
Compatibles, copies and the end of MITS
IMS Associates announced the IMSAI 8080 in October 1975, plug compatible, easier to assemble and priced at $439 as a kit, with the first 50 units shipping that December. MITS answered with the Altair 8800B in June 1976, which corrected several of the original's design faults. It was not enough. Pertec Computer Corporation bought MITS in 1977, and sold the same computer under a new badge as the PCC 8800 in 1978.
What survived was not the hardware company but the ecosystem around it. The Altair gave hobby electronics a product to gather around: clubs formed in response to the January 1975 cover, the most famous of them the Homebrew Computer Club in Silicon Valley, where Steve Wozniak showed the first Apple computer in 1976. Almost everyone who built a personal computing business in the following decade looked at the Altair and decided they could do it better.