MITS made the Altair 8800. It was their first computer. Before that they made calculators and instrumentation.
I believe they shipped an earlier home computer under the MITS brand. It may only have been a kit.
In 1971 MITS developed a home computer based on the National Semiconductor SK0076. It was a 76 bit serial processor with custom instruction sets for different applications. NSC built the part but couldn’t test it, so MITS dropped the project and changed vendors to Intel.
Crane
The first small computer I ever saw (then used for a couple of years, ca. 1974-5) was an HP engineering computer that I later learned had a 4-bit processor and ran a fairly sophisticated form of BASIC. Can’t remember the model number - 9830A? - yup, there it is. Other than having a single-line LED alphanumeric display, it was pretty much indistinguishable from every computer I used for the next ten years.
About a year later, the lab that had the HP got one of the first Commodore PET computers, which predated both the TRS-80 and Apple products. I don’t recall it as being much more capable than the HP except that it had an 80x24 CRT display.
How popular was the COSMAC Elf, based on the 1802 processor, in the hobby market? I only first heard of it a couple years ago at a vintage computer convention (I was born in 75 so the Elf predates my computer using days.) I’m thinking of putting one together for fun.it looks a little bit easier to put together than a basic Z80 or 6502-based computer (although I guess it’s not all that much different.)
The HP-9845 was the big brother. I remember that one - a couple of guys from the local HP office would bring one along to the local computer club meetings. We were in awe of it. Just a ridiculous amount of capability in a box. We were still on punch cards and a mainframe for most real work.
Of course we had no inkling of what was being done by Xeros over at PARC.
Technically, back then – or at least, during the prime of the mainframe boom and IBM’s dominance thereof – IBM didn’t sell computers at all, nor their major peripherals; you could only lease them. The standard way of referring to all mainframe related costs was dollars per month. IBM touted themselves not as a computer vendor but as a service-oriented partnership, and pushed the idea that the monthly costs should be related to the business value provided and not to what the thing cost to make. With that kind of strategic approach to product installations, IBM might not have had much interest in a rich eccentric geek who might present a PR downside if things went badly (this was the company that for a time practically gave their mainframes away to universities, because nothing warmed their corporate hearts more than the thought of thousands of students graduating each year to become business and academic decision makers for whom “computer” automatically meant “IBM”).
A little later, though, Digital Equipment would have been just the opposite. They loved geeks because they were geeks themselves, and a great many of their university installations both of minicomputers and the venerable PDP-10 resulted in many enthusiastic and productive collaborations. I’m not aware of anyone who had PDP-8s or -11s at home in those days, but a few of the real diehards eventually did when they basically became surplus castoffs.
“Ran on” is a fair statement when the computer had no other storage peripherals, and an ASR33 teletype was its only link to the world other than switches and lights, and every program required a long paper tape load. There was no such thing as an “operating system” – the operating system was the user, with a sheet of instructions for how to key in a minimal loader, and then read in a real loader on paper tape prior to every program load. The loader occupied high memory and if you were very lucky it might still be around (and uncorrupted) after the previous program ran, but often that precious memory was needed by the program.
Typical memory for a PDP-8 in those days was 4K, and that was enough to run, for one example, quite a sophisticated interactive interpretive programming language. Today we need more memory than that, for sure, but this is why I roll my eyes when someone says “you need at least 8 GB to run Windows”. Really? Doing the math, that’s 2,097,152 times the memory of a standard PDP-8 in terms of addressable units. But since one is 12-bit words and the other is bytes, it’s about 1,400,000 times the amount of memory in a standard PDP-8, bit for bit. And no way it’s 1,400,000 times better or faster. With Win-Doze, some things are slower. 
Francis,
DEC was an amazing outfit. I gave a seminar there sometime in the 80s. The campus was like a SciFi movie set - slick layout, mega-geek audience and helicopters to take VIPs from building to building. Then it seemed like they were gone overnight.
The loader on the IBM704 was a Read Select of the card reader, 24 copies (punched cards), and a transfer (Jmp) to zero. From there you were on your own!
We wrote most diagnostics in octal object code. That was a lot easier than trying to debug around an operating system.
Crane
Do you have a handy citation? Google isn’t helping. AFAIK, the 8800 was MITS’ first computer product, unless you count the programming add-on for some of their calculators, none of which appear to have anything to do with that part number.
Where was this, out of curiosity? DEC didn’t really have a central campus in the sense that Microsoft or Apple has one. The original facility in Maynard was hardly “sci-fi” – it was an old former wool mill. The company itself was indeed amazingly high-tech, but “the mill” was mainly amazing for the contrast between the ancient rustic environment and the state of the art computers and technology standing within it. It was originally chosen because the square footage was cheap, and Ken Olsen was famously cheap!
As the company grew, however, the R&D centers became distributed throughout rural Massachusetts and a few locations in New Hampshire in the form of modern buildings. There was never a single campus. That’s why the helicopter fleet existed, which also ferried staff and customers between major facilities and Logan airport. And Olsen’s thrift prevailed even in the new real estate – some of the buildings were not new construction, but at least in some cases were buyouts from prior enterprises.
The company eventually disappeared after a number of takeovers because of the same forces that were transforming the entire industry at the end of the last century, and that threatened even IBM.
I thought it was in Natick or something like that, but it’s been forty years.
I got in a car and the field guy drove me there. Sorry.
Crane
gnoital,
Around 1976 or 77 a guy who worked for me made a computer from a kit. He referred to it as a MITS, not an Altair. MITS may have made that distinction between kits and assembled units. I have nothing to go on but a recollection.
At the time we ran a school where we used National Semiconductor IMP16s. Hardly a household word. They were later sold with IMP8 processors as Starplex.
Crane
A friend from college is one of those diehard DEC PDP fans. In the pantry of our off-campus apartment, he kept his PDP-10, powered off half a dryer circuit. After college, most of his collection was in storage. Once, he described how some guy named Paul or some such thing needed an old DEC tape read. So this guy sent over someone from Seattle with the tape, which my friend copied to floppy disk. My old roommate didn’t recognize who had asked him to read the tape but I did. My friend helped the co-founder of Microsoft to read the BASIC interpreter that he and his friend Bill wrote on a PDP-10 at Harvard in 1975.
The name “Altair” is right there on the cover of Popular Electronics, January 1975. (PE actually gave it the name.) Note the it is called a kit.
The Altair brand persisted a bit after MITS was bought out in 1976.
Wow, pretty cool story! 
What Bill and Paul Allen initially wrote was an interpreter for the Altair 8800 (basically, for the Intel 8080 CPU) and then wrote BASIC for that. The first time the BASIC code ever ran on a real Altair was when Allen arrived with it at MITS in Albuquerque. Developing code entirely on a simulated machine was a pretty cool idea especially back then. The irony of it all is that Bill and Paul basically stole the computer time on Harvard’s PDP-10, since the machine was supposed to be for military (DARPA) use only, and they got into a mild amount of trouble over it, but that didn’t stop Bill from getting all sanctimonious and whining that the members of the Homebrew Computer Club out in California were “stealing” his BASIC code! Nor did it stop Microsoft from becoming one of the most possessive, license-obsessed and litigious software companies ever.
Are you sure it was actually a PDP-10 that your roommate had? Lots of other machines used DECtapes, too. Even a minimal PDP-10 would be huge and require a big room with serious air conditioning, and I doubt it could be operated off a dryer outlet. Even the newest-generation KL10 had a power draw of around 12.6 kW – the main power cable alone was the size of a fire hose (pic 1 and pic 2, comparing it to a standard power cord!).
