IBM: It was, when, 1890, that the US Census took more than 10 years to process one census that is supposed to happen each 10 years? So, US Census needed a bet

IBM: It was, when, 1890, that the US Census took more than 10 years to process one census that is supposed to happen each 10 years? So, US Census needed a better tool. Enter H. Hollerith who borrowed the idea of punched cards from, maybe, the Jacquard loom. Tom Watson was with National Cash Register or some such, saw the Hollerith equipment, concluded that it was the secret to automating routine business record keeping, especially accounting, and IBM was born. Watson sold the heck out of his equipment. And he had a team of good electro-mechanical engineers in Poughkeepsie designing and building the equipment. That takes us to about 1960. They made a lot of money. About then, Tom Watson, Jr. visited Columbia University and reported that there was a guy there doing things, whatever, 200,000 times a second. An IBM life insurance customer in NYC complained that the IBM punched cards were taking up too much room. So, some changes were needed, e.g., magnetic tape. Due to US WWII work on vacuum tube digital computers and pushed by Univac, IBM got into the computer business, with magnetic tape, software, etc. In the late 1950s Watson was told that he had no research division and that to have a good one he needed a leader good researchers would respect. So, Watson got a chunk a land in Westchester County, NY and hired Eero Saarinen to build him a fancy research lab on a hill near Yorktown Heights. By the mid 1960s, IBM had an idea, a line of compatible computers called System 360 (the origin of the IBM 'mainframe' computers). So, a customer could start with a cheap, small, slow System 360/20, 30, 40, etc. and when ran out of capacity upgrade to a 360/50, 65, 85, 91, 95, etc. and just keep running the same software. IBM regarded itself not as an electronics company or a computer company but a marketing company. The marketing/sales people had nearly all the power. Here the focus was still on routine business record keeping, 'business machines', and not general purpose computing. IBM had some good work in general purpose computing, but the executives stayed with their 'good IBM customers'. So, when MIT did Multics in 1969, IBM was slow to respond. When DEC was selling good general purpose time sharing on the DEC 10, IBM was slow to respond. As a means of a time-sharing computer to be used for developing operating systems, IBM Cambridge Scientific Center did CP67/CMS for the IBM 360/67 virtual memory computer of about 1967 and where CP67 was virtual machine. Lots of people loved CP67/CMS, later VM/CMS, but IBM's marketing people didn't want to sell it. Internally, VM/CMS ran the company through at least 1994. Also there was VNET, roughly like the Internet except the computers were also the routers -- did essentially of the IBM internal computing through at least 1994. Use it? Yes. Eagerly sell it? No. 'Dog fooding'? No. In the 1970s, IBM did move to System 370, a 'mainframe' which had virtual memory. Customers wanted to do interactive, on-line applications, and IBM responded slowly. For the communications, IBM did Systems Network Architecture, which was about as flexible and easy to install as a railroad and about as costly. IBM did notice that with much in on-line activity, they could sell a lot more System 370 computers and did. IBM was slow to let the capacity of the 370 computers increase as quickly as customers wanted, but by 1980 IBM had a collection of faster boxes. As 18 wheel trucks lined up to receive these computers, the line backed up to the NYS Thruway, and there was a blip in the US GDP. A lot of IBM customers were doing 'personal productivity', e.g., word processing and electronic spreadsheets, on IBM's mainframe computers. But that was 1980, and DEC, Data General, Prime and others were doing well. Prime was a single board, bit-sliced, virtual memory computer with some extra register sets for fast process switching and a darned efficient time sharing box, much more efficient, and much easier to use, than anything from IBM. And in 1980, Prime gave the best ROI on the NYSE. Also about here came the microprocessors, e.g., Intel 8080 and 8086. Then PCs began to explode. Then work started moving off IBM's mainframes. By 1986 or so, DEC was getting more revenue from DuPont than IBM was. There were well done IBM internal reports on technology and markets that outlined the future and the threats to IBM's business, but IBM's leadership, really successful mainframe salesmen, essentially ignored the reports. The mainframe people had the power and worked to kill off anything else inside IBM. At one time, IBM CEO Gerstner said that "IBM is the most arrogant, inwardly directed, process oriented company I've every seen.". By 1986, at an internal top management meeting, it was possible to go around the table and find that nearly no one had made their projections. The conclusion was that God had ceased to smile on IBM. IBM laughed at Intel and Microsoft, and those two came to have by far the last laugh. For a while IBM's Cocke's work on reduced instruction set computing (RISC) gave IBM an opportunity to grab the high end desktop and workstation markets, e.g., for finance, engineering design, graphics, but the mainframe people didn't like the competition. E.g., when an IBM mainframe had a processor clock of about 10 MHz, Cocke's discrete component board with RISC had a clock of 80 MHz. Near 1994, IBM in three years IBM lost $16 billion and went from ~400,000 employees down to ~200,000. The research division phone book went from 4500 names down to 1500 with about 500 of those recent temporary employees. Then IBM pushed services, e.g., they would run your data center for you. IBM bought companies with attractive products and put the IBM marketing force behind those products. Net, the first and last 'visionary' thing IBM did was to grab Hollerith's work, although, of course, it was Hollerith who was the real visionary. Since then IBM has focused on selling 'business machines' for routine business record keeping for large banks, insurance companies, manufacturing companies, and governments. That's their 'market'. They haven't tried very hard to look for other markets. In technology, IBM lets others take the first steps and maybe does something similar or maybe just buys a company. That's what they are still doing. IBM had a lot of opportunities they just dropped, apparently deliberately: They could easily have had all of Cisco, Intel, Microsoft, and Oracle. At one time, they ran NSFNet, that is, the Internet; had IBM stayed with that work, they could have be running some huge fraction of all of the Internet. IBM was long a leader in laser printing, e.g., for printing bank statements on a roll of paper moved with a fork lift truck, but in the US HP made big bucks in that market. In the 1980s IBM saw the need for video servers and had some working; same for wearable computing. For the Internet of things, IBM has long had the TCP/IP stack on a chip. For relational database, IBM invented it, roughly parallel to E. Wong's work at Berkeley, but now the revenue for relational database goes to, right, Oracle, etc. IBM did early work on the tricky 'passive' amplifier wrap around an optical fiber to amplify the digital signal without converting from optical to electrical; likely that amplifier is crucial for the backbone of the Internet, but IBM is not running that backbone; maybe IBM is getting some patent license revenue. IBM did good, early work on giant magneto resistive disk heads and associated vertical recording and long was the leader in magnetic disk, but now people buy hard disk drives from Seagate, Western Digital, Maxtor, etc. IBM did a lot of high end work on large disk storage systems, but people buy from EMC, build their own, etc. It goes on and on: At one time or another, IBM had in their lap the beginnings of nearly everything we see today in information technology but dropped it. IBM remains focused on being a 'marketing' company, but now they are a bit short on what to market. Ah, yet another mainframe salesman bites the dust.