The Word Two Men Cannot Agree On

Last time I went looking for who invented the term “IT asset management” and came back with nothing. No person, no date, and nothing anywhere you could point at and call a first printing. The thing had just accreted, the way a path wears across grass, and every institution that looked like the parent turned out to be a foster home.

So I thought I’d try an easier one.

Who invented “cloud computing (opens in new tab)”?

And this time there is an answer. A date, and a document you can open and read for yourself, with two men’s names on the story. Which sounds like the end of it, except that when you look at what the answer is made of, it’s two pieces of paper that both belong to the people claiming credit, and the two of them still can’t agree which one said it first.

Bear with me. This one’s better.

The answer everybody gives, which is wrong

Ask around and most people will tell you Eric Schmidt (opens in new tab), Google, 2006. And he did say it, in public, on the ninth of August that year, at a search conference in San Jose (opens in new tab). “It starts with the premise that the data services and architecture should be on servers. We call it cloud computing — they should be in a ‘cloud’ somewhere.”

Read that again, because the sentence gives itself away. We call it cloud computing. You don’t say “we call it” about a phrase you’ve just invented; you say “I’m calling this.”

What Schmidt did was enormous, but it wasn’t naming. The term spread the following year as Amazon, Microsoft and IBM started touting cloud-computing efforts, and that’s when it first turned up in newspapers, including a New York Times piece in November 2007 headlined “I.B.M. to Push ‘Cloud Computing,’ Using Data From Afar.”

So the popular answer is off by a decade.

An office park outside Houston

In 2011, MIT Technology Review’s Antonio Regalado went looking properly (opens in new tab) and came back with something remarkable: a 50-page internal Compaq (opens in new tab) analysis titled “Internet Solutions Division Strategy for Cloud Computing”, dated the fourteenth of November 1996.

Nineteen ninety-six. Everyone was excited about the Netscape browser (opens in new tab). The Yankees were playing Atlanta in the World Series (opens in new tab), and the word we now build trillion-dollar businesses on was sitting in a filing cabinet in Texas.

And it isn’t a phrase used once in passing. It predicts that enterprise software would give way to web-delivered services, and that in future “application software is no longer a feature of the hardware — but of the Internet.” There’s a business plan from the same period with an imaginary bill for the e-purchases of one George Favaloro: $18.50 for 37 minutes of video conferencing, $4.95 for 253 megabytes of internet storage, and $3.95 to watch a Mike Tyson (opens in new tab) fight.

Somebody in Texas in 1996 invoiced a fictional man for streaming boxing and cloud storage. They were describing 2015 while Netscape was still the news.

Two men were in that room. George Favaloro, a Compaq marketing executive who’d just been handed a new internet services group, and Sean O’Sullivan, a young technologist whose startup NetCentric was negotiating a $5 million investment from Compaq. And here the story splits in a way I didn’t expect.

The two of them walked out of that meeting into completely different decades. For Compaq, cloud computing was the start of a $2-billion-a-year business selling servers to internet providers. For O’Sullivan’s company, it was a step towards disenchantment and insolvency. NetCentric got drawn into trying to market an internet fax service – about the only application the patchy network of the day could actually handle – and eventually went under. O’Sullivan took a sabbatical from technology altogether and went to film school. Favaloro, meanwhile, went from being treated as a heretic inside Compaq to being treated as a prophet, and now runs an environmental consultancy in Massachusetts.

Whose paper is it, though

Now. I said there’s an answer, and there is. But I’ve been caught out before by evidence that looks firmer than it is.

Two artefacts. The Compaq document, dated November 1996, which Regalado scanned and published (opens in new tab) so you can look at it yourself. And a daily planner belonging to O’Sullivan, dated the twenty-ninth of October 1996, in which he’d written “Cloud Computing: The Cloud has no Borders” after a meeting with Favaloro that day. Two weeks apart. Those are the earliest documented references Technology Review was able to locate.

Regalado is being careful there in a way most people quoting him aren’t. The earliest we were able to locate. He’s describing how hard he looked.

And both documents were produced by the claimants themselves, fifteen years after the fact, out of a storage unit and a drawer, at the request of a journalist writing a piece called “Who Coined ‘Cloud Computing’?” Nothing external corroborates either date. A planner entry is handwriting on paper; there’s no timestamp on ink.

I’m not suggesting anybody lied. I don’t think they did, and the Compaq document is corroborated by its own contents in a way that would be very hard to fake: nobody in 2011 invents a 1996 business plan that predicts 253 megabytes of internet storage. But by 2011 the term had become, in Regalado’s own words, a disputed prize, with billions of dollars of IT spending in play. Dell had tried to trademark “cloud computing” in 2008 (opens in new tab) and drawn outrage. IBM and Oracle had been accused of “cloud washing”, slapping the word on old product lines. That’s the weather in which fifteen-year-old paper gets dug out of storage units.

And then there’s the detail that undoes the whole thing. The two men don’t agree.

Favaloro believes he coined it. O’Sullivan thinks it might have been him, and reasons that he must have done, because why else would he have tried to trademark it afterwards. Neither remembers precisely when the phrase was conceived. The hard drives that would hold the emails are long gone.

O’Sullivan’s own summary is the most honest sentence in the entire story: “There are only two people who could have come up with the term: me, at NetCentric, or George Favaloro, at Compaq … or both of us together, brainstorming.”

Or both of us together, brainstorming. That’s a man who was in the room, who has skin in the game, who filed the trademark (opens in new tab), admitting he cannot tell you who said it. If the participants can’t resolve it fifteen years on, nobody downstream is going to.

The other people who’d like a word

Once a term is worth money, claimants arrive. Each of them falls over in a different and instructive way.

Ramnath Chellappa, 1997. Widely credited, still credited, in a lot of places. An Emory professor who used the term in a lecture and defined it rather well. The problem is arithmetic: 1997 comes after 1996. Sam Johnston, who edits the Wikipedia entry on cloud computing (opens in new tab), removed the citation crediting the Emory professor. Being early is not the same as being first.

Dell, 2008. Tried to trademark the phrase and drew outrage from programmers, who by then considered the word common property. And nobody trademarks in 2008 a word they coined in 1996.

AT&T. Made a video claiming it invented the cloud. David Linthicum, writing in InfoWorld (opens in new tab), pointed out that this proves AT&T made a video. He was not convinced it proved anything else.

The 1960s crowd. J.C.R. Licklider, John McCarthy, time-sharing, ARPANET. All genuinely the ancestors of the idea. McCarthy suggested in 1961 (opens in new tab) that computing would one day be sold as a utility, and IBM virtualised operating systems in 1967 to let multiple users timeshare the same resource. But that’s the concept, not the name, and conflating the two is how most cloud histories go wrong. Nobody in 1961 said “cloud.”

And the metaphor was never theirs anyway

The coinage arguments all share a blind spot. Even if you accept the November 1996 document, all it dates is the phrase. The metaphor was already old and already communal. Tracing the etymology is difficult (opens in new tab) precisely because the cloud metaphor was widely used by early internet designers to denote the wide-area routing and switching infrastructure between network nodes. Engineers had been drawing a little cloud on network diagrams for years to mean and then some stuff happens in here that isn’t your problem.

The two men who wrote it down say the same thing. They agree cloud computing was born as a marketing term, that telecom networks were already being called the cloud, and that in engineering drawings a cloud already represented the network. What they were hunting for was a slogan to link the fast-developing internet opportunity to businesses Compaq understood. “Computing was bedrock for Compaq, but now this messy cloud was happening,” Favaloro said. “And we needed a handle to bring those things together.”

A handle. Two existing words welded together to shift servers, and one of them had been sitting on whiteboards for a decade already.

The bit that made me laugh out loud

Two months after that document, the term met its first serious opponent. In January 1997, Compaq drafted a press release announcing its investment in NetCentric, describing the deal as part of “a strategic initiative to provide ‘Cloud Computing’ to businesses.” Compaq’s internal PR team objected, and changed it to “Internet computing” in the final version.

Compaq then dropped the term entirely, along with its plans for internet software.

So the first company to write the phrase down killed it in-house within eight weeks, on the advice of its own communications department, for sounding too vague. The word we now hang a trillion-dollar industry on got spiked by a marketing committee in Houston for being bad marketing.

And it stayed dead for the better part of ten years. Their new marketing term didn’t catch fire, which is Regalado’s dry way of saying nobody cared. Whatever brought it back in 2006 was not those two men. They’d shelved it, one of them had gone bust, the other had moved on to selling servers and, eventually, environmental consulting.

Meanwhile, people were building the thing

I was around for some of this, which is the only reason I have anything to add to it.

While the word was in its coma, the architecture was being built by people who had no vocabulary for what they were doing. David Linthicum (opens in new tab) took the CTO job at Grand Central in 2003 – Halsey Minor (opens in new tab)’s company – where they built a multitenant cloud, a message broker “in the sky”, essentially integration-as-a-service. And the obvious problem, as the Forbes profile of him (opens in new tab) puts it, is that the term “cloud” hadn’t even been coined in 2003.

Except it had, seven years earlier. The man building cloud infrastructure simply had no word for the thing under his hands.

It wasn’t even his first run at it. Project Archangel at Saga was “SAGAVista-as-a-service”, and that was too early as well. Minor lost his nerve and shut Grand Central down in 2005, which is the year before Schmidt stood up in San Jose and the whole industry decided this was the future.

I met Linthicum around 2010, when Bick Group (opens in new tab) – a St Louis firm that had been designing and building data centres since 1964 – had just acquired his cloud consultancy, Blue Mountain Labs, and brought him in as CTO. Which I’ve always found funny, given everything above. The metaphor exists to make computing sound weightless, a little cloud on a diagram, and the man most associated with explaining it had just been bought by people who pour concrete and worry about cooling.

That’s the joke at the centre of the whole word, really. There is no cloud. There’s a shed in Loudoun County, Virginia (opens in new tab) with very good air conditioning.

So who named it

Favaloro and O’Sullivan have had three decades to decide they did it. Neither has.

That is not how this normally goes. Priority disputes are one of the reliable engines of human unpleasantness, and the usual shape is two people entirely certain and one of them wrong. Here are two men with a documented claim and every commercial reason to plant a flag on it, and what they say instead is it might have been me, or it might have been him, or it might have been both of us talking. O’Sullivan reasons his way towards himself and stops short. Favaloro believes it was him and won’t say the other man is lying.

My best guess is that neither of them did it in the way the question assumes. The question wants a moment: a man says a thing, the thing is new, the room goes quiet. But both halves were already lying about the building. If somebody put them together and nobody’s head turned, there would be nothing for either man to remember now. A coinage is a thing the future performs on a sentence, and the people in the room never get to see it done.

Which would explain why the word died eight weeks later without a fight. You do not defend a phrase you don’t know you’ve made.

Although I should say that I have now spent several weeks on this and I am no longer sure the mystery is a mystery. There is an old joke about two riders crossing the prairie. Something goes past them at tremendous speed and is gone over the horizon. Who was that, says the first. That, says the second, is Uncatchable Joe. Uncatchable, says the first, has nobody ever caught him? Nobody, says the second, has ever tried to.

Nobody has tried to catch this either. In thirty years the only investigation anybody cites is Regalado’s, and the earliest document he found was one that a man went and fetched from his own storage unit. One shovelful, and that is where the digging stopped. Dell went after the trademark, AT&T made its video, Chellappa was credited by other people, and none of those contests turned on who said it first, because that has never been the valuable part. The question has gone unanswered for want of anybody minding, which strikes me as a stranger fact about the cloud than anything to do with Compaq.