Index Home About Blog
From: (John Mashey)
Newsgroups: comp.arch
Subject: Re: Canadian Technology [really high-tech innovation, long]

In article <1992Mar07.175204.1313@skynet.uucp> ice@skynet.uucp (Ice) writes:

>Why do Canadian startups of any kind always falter?

I think you may be a little too hard on Canada.

>First the Arrow, then finally a chance at supercomputing... and what
>Anyone out there who can comment on what it takes to be innovative, and
>why nothing ever happens around here?

It isn't a question of being innovative.... innovation can happen almost
anywhere.  The problems are:
	getting it financed
	getting it staffed and keeping it that way
	getting the product out
	having enough infrastructure around to make it possible
	getting enough initial support for it to survive
	surviving mistakes
	doing it again and again
	making it a continuing business.

If you study technology history, innovation patterns, and general business
texts, one thing that becomes clear is that startups can be made to work
almost anywhere; nevertheless:
	a) Anywhere, most startups either go out of business, struggle along,
	or get bought out fairly quickly.  Relatively few turn into even
	a $50M business.  Unfortunately, these days, it probably costs
	something like $50M investment to start an innovative computer
	system company, which is why Venture Capitalists don't do much of
	that any more: it's too risky.  New software companies cost less to
	try.  Supercomputers take a *lot* of money and time, hence most
	supercomputer startups fail....
	b) In any given particular technology area, there turn out to
	be geographic locations that have spectacularly better success
	than others.  This has to do with patterns of innovation, ability
	to have people leave one company and start another, ability to hire
	the right people locally, existence of a good local market to make
	it easy to get going, and existence of the infrastructure you need to
	do something without investing huge bunches of money.
	What happens is that once one company in an industry gets successful
	someplace, there is at least some chance that related industry will
	grow around it.  At some point, the competitive advantge of such places
	gets very difficult to overcome.  This is true in many industries, not
	just computers, although computers, I suspect, are one of the more
	extreme cases.
	c) Good current publications to read about such things
	include INC, UPSIDE, FORBES, California Business
	(or the equivalent elsewhere).

For this specific topic (computer design), the most recent issue of
Computer Architecture News happens to have a breakdown by geography of
the subscribers.  One should not overgeneralize from this, but the numbers
do give some hints of what goes on, showing the largest numbers in each
category [US States, other countries]:
4832	Total members
	3590	USA (74% of total)
		951	CA (20% of total, 26% of US)
		307	MA
		292	NY
		262	TX
		156	NJ
		155	IL
		141	PA
		102	VA
		 83	CO
		 77	OR
		 74	MN
		 74	WA
		 73	OH
		 66	MI
		 66	FL
		 64	WI
	 176	Japan
	 131	Germany
	  98	S. Korea
	  94	U.K.
	  90	Australia
	  70	France
	  62	Italy
	  60	Canada

Note of course that CAN is published in English.
Of course, some of these numbers correlate with total populations,
although Australia's 90 is higher than you'd expect from sheer numbers.

However, it is quite likely that ~500 (~10% of total) live within
an hour's drive (well, maybe not during rush-hour :-) of Sunnyvale.
A one-hour radius gets you to San Francisco, Berkeley, Santa Cruz, 
the Peninsula, Livermore & the I680 corridor,and obviously Silicon Valley.
Obviously, the Boston area has a similar concentration
within a 1 hour radius [include southern NH).
A slight difference between Rt128 and S.V. is that 128 is spread out 
*around* Boston, where is S.V. is much more densely concentrated, with
companies right up against each other.

Many places have tried to duplicate S.V. with various degrees of success,
and I've often had discussions with people about why it has worked, and why
it doesn't always.  Briefly, it turns out that a whole lot of factors need to
exist to create an environment in which startups continually breed.

S.V. historical factors include the following:
1) EXPOSURE: During W.W.II, CA had a lot of aerospace industry; also many
servicemen who'd never seen California went out thru CA ports and decided
CA would be a good place to live. Some of them got paid on the G.I. bill to
go to school after the war, some at Berkeley or Stanford, for example.

2) UNIVERSITY+BUSINESS: Stanford, especially, and especially Frederick
Terman (of few of whose students were Hewlett & Packard) had an enlightened
attitude towards letting faculty capitalize on inventions by licensing them,
or by taking short sabbaticals to start companies, and still be able to
come back. Obvious Stanford spinoffs, where the first version of something
was done there, and folks left and licensed the technology [sometimes for
stock], or left and came back include the obvious Sun, SGI, and MIPS.
Stanford also leased land for R&D facilities; you can see where Silicon
Valley really started and spread from: drive up Page Mill Road in Palo Alto,
where Stanford leased land to HP, IBM, XEROX PARC, etc.
*It is widely understood that 1-2 strong universities are very important for
anchoring a high-tech haven, i.e. MIT/Harvard for 128, Stanford/Berkeley for
S.V., U.T. for Austin, etc.  What is often missed is that it's not enough to
just have a good unveristy, the university must have an attitude of working
with business in an appropriate way, or it doesn't take very well.*

3) FINANCING: Venture Capitalists and other sugar daddies.
Startups have to get financed.  There are a lot of ways to get this money;
VC's are certainly not the only way. On the other hand, a GOOD VC
(there are good ones, there are bad ones, like anything else) can
really help a potential startup that it chooses to fund (maybe 1 of 100
business plans it looks at?)
	a) Advise on improving business plan.
	b) Help complete the management team, if not already.
	c) Help get more financing from other VC friends.
	d) Talk to banks or lawyers.
	e) Give advice, help make connections with potential suppliers
	and customers.
That is, money may not be the critical factor, but these other things.

This is a pretty tight community: a huge percentage of them have offices
at Sand Hill Road about 5 minutes from here.  I once attended a large AEA
dinner in honor or Art Rock and Tommy Davis (two of the premier, and early
VCs here, from which many other VC firms are spinoffs). The attendees were
such that if the building had blown up, a large chunk of the electronics
business would suddenly be missing its senior management.  They asked how
many of those present had sometime worked for companies bankrolled by
either of these two, and most people raised their hands.

* Sometimes government support for startups works, sometimes not; when not,
I'd suggest it's not usually lack of money, but of the other support.*

4) BANKS: when you're a startup, and you go to the bank for a short-term
loan, does the bank say:
	a) Maybe, where's your funding, and who's your lead VC?
	b) Where's your last 5 years' of profit statements?
	c) Are you kidding?
In some places in the world, you get b) or c).

5) INFRASTRUCTURE: it helps not to create everything from scratch.  By now,
around S.V., there are companies that do *anything* you can imagine connected
with electronics.  They'll stuff and test boards [like Solectron, which does
boards for many companies]; bend metal; do contract software; do specialized
testing. Need a very expensive ion-milling machine to patch an early chip
so you can keep debugging? Don't want to buy one? No problem, there's
somebody who'll provide that service.

* This is one of the most commonly overlooked necessities. You MUST get
enough on one industry in a place to induce the needed infrastructure.

5) LAND;FACILITIES; TRANSPORT: it helps to have land to build buildings.
S.V. *used* to have a lot of orchards :-) 
A startup needs to get basic facilities cheap, then expand as necessary,
without spending too much money.  Around here there are plenty of
buildings you can lease a piece of, then move.  Also, you can sometimes
buy furniture, partitions, etc. cheap.  MIPS bought some cheap from
Activision; when we moved, we sold them to Ardent; when they moved, they
sold them to sombody else.  They could be anywhere by now.

Many places have created "incubators", which are low-overhead setups for
early offices and maybe labs, with shared secretarial support, meeting
rooms, etc, and usually right next to an industrial park with space to
build new buildings as successful startups need more space.  These are
all over the place; one nice example is the one near Perth, Australia.
Another way is to renovate some old mills.

As a company grows, transportation can become important. It is very helpful
to be located within 1 hour's drive of a major international airport.
If it takes an hour to get to an airport, and from there you need another
flight to get to a major international airport, life can get painful.
If you have to make 2 flights to get to the M.I.A., life will be really
hard if you ever expect to do serious international stuff.

*Note that software companies are much more easily located all over the
place than hardware companies, as the required infrastructure is much

6) (US special): The US has 70% of the world's lawyers, and high-tech
startups can be a contentious lot. People leave A to start B.  A thinks
they took technology with them and wants to sue.  It just happens, that
around S.V. a large number of firms are represented by one particular
law firm that always had a habit of being supportive of startups,
and knows how to handle them. [Wilson, Sonsini, Goodrich & Rosati].
If it turns out that A & B both use WSGR, what often happens is that
it's pointed out to them that their regular lawyers can't represent them
due to the conflict of interest, and why do they want to spend their time
in court rather than building things, and can't they find an easier way
to resolve it?  Often they do.

7) PEOPLE: you need lots of the right people, not just engineers/scientists,
but business management, finance, marketing, manufacturing, etc.  It also
helps to have customers close by, so you don't spend all of your time traveling
just to find the early customers, and the expense of remote support.
The relevant people have various ideas of where they might like to live;
thank goodness not everyone's priorities are the same.  On the other
hand, the following are often Good Things: good weather, proximity to city
large enough to have cultural amenities, beaches, mountains, recreation,
nice housing [S.V. so far so good], reasonably-priced housing [S.V. oops].

Company walls around S.V. sometimes have a genealogical history chart that
shows the evolution/spinoff of semiconductor companies,
starting all the way back with Shockley.  The systems companies often
trace some lineage to HP.

8) GOVERNMENT: if your tax laws say that if you leave a good job,
work 80-hour weeks, and make something happen, that you might get to be
well-off, you'll get startups.  If not, maybe not.  If government rules
make it hard to start businesses, they'll be hard to start.

9) ATTITUDES: if the surrounding society thinks that if you try to start
a business and it fails, that you're stigmatized forever as a failure...
you won't get a lot of startups.  If the area is permeated with risk-takers
whose reponse to hearing your business went under is "Oh, too bad. What's
your next one going to be?" it's different.
It is sometimes said that the US West Coast is populated by people who
(or whose ancestors who) were continually willing to take the risk of
leaving everything they knew and move West ... until they couldn't go
any further, and here they are. [There are plenty of risk-takers elsewhere,
of course, but there is a grain of truth here, as there are a LOT of people
around here wlling to try something different...]
*This is one of the most important factors, and is often the fundamental
reason that S.V.-clone efforts have failed.  You get entrepreneurial folks
everywhere, but some places they're encouraged....and some not*

OK, that's my summary of what it takes.  With the current state of the
industry, until the next serious discontinuity in hardware design,
it's going to be very difficult to start technically innovative new
computer systems companies.... For sure, the number of startups with
successful new general-purpose *instruction sets* is likely to be rather low...
Systems integration, networking,
software seem like better bets for startups, although there is always
room for clever folks to figure out how to put together technology
that used to be too expensive, but now isn't to solve somebody's

Now: back to Canada, which has long had plenty of good computer
people.  Unfortunately, it suffers from the concentration issues
mentioned above, i.e., it's hard to get critical mass in one place,
especially for big projects like supercomputers. [After all, most
supercomputer projects *anywhere* have failed also :-)]  
Fairly clear, if you really want to work on supercomputers,
there's a fairly small number of places in the world to do it....
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: OR {ames,decwrl,prls,pyramid}!mips!mash 
DDD:  	408-524-7015, 524-8253 or (main number) 408-720-1700
USPS:	MIPS Computer Systems M/S 5-03, 950 DeGuigne, Sunnyvale, CA 94086-3650

From: (John Mashey)
Newsgroups: comp.arch
Subject: Re: Canadian Technology [really high-tech innovation, long]

In article <> (daniel lance herrick) writes:

>if you would continue the discussion by commenting on Chipewa Falls.

Sure: I assume the question is why a small place could ever be the center
of the supercomputer business?
1) Somebody more familiar with the history than I should comment.
2) However, one of the tenets of the usual tracts on this topic
(technology concentration and development patterns) is that at the beginning,
the early locations can be almost random; it's later on where there
is strong motivation to keep doing it wherever it already is.
3) Minneapolis has long been a hub of large-computer design,
with CDC, Honeywell, Sperry.
4) Chippewa Falls is not that far away.  and finally:
5) Back then, the center of supercomputer-dom would be wherever Seymour Cray
wanted to be located. [This issomewhat akin to the way company headquarters
moves are planned. (See Joel Garreau's excellent book EDGE CITIES that just
came out last year).  I.e., a big company will do massive studies of all the
effects of various site selection ... but somehow, the one thing you can
guarantee is that whatever move happens will lessen the CEO's commute :-)]

-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: OR {ames,decwrl,prls,pyramid}!mips!mash 
DDD:  	408-524-7015, 524-8253 or (main number) 408-720-1700
USPS:	MIPS Computer Systems M/S 5-03, 950 DeGuigne, Sunnyvale, CA 94086-3650

Index Home About Blog