Supercomputer projections from top500

This is the projected performance chart from the top500.  They track the top 500 supercomputers in the world, releasing the list every 6 months.  It has taken just about 4 years for every 10x increase in performance.  But even if it slows down to every 5 years, we are still talking rapid increase in performance.

In 2016 we would see a 20 petaflop system.  That is over the 10 petaflops the pharmacuetical companies said they need to do timely biological simulations at the cellular level.  Although the first system would be very expensive.  20 petaflops is the estimate that Ray Kurzweil has guestimated for the speed of the human brain.

In 2021 there would be a 200 petaflop system.  And in 2026 a 2 exaflop system.  Some of the major nations are already rumored to be aiming for an exaflop system by 2020.  Surely by the mid 2020's a 10 petaflop system would be relatively cheap for the pharmacuetical companies.

Computing will drive future electric growth

The supercomputer in the picture is MareNostrum, installed in an old chuch in Spain.


An interesting example is to look at the growth of supercomputing power consumption.  The biggest supercomputers in the world 10 years ago used about 500kw of power.  Today they are up at the 5 megawatt level.  At the current pace in 2020 they will be at 50 megawatts and by 2030, 500 megawatts. 

A modern nuclear reactor is about 1200 MW, a wind turbine 2 megawatts.  As the supercomputers grow in computational power they have more applications that we want to use them for.  The big pharmacuetical companies have estimated that they need 10 Petaflops of performance in order to do timely and useful simultions of biological interactions.  These are companies with 5 billion dollars a year in research budgets, so when it becomes useful they will not have a problem spending what it takes.

But right now the biggest 3 supercomputers in the world have just surpassed 1 Petaflop of performance.  But in under 5 years we should see 10 petaflop computers installed. 

Of course supercomputers are only one part of the world computing useage..and in 2010 a quite minor one.  There is hosting servers, datacenters, carrierhotels and cloud services.  And in the home there is more and more computing devices.  An estimate I saw said computing is already 20% of US electric demand and is growing at 10% a year.  That means just computing would cause a 2% growth in total US electric demand, all else being equal.

And 2% is a non-trivial amount for one year in a nation with 1,000 GW of installed capacity. That would require 20 gigawatts of new capacity to meet 2% demand growth.  Or 20 nuclear reactors.  (or 10,000 wind turbines).

Prediction for China car sales 2011

Production in millions units from wikipedia.  Passenger and commercial vehicles combined.  Growth over previous year in brackets.  http://en.wikipedia.org/wiki/Automobile_industry_in_China

1992:  1.0
1999:  1.2
2000:  2.07 
2001:  2.33  (12.6%)
2002:  3.25  (39.5%)
2003:  4.44  (36.6%)
2004:  5.07  (14.2%)
2005:  5.71  (12.6%)
2006:  7.28  (27.5%)
2007:  8.88  (22.0%)
2008:  9.35  (5.3%)
2009:  13.83  (47.9%)

Last year I predicted 2010 sales would grow to 17.5 million unit.  I simply took the last 9 year average of 24.2% growth and added that to the 2009 total, which came to 17.2 million units.  Then because I was bullish I added a little bit. 

According to Xinhua sales for the first 11 months of 2010 were 16.4 million units.  That puts them on pace for 17.89 million units in 2010.  http://news.xinhuanet.com/english2010/business/2010-12/09/c_13642345.htm

2010:  17.89(est)  (29.4%)

So that brings the 10 year average growth rate to 24.7%.  So I will guess 24.7% growth for 2011, which would bring 2011 sales to..

2011:  22.31(my guess) 

World added 15 UKs over decade for electricity.

One thing that amazed me is the increase in world electric production over the decade of 5,440 billion kwh, was the equivalent of adding 15 United Kingdoms(361 billion kwh) in electrical production.

I've been trying to find statistics to relate to people how broad this global economic growth is right now. 

World Electric prediction for the decade

In 1998 world electrical production was 13,663 billion kilowatt hours.  I predicted my usual, a 2.7% rise in production per year.  Because that is the long term historical average.  So my prediction was for 2008 production to be 17,830 billion kilowatt hours.  Back then everyone I told said it wouldn't happen, because of limited resources.

The numbers are out for 2008 now, the world production was 19,103 billion kwh.  The United Kingdom produced 361 billion kwh in 2008.  Or ~1.89% of world output. 

My prediction for the 2018 numbers which will be out in 2020.. is simply an annual 2.7% increase from now.  Or 24,929 billion kwh.  That growth if it happens will be like adding 16 United Kingdoms.

How unemployment creeps up

The 2000-2010 decade was the first decade in American history where the country went out the decade with less jobs than it went in.  The population grows by about 1% a year, the population grew by somewhere around 28 million people over the decade.

In the year 2000 America had 280 million people and about 140 million active in the labour force.  For a labour force participation rate of 50%.  In the year 2010 America had ~310 million people, but the labour force had declined to 138 million.  If the labour force participation rate had remaned at 50%, there would be ~155 million active in the labour force.  So this difference amounts to 17 million jobs.  In fact because the country grows by about 3 million each year, at a 50% labour force participation, about 125,000 new jobs must be created each month to stay even.  17 million jobs is 12% of the current labour force of 138 million.

That works out to 1.2% of jobs lost each year.  For a random economically active individual, the chances of you becoming out of work this year are 1 in 100, if things continue at this pace.  And it probably isn't even that high.  Corporations and governments usually reduce their workforce through attrition.  As people retire they do not replace the person.  As people leave for other opportunities, they do not replace the person.  When people die or become disabled, they do not replace the person. 

So where we are really seeing the unemployment is the young, coming out of school.  And in many western cities the unemployment rate among 18-25 year olds not in school is catastrophically bad.  It was reported in New York City, the rate was 50%. 

 

Growing unemployment from tecnology

One of my main fascinations at the current time is the rising unemployment across most of the developed nations.  My contention for years has been that automation is the primary driver.  At some point technology begins replacing jobs faster than the same technology creates new opportunity for people.

If you look at many of the big jobs of the post war era, jobs that could support families, ones like auto workers, steel workers, port workers, rail and electrical workers, most of these fields have seen a huge reduction in the number of workers needed.

Today steel needs less than 1/5th the workers that it needed in the 1960's to produce 1 tonne of steel.  So you get the strange situation of old steel towns which appear as ghost towns, and the assumption is that the mill has shut down.  But in actuality the mill is still operational and producing the same output as it was 50 years ago.

In addition to physical automation there has been vast automation for white colar workers.  Even in the 1980's there was armies of clerks at all the major corporations who were managing information.  Filling out invoices, filling out forms, getting signatures, filing copies, keeping track of accounts and so on.  This is why every morning tens of millions of people travelled from their homes into city centers and worked in office buildings.  Each worker working in a cubicle, with hundreds of other cubicle drones on his floor, and a building tens of stories high.

But today most of the information is handled electronically with computers effortlessly updating accounts, and effortlessly filing the information.  So you have a few people working in a server farm out in suburbia.  This is a big reason why the downtown centers of cities have hollowed out. 

In some of my coming posts I will look into what it means for our society and economy with tens of millions of these middle class jobs permanently disappearing.  That is something to realize, these jobs are never coming back.  For decades after farm automation took over during the 1930's, politicians tried to revive the 'farm economy'.  But they never were able to get the rural economy going, they were not even able to stop the loss of jobs from continuing in the farm sector.