[IAEP] Future of Sugar

Edward Cherlin echerlin at gmail.com
Mon Jun 9 07:13:50 CEST 2008

Since Moore's Law is alive and well for the foreseeable future, just
as it as when Alan Kay was first thinking about Smalltalk and the
Dynabook, it is worthwhile to think about what we can put on future
XOs and compatibles. The general rule is that capacity on chips
doubles about every 18 months at increased cost. So we can think about
an unconstrained XO would look like, and then work out roughly when it
will be available.

Back in 1981, the original IBM PC came with 640K (sic: K!) of memory,
and people asked what you would do with it all. Now 1G is ordinary
(and costs less), a factor of 1600 improvement in capacity in 27
years. So let's see, 27th root of 1.6e3...31% annual increase in
capacity. Nice. So it takes about two and a half years to double
capacity at the same or lower price. Factor of four in five years, 16
in ten,...

We have a few datapoints to start from.

Processor: 433 MHz AMD Geode
RAM: 256 M
Flash: 1G

Same specs, roughly

To get to capacity comparable to a cheap notebook computer of today,
we want to multiply memory by 8, and storage by more. It isn't at all
clear what is going to happen in coming years in the low-cost,
low-power CPU space, except that it is about to go from a quiet
backwater to full frenzy. I expect, therefore, that clock speed will
go up fairly rapidly while holding power down to levels we will need.
We are already seeing such chips announced by Intel, AMD, nVidia, and
others. So, more than five years for sure, less than ten in all
likelihood. At some point the political process will kick in for
create national rural electrification and Internet plans.

In less than ten years, then, some countries will be able to go for
the relatively unconstrained version of everything, and some group of
laggards will still need to use the cheapest alternative, which may be
a $50 laptop, more or less.

XO sales are running at an annual rate of a million units or so. They
can be expected to double or better in 2009, and significantly better
whenever the $75 units become available in quantity. The steady-state
purchase rate for a billion schoolchildren is 200-250 million units
annually, at a cost less than $2 billion for the low-end models.
Countries will move up the product ladder as fast as they feel they
can afford it, particularly when economic results from education start
to be measurable, so that one can calculate the Return on Investment
and compare it with real interest rates.

One of many possibilities:

2008 1M
2009 2M
2010 5M
2011 10M

after which the growth rate is unknown, but we will only need four and
a half more doublings of unit purchases. It is conceivable that the
growth rate will fall to under 100% annually, as is generally the case
with consumer products after takeoff, or that it will increase even
faster once good studies on results come in, countries all scramble to
get it, and the standard funding sources put it into their regular

So ask yourself, How long is it going to take to create the next
generation of educational software that you have thought about doing,
and what will be available to run it on when it is ready? Does that
mean I should start now? Which of the available apps and engines will
be practical to port to Sugar, year by year?

Edward Cherlin
End Poverty at a Profit by teaching children business
"The best way to predict the future is to invent it."--Alan Kay

More information about the Its.an.education.project mailing list