The Software Manufacturing Delusion

September 13, 2010

“Software is largely a service industry
operating under the persistent but unfounded delusion
that it is a manufacturing industry.”

Eric S. Raymond

A trip back from OSCON 2010 brought the nice opportunity of having five hours available for book-reading.

The lucky choice was:

The Cathedral and The Bazaar,
by Eric S. Raymond

Which in the back of my mind was a “classic” book from the “early” days of open source.

I thought on reading it for historical interest…

Alas, how wrong and misled I was!

The chapters of The Cathedral and the Bazaar still read perfectly today, and reveal with pristine clarity many of the concepts that, despite the widespread adoption of Open Source in the corporate world, still have many minds struggling.

One of the most revealing chapters was The Magic Cauldron (original here), where Raymond goes in depth analyzing the economy of software development an its implications in open source.

The essence of this chapter to me is summarized in one of its subtitles:

The Manufacturing Delusion

In this chapter, Raymond tell us

We need to begin by noticing that computer programs like all other kinds of tools or capital goods, have two distinct kinds of economic value. They have use value and sale value.


The use value of a program is its economic value as a tool. The sale value of a program is its value as a salable commodity. (In professional economist-speak, sale value is value as a final good, and use value is value as an intermediate good.)


When most people try to reason about software-production economics, they tend to assume a `factory model’ that is founded on the following fundamental premises.

  1. Most developer time is paid for by sale value.
  2. The sale value of software is proportional to its development cost (i.e. the cost of the resources required to functionally replicate it) and to its use value.

In other words, people have a strong tendency to assume that software has the value characteristics of a typical manufactured good. But both of these assumptions are demonstrably false.


I decided to verify this assertion using the following public data:

Six of the largest producers of software for sale

Software Publisher Balance Sheet Annual Revenue
Cost of
Gross Profit
Microsoft Corp.  $ 62,484 M $ 12,395 M $ 50,089 M
IBM Corp. $ 95,759 M $ 51,972 M $ 43,787 M
Symantec $ 5,985 M $ 1,105 M
$ 4,880 M
Apple Inc.  $ 42,905 M $ 25,683 M $ 17,222 M
Oracle  $ 26,820 M $ 5,764 M $ 21,056 M
Adobe  $     943 M $ 107 M  $ 835 M

Note that you must click on the “Annual” tab in those financial reports in order to find the same number shown in this table. (The numbers for this table were taken on September 10, 2010).

My list is not the same as Forbes list of top 100 software companies, so these numbers will deserve a review.

The amounts for Apple Inc. and IBM Corp. are arguable, given that both of these companies deliver a combination of hardware, software and services.

IBM’s 2009 Financial report
in page 10, partition its business as

  • 7% Hardware
  • 9% Financial Services
  • 42% Services
  • 42% Software

Apple’s third quarter financial report of 2010
in page 35, partitions the sales as:

  • 8.3%  Mac Desktops
  • 19.7% Mac Laptops
  • 9.8%  iPods & related products and services
  • 7.7%  Music related products
  • 13.8%  iPads & related products and services
  • 34.0%  iPhones & related products and services
  • 2.5%  Peripherals and other hardware
  • 4.1%  Software, services and other sales

This probably deserves some discussion, maybe in a follow up blog, since it is unclear how much of the sale cost of a Mac laptop, or an iPhone, or an iPad is the software included with the device; and therefore, whether such fraction of the cost should be considered hardware or software. As a reference, note that when you buy a Cisco router, the router includes software, but the price is not split into physical hardware and software, at least not to the buyer.

Back to our original exercise,

How big is the “software for sale” industry?

When we add the revenues on the table (weighting IBM and Apple by their fraction of software business), their cumulated revenues are in the order of $150B a year.

Note however, that the cost of developing that software for sale is actually a lot lower. For one of the largest vendors in this list, its revenue is $ 62,484 M the actual costs of revenue is only $ 12,395 M which leaves the non-insignificant sum of  $ 50,089 M as gross profits. A ratio of 5:1.  This also is an interesting observation and, should probably be discussed as well in a follow up blog.

Let’s look now,

How big is the “software for use” industry?

This is, how much is invested in developing software that is not intended to be sold (or licensed) outside of an institution, but to be used internally in the process of running the organization. This is typically the cost of developing, customizing and maintaining the software that is used in every company, academic institution, and government agency for doing things like rolling the payroll, maintaining inventories, hosting websites, managing online stores, providing customer services, storing and managing internal documentation.

This is not a trivial question, and it is well beyond my limited understanding of economics, so I will rely on smarter and more capable people to give us the numbers.

Here is my source:

Roger Sessions, in his blog post “The cost of IT Failure” reports that:

According to the World Technology and Services Alliance, countries spend, on average, 6.4% of the Gross Domestic Product (GDP) on Information Communications Technology, with 43% of this spent on hardware, software, and services. This means that, on average, 6.4 X .43 = 2.75 % of GDP is spent on hardware, software, and services. I will lump hardware, software, and services together under the banner of IT

In Roger’s account, the global GDP is about $ 69,000 B

This lead us to $ 69,000 B x 2.75% = $ 1,897 B in IT expenditures a year.

At this point we need to know what fraction of IT expenditures are invested in developing and maintaining software. We will arrive to this number by a maneuver of global accounting Jujitsu. We will take the revenue of the hardware industry and subtract it from the global expenditure in IT.

Here is our estimation:

Hardware Vendor Balance Sheet Revenue Hardware Factor
HP $ 114,552 M    100%
IBM $   95,579 M     42%
Dell $   52,902 M    100%
Cisco $   40,040 M    100%
Sun  Now part of Oracle, so we use numbers from the Wikipedia $   11.700 M    100%
Apple $   42,905 M      27%

Again, for IBM and Apple we apply the factor of how much of their sales are related to hardware, and we get to an estimate of $ 270 B of annual expenditures in Hardware.

Note that for vendors, other than IBM and Apple, the nominal value of 100% was used, but it could probably be estimated below that number.

A better estimation could be obtained by working it from the list of the 100 largest hardware companies:

We subtract this hardware estimate from the annual IT expenditure and get

  $ 1,897 B  $ 270 B = $ 1,627 B

That we can claim as the annual investment in maintaining IT  infrastructure. We finally compute the ratio between these two numbers and get:

Software for Sale $150 B : Software for Use $1627 B  =  9%

Granted, many steps in this estimation are worth of a larger discussion, but still we get to an estimate quite close to the one Raymond presents in his book.  The industry of software “for sale” is about 10% of the larger in-house software industry.

This is an interesting verification of Raymond’s account for the real nature of the software industry: the large majority of software is developed as in-house projects in companies, governments agencies and academia, and it is not intended for sale, but for direct use in the organization that developed it.

In this context, the proposition of sharing source code, with the goal of distributing the burden of maintaining and improving the software, is one that make perfect economic sense.

If you have read this far, you probably notice that I left two homeworks half done:

A) A better estimation of the software for sale industry, based on

B) A better estimation of the hardware industry size, based on

We can anticipate that a more complete account will result in a larger value for (A) as well as for (B), and that will lead to a ratio estimate larger than the 9%  presented here.

DISCLAIMER: Yes, I understand that software is “licensed”, not “sold”. So, the term “sale value” used here is to be understood as the value that licensees pay to software publishers for the permission to use the software.

There are very few actual “software vendors” in the industry. Real software vendors are closer to construction contractors who will  build a house for a customer, and then the customer will fully own that house.

In practice, the software licensing industry is closer to the car rental industry than the car manufacturing industry. After all, as a software customer, you’re paying for permission to use goods for a limited period of time and are subject to a precise list of conditions and restrictions without ever becoming the owner of the goods.

Leave a Reply