An industry built on sand

By Malcolm Penn

The first transistor was a crude device invented at Bell Laboratories

Sixty years ago three scientists in the US invented the transistor - the tiny switches at the heart of all silicon chips.

Now they are found in everything from cars and aircraft to MP3 players and mobile phones.

Analyst Malcolm Penn looks back on the phenomenal growth of the industry.

Following the transistor's invention by John Bardeen, Walter Brattain and William Shockley at Bell Labs in 1947, no one really knew quite what to do with the invention.

Little more than a laboratory curiosity, it was not until manufacturers realised that the tiny switches would enable products to be built smaller, more reliably and with less power consumption than with conventional electronic valves that the market started to develop.

It took the technologists five years to come up with the first practical transistor application - a hearing aid in 1952.

Whilst not a large market, the hearing aid did demonstrate two key features of the new technology; it enabled something to be done that was impossible with the previous delicate valve technology and it helped enhance the quality of life for those people who were hard of hearing.

TRANSISTOR HIGHLIGHTS 1947: Transistor invented 1952: First - commercial application - hearing aid 1954: First mass-market device - transistor radio 1956: First transistor computer Digital timeline

These two features, along with the industry's ability to reduce the production cost of a transistor by around 30% per year, have subsequently proved to be the fundamental drivers behind today's trillion plus dollar electronics industry.

Eight years after the transistors first laboratory demonstration, launched in December 1955, the watershed product that truly grabbed the world's attention was the first transistor radio.

Working off batteries, the $49.95 (over $1,000 in today's equivalent terms) pocket-sized Regency TR1 was the smallest and most affordable radio that the world had ever seen.

Silicon hotbed

For the early part of the industry, products were made out of individual transistors, manually assembled with the other necessary electronic components (resistors and capacitors).

TRANSISTORS Transistors are basic electronic switches found in silicon chips Each transistor can be switched on or off, representing a "1" or "0", known as binary code All computation is done using different combinations of these two outputs to do calculations The number of transistors on a chip determine its speed Modern chips contain millions of transistors allowing them to execute millions of calculations per second The tiny devices consist of a source, drain and gate A voltage applied to the gate and drain turns the device on Removing the gate voltage switches it off again Switching on the digital age

By the beginning of the 1960s, individual transistors were replaced by the integrated circuit (IC or chip), a process whereby several transistors were built at the same time on a single piece of silicon.

Most of the early device manufacturers were traditional valve companies, soon to be displaced by the fledgling specialist semiconductor companies, the most famous of which being Fairchild Semiconductor in Palo Alto, California and Texas Instruments (TI) in Dallas, Texas.

A spin out from Shockley Semiconductors, Fairchild proved to be a hotbed of technology innovation throughout the whole of the 1960s, including the invention of the Planar manufacturing process - the ability to print tiny patterns on the silicon surface using photographic techniques - still the foundation for today's mainstream technology.

It was at Fairchild in 1965 that Gordon Moore, one of the co-founders of chip-giant Intel, sketched out his prediction of the pace of silicon technology.

Subsequently to be known as Moore's law, his remarkable prediction that IC complexity would double every two years or so has stood a 40 plus year test of time.

Shrinking sizes

TI, together with the Boston Consulting Group, was to develop most of the industry's fundamental business concepts, including forward pricing and the industry learning curve.

Whilst Fairchild recognised the industry's ability to make transistors smaller - and therefore ICs more complex - TI was the first to realise that transistor production costs would plummet by around 30% per year. This enabled them to price future device deliveries at less that the current cost of production and yet still turn in a profit.

The Sonotone 1010 was the first transistorised product

Such forward price projections enabled the industry to attack and open up new market after market, a concept that subsequently became know as market pervasion - what's expensive today will be readily affordably tomorrow.

Since the 1961 launch of the first commercial IC, containing just four transistors on a single piece of silicon drawn with patterns that were 125 microns (millionths of a metre) in size, the technology has been evolutionary rather than revolutionary.

By 1971, the technologist were cramming 3,500 transistors onto a single piece of silicon, using line patterns now 20 microns in size and by 1996 more than four million, 0.5 microns in size.

Using today's 45-nanometer (billionths of a metre) technology several million transistors can be built on a pinhead, with each transistor costing less than one-thousandth the price of a single office paperclip.

The consequence has been to increase the overall market from nothing to around $300bn, a staggering average annual growth rate of 14% per year.

Growth has been nowhere near steady however, with some years growing over 30% and others declining by 20% or more, making the industry the wildest white-knuckle ride yet invented.

Price crash

The industry has been the result of successfully combining several branches of science, including physics, chemistry, mechanics, mathematics and computer sciences.

The number of transistors it is possible to squeeze in to a chip for a fixed cost doubles every two years First outlined by Gordon Moore, co-founder of Intel Published in Electronics Magazine on 19 April, 1965

In the early years, the semiconductor firms had to make their own production equipment and process materials, before eventually these became separate industries in their own right.

Early production was a hit-or-miss process; with yields desperately low (sometimes less than 20% of the devices made actually worked) until the right combination of process temperatures, timing and chemicals was stumbled across.

Nowadays alchemy and black-art have been replaced by well understood dial-up process recipes, with typically well over four out of every five manufactured chips working.

This does however mean that the production equipment is not only very specialised but also high precision - with the IC transistor structures now built an atom at a time - and thus very expensive.

You could buy a fleet of over twelve A380 Super Jumbos for the price of a single chip factory.

Yet, despite their high cost, technical obsolescence is measured in just a few years; just to stand still the industry needs to invest around 30% of its annual sales in R&D and investment.

This constant flux of change has had a dramatic effect on the supply chain, with companies entering the market only to peak and drop from favour a decade or two down the road, as a new technology and/or applications assumes the so-called "killer" market role.

Just as each past decade has seen a dramatic change in the top 10 rankings, so we can expect more change in the current decade, the seventh decade of the transistor revolution.

A decade is a "lifetime" (five Moore's Law cycles) in the semiconductor industry.

Aside from the more obvious semiconductor applications - our mobile phones, laptop computers, mp3 music players, televisions and digital cameras, cars today are now computers on wheels and almost every aspect of our daily lives is driven or enabled by the semiconductor technology, with no let up in sight over the foreseeable future.

If the same advances had been applied to, for example, the airline industry, a commercial flight between say New York and Paris costing around $900 and taking seven hours in 1978 would today cost about one cent and take less than one second.

Not bad for an industry essentially built from sand.

---

Malcolm Penn is chairman and CEO of semiconductor analyst firm Future Horizons.