Moore’s Law, Intel co-founder Gordon Moore’s prediction that computing power would double every two years, has enjoyed golden rule status since 1975.
Moore forecast that the size of transistors – semiconductor devices that control the flow of electrons – would halve every two years, allowing twice as many of them to fit on a computer chip. He was right – in fact, between 1975 and the late 1990s, components per chip doubled every 18 months. His prediction holds such weight that more than forty years later, we have come to an almost intuitive acceptance around the life cycle of our own personal technology. Yet the power of Moore’s Law seems to be waning.
Moore’s Law is about costs
Integral to the law was a reduction in cost – the smaller the transistor, the faster, more reliable, and crucially, the cheaper it tended to be. Since 2011, however, this happy relationship between size and cost has been distorted. Manufacturers – or foundries – started to signal slowing cost declines. Designers such as AMD and Nvidia reported increased costs and even skipped production on 2011’s model, deciding to wait a further five years for a smaller node. It has become increasingly difficult to sustain the halving of transistor costs that Moore’s Law requires.
This isn’t the first time that Moore’s Law has been challenged. Gordon Moore himself famously revised his own earlier prediction, from a doubling of transistors every year in 1965 to a doubling every two years in 1975. But signs that time has been called on the halving of transistor costs threaten to end Moore’s Law once and for all. The impact of the death of Moore’s Law on the semiconductor industry, which has embraced the law as a way of setting targets, would be significant. One of the benefits of this perpetual collapse in transistor costs had been the creation of new markets, not only in progressively higher performance computing, but also in lower spec markets such as microwaves, fridges and washing machines – markets now able to use the technology at a lower cost.
Why has the relationship between cost and size broken down?
The root of the breakdown in the relationship between transistor cost and size is complex, spanning multiple problems. For example, the investment required to build a cutting edge manufacturing plant has exploded from the tens of millions to the billions. Meanwhile, with each successive node, the costs to design a chip seem to be rising faster than the benefits, with average design costs around $300m for a 7-nanometre chip and around $540m on a 5-nanometre chip. And crucially, the cost of photolithography – the process by which designers imprint the pattern of the circuit onto a semiconductor – rises exponentially once the size of the transistor falls below 28 nanometres.
What are the investment implications?
The fact that semiconductors power all of our everyday electronics, from radios to computers and mobile phones, means that they are a core focus in our thematic investment approach, particularly within our digitalisation theme. We are mindful of the challenges ahead but we believe that continued innovation holds the key to the future of Moore’s Law.
Foundries are clearly challenged on cost and complexity, with participants now shaken out at every successive transistor size. TSMC, a stock we hold, and Samsung, remain the only foundries still producing at 7 nanometres.
There are two new developments of particular interest. While the current photolithographic process only works up on 7-nanometre transistors, next generation lithography, known as Extreme Ultraviolet Lithography (EUV), may enable the development of transistors down to 1 nanometre. The only supplier of EUV is ASML, a capital equipment provider we are exposed to in our portfolios.
We are also closely observing a number of Taiwanese and Chinese companies; these companies have a reputation as innovators in the manufacturing process, with suggestions of a new technology that could even replace photolithography.
With doubts cast over the future of Moore’s Law, the semiconductor industry faces a challenge. But while we acknowledge the difficulty ahead, we believe there is cause for optimism; so far, each new node has brought with it innovation and novel technology. Moore’s Law is unlikely to continue in its current form, but with so many markets dependent on semiconductor technology, for the foreseeable future, we see the law changing, rather than disappearing altogether. Reports of the death of Moore’s Law may be greatly exaggerated – at least for now.