Cheap as chips: the end of Moore’s law and its impact on technology
Have you ever wondered how the electronic devices we use for everything from sending WhatsApp messages to capturing Pokémon can possess enough computational power to blow the Apollo Guidance Computer out of the water (and into space), yet cost a mere fraction of the price? The answer was famously prophesised by the cofounder of Intel, Gordon E. Moore.
“Moore's law” states that the number of transistors on a microprocessor chip will double approximately every two years. And because the transistor is the element within a chip that enables a current or voltage to flow around it, essentially making it work, this exponential growth has generally meant that the performance of chips has improved just as quickly. That’s because transistors have the unusual quality of getting better as they get smaller; a small transistor can be turned on and off with less power and at greater speed than a large one. And keeping pace with predictions, every year the industry churns out new iterations of chips that are smaller, cheaper and faster – without fail.
But the economic engine based on such a growth model seems to be drawing to a close as components approach a fundamental limit of smallness.
Making chips larger and transistors smaller isn’t an easy task. Currently 7 nanometres (nm) (as of 2016, no 5nm scale chips have been commercially produced), transistors are becoming ever more prone to overheating. As a result, they can interfere with neighbouring components due to their close proximity, meaning the chips are unpredictable and unreliable.
But the technology industry faces more than this scientific problem: it also faces a financial one.
For decades, semiconductor companies have spent heavily on research and development (R&D); it has also become far more expensive to build the environments that facilitate chip manufacturing and development. But as transistors shrink, the market for them grows, allowing manufacturers to recoup their investments and giving them further incentive to reinvest.
The increasingly sophisticated equipment required to make semiconductors becomes more expensive with every iteration of Moore's law; soon it won’t be financially viable to make chips any smaller. To manufacture the most advanced chips requires tools that cost more than $50 million. These prohibitive costs have led to the depletion of chipmakers in the industry. Only a few remaining big hitters such as Samsung, Intel and IBM have the financial clout to reach the economies of scale required to make such investments worthwhile. And even then, these rivals have been forced to share their financial pain by teaming up to share chip-making technology.
A silver lining
So, is it all doom and gloom? I don't think so.
For too long, the semiconductor industry has relied on this seemingly guaranteed path to innovation, and as a result it has become complacent. In the past, the relentless doubling and redoubling of computing power meant there was little incentive to experiment with other avenues that might present alternatives to the silicon chip. The end of Moore’s law could be a blessing in disguise – the industry will now be forced to build better machines, to become more inventive and to invest in more adventurous R&D. Hopefully a variety of new computing technologies will come into their own; perhaps we will see the emergence of new and innovative tech companies that under normal circumstances would be unable to enter the market because of the eye-watering costs required.
A number of companies including IBM, Intel and Micron are now experimenting with stacking chips on top of each other, allowing room for more transistors in a given area. It’s a tough process though. Fabricating a silicon chip requires temperatures close to 1,800°F, making it extremely challenging to build a silicon chip on top of another without damaging the layer below. In order to combat this hurdle, two silicon chips can be constructed separately before being stacked and connected via thousands of wires – using non-silicon materials that can be fabricated at much lower temperatures. Conventional 3D silicon chips are still prone to “traffic jams” – where data is bottlenecked due to a lack of wires – and it takes a lot of energy to push it through. Cooling these stacks has also proven to be a nuisance due to the restricted accessibility of each layer.
IBM is also investing heavily – almost $3 billion – in the search for viable alternatives to silicon. One of the best-known options is graphene, a single atom-thick form of carbon. Engineers at IBM Research have built the world’s most advanced graphene-based chip, which performs 10,000 times better than previous graphene incarnations. But doubts remain as to whether it will drive the future of chip production. Graphene conducts electricity extremely efficiently, but that’s where its benefits end. It can push a charge at an incredible speed, but it cannot retain it. In a binary system we may need to retain data so that our running programs don’t close the instant they open. It’s important in a RAM chip, for example, to ensure that the data inside it remains readable for the foreseeable future. The difficulty arises with making it stop. When a transistor is in the “on” state, it registers a 1 and in an “off” state, it registers a 0. Graphene would be unable to “switch off” because the difference between “on” and “off” voltage is very small.
Another alternative is a spintronic transistor. Where electronic devices use the charge of an electron to represent information, spintronics use “spin”, another property of electrons. Research into spintronics has been going for a number of years now, but no transistors have yet made it into production. Appealingly, the voltage needed to drive spintronic transistors is tiny. That eliminates the heat problems experienced by graphene, but brings problems of its own as such low voltage, which means distinguishing between a 1 or 0 from electrical noise becomes tricky.
The semiconductor industry is approaching a crossroads – and the impact will be felt across many areas of everyday life. It may prove to be the end of an era, but I am certain there is a new wave of technological innovation just over the horizon. The end of Moore's law does not spell the end of all technological advancement. It merely suggests that the gains we are used to seeing will occur less frequently. Progress will march on and there will be ample time to continue developing new tech and pick up where silicon chips left off.
When the chips are down, I’ll bet that chipmakers are going to come out fighting.
Gows has left The Frameworks.