For decades chip makers have relied on a lucrative and effective strategy to drive innovation and market growth: Cramming as many increasingly miniaturized transistors as possible onto a microprocessor – to drive the pace of innovation predicted by Gordon Moore in 1965. Moore predicted that the number of transistors per processor – and thus computing power – would double roughly every 2 years, while the cost of computing would be halved. Moore’s Law has held more or less true for 50 years, contributing to massive productivity gains and underpinning major innovations in nearly every sector of the economy.
Moore’s Law has quite clearly hit a wall, at least insofar as the strategy of transistor miniaturization has run up against the laws of physics and increasing cost. Some argue the industry can continue its breakneck pace of innovation through other means – specialized architecture, more efficient software, materials innovation, quantum computing – thereby continuing Moore’s Law. While all this may be true, these arguments gloss past the fact that there’s a fundamental shift in the nature of semiconductor innovation underway. The next ten years of computing innovations will look nothing like the last sixty, and silicon manufacturers will need to decide where to place their bets.
Four thoughts on what to expect as we enter the next era of semiconductor innovation:
Improvement in software is perhaps the most direct and cost-effective path to unlocking the potential of hardware innovation. Semiconductor manufacturers have a role to play, both in developing efficient firmware that takes advantage of hardware improvements and in providing tools to their developer ecosystems to ensure their architecture remains relevant and optimized for targeted applications.
One of the major implications of the prevailing innovation strategy of the past 60 years has been that processors have been a general purpose technology – a building block for countless industries to adapt to their purposes. General purpose CPUs/SoCs will always have a role to play, but many high-growth applications benefit significantly from application-driven hardware design. There are massive opportunities in IoT (automotive, industrial) and server segments (AI/deep learning workloads) to design chips more efficiently for targeted workloads.
Materials innovations are an alluring pathway to performance breakthroughs in further miniaturization and quantum computing. But the price tag and development timeline for novel materials are significant and unpredictable, presenting management challenges for any manufacturer who bets the next generation on a novel material. From sourcing and supply chain, to process redesign, to new manufacturing equipment, to uncertainties in scalability, this innovation pathway carries significant risk.
The costs of innovation are rising, and semiconductor R&D expenditures are beginning to tick up accordingly. With such a range of problems to solve and opportunities for growth, no company wants (or expects) to solve them all. Expect more partnership on everything from materials to manufacturing technologies to software, along with ongoing M&A activity to acquire advanced technologies and enhance vertical market positions.
Companies will continue to drive innovation, but the pace and nature of that innovation has changed, and is likely to be more uneven and unpredictable than ever before.