The growth challenge: How can the semiconductor industry double in size and become more sustainable at the same time?

We are in the midst of a golden decade for semiconductors (despite temporary post-pandemic dips seen last year). The global semiconductor market is forecast to roughly double in size in the next ten years, driven by an ever-increasing demand for computing power and data storage, fueled even further by the surge of (generative) AI. In action you can see it in your smartphone which will become more powerful and require more computing power and in your car which will require more (and more powerful) chips to become even safer and have self-driving features. What this means: more computer chips and more materials needed to produce these chips.

A growing industry means technological advancement, many job opportunities and new products that are just thoughts today. Growth also means environmental responsibility. Because one thing is clear: if our industry continues to work in the same way, our environmental footprint will be an ever greater burden tomorrow.

Let’s look at the carbon emissions of the companies that make the world’s most popular electronic devices. Big tech players in this space have set themselves sustainability targets to reduce their carbon footprint, ideally to (net) zero. An important aspect of this ambition is reducing the footprint of semiconductor manufacturing. For example, around 40% of a smartphone company’s carbon footprint today comes from the process of making integrated circuits.

Now let’s look at recent emission figures in our industry. Emissions of greenhouse gases per wafer are going down, that’s good news. However, the total emissions of the industry go up because of increasing volumes. Processors are becoming more energy efficient, but this is happening at a slower pace than their rapid growth. This is a growth challenge we are facing.

Like others in the industry, material manufacturers have an important role in achieving the industry’s sustainability ambition.

From the perspective of our customers, who operate major semiconductor fabrication facilities, there are two important ways that materials companies can move the needle on sustainability.

First, the process of making semiconductor chips is extremely energy intensive – and this is not just limited to the energy our customers use for making the chips themselves. Energy is also needed to produce the materials that go into chips – chemical synthesis, purification, and waste treatment. As we all push towards more renewable energy sources, we all have a part to play. This is one of the reasons that we at Merck, have aggressively pursued renewable energy agreements. Once a chip is made, one also has to consider all of the energy used to run these chips in servers and personal devices. Today, for example, global data centers consume around 340 TWh of electricity annually (to provide some context; this is more than half of Germany’s annual consumption) – both to run semiconductor chips and to provide the infrastructure to cool them while they are working, and this will only increase. Devising new, more energy-efficient chips and computing paradigms, such as neuromorphic computing, depend heavily on all of us in the value chain, including us with our material and process innovation.

Second, the materials that are used to make semiconductors sometimes have an environmental footprint. For example, the photoresist processes that have enabled transistor scaling down to only a few nanometers across have typically included toxic solvents that need special handling, for example. Meanwhile, the etch gases used to make ever denser computer memories can contribute to global warming with a potency many times that of carbon dioxide. As the science and technology company behind the companies advancing digital living, we are meeting these challenges with material intelligence. We have already released our first generation of photoresist rinses based on green solvents, and we are working with our peers in the semiconductor climate consortium and our customers to help the industry adopt etch gases with lower global warming potential. We have taken many steps along our sustainability journey, but the rate of innovations can’t come quickly enough to meet the challenges before us.

AI is a game-changer

Historically, when looking for material innovations, scientists turn to the periodic table. In the 1800s, scientists carefully measured atomic properties, used the collected data to make models, and came up with predictions. Mendeleev prediction of ekasilicon, now known as germanium, was a major feat of this early data science approach. Those early material discoveries fueled by data, and all of the subsequent discoveries that built upon them, played a part in creating today’s digital era (the first transistor, incidentally, was made from germanium).

Today, the process of material discovery is still the same data-driven enterprise, but AI and machine learning (ML) tools facilitate a step-change in materials research needed to address the complexity of chemistries we must use. We have shown that our data-driven models and predictions enable a more rapid discovery cycle, for example, 50% faster for formulations in R&D.

We see an explosion of data and new models in material science, and these huge datasets enable rapid in-silico screening of performance and sustainability. For example, we were able to build a database of 65,000 potential etch molecules and models that predicted both the etch performance and the global warming potential of a gas. This process has already led to the discovery of new etch gases in our R&D lab, and we are working to couple generative AI to models like these to provide us with an even broader pipeline of novel candidates.

Fostering collaboration is key

Sustainability in our industry is a collective enterprise. The interdependency of sustainability on material properties, semiconductor manufacturing processes, and device performance and use mean that we must foster collaboration between suppliers, manufacturers and other stakeholder through digital platforms, such as Athinia. Here, all partners can share systematic, reliable and harmonized sustainability data in a secure environment. By correlating material and process data, AI can predict the performance and the full life-cycle view of sustainability. Material suppliers, for example, can screen numerous alternatives, based on data and digital models, to improve the speed and cost-effectiveness of bringing new, more sustainable materials to market.

The crux of the issue really is that: the semiconductor industry requires a new way of thinking and collaborating. If we transform the way we operate across the entire value chain to a data-based and increasingly collaborative approach, we can already improve the manufacturing process, speed up material innovation and, ultimately, push forward sustainability across the semiconductor value chain. The key to reducing our industry’s environmental footprint is using the digital discovery tools we have at hand and collaborate with customers and partners. It is our responsibility as industry leaders to rethink our business to ensure that doubling production does not mean doubling our environmental footprint.