Advanced materials and fine chemicals don’t appear to have the same ‘cachet’ of some of the hotter innovation areas such as metaverse etc. because at first glance they are not direct plays on consumption. They also don’t have the scaling and networking effects that ‘successful’ digital platforms retain. However, the sector is experiencing some of the similar benefits of AI that the pharmaceutical industry is going through – namely reducing the time spent on R&D while increasing the probability of successful product launches.
A recent FT article (see AI to drastically cut time to develop new battery materials, say executives) pointed to evidence by Belgium’s Umicore and Sweden’s Northvolt that machine learning was allowing AI generated patents to be filed much faster. For example, Microsoft in co-operation with the Pacific Northwest National Laboratory is using AI and high-performance computing to narrow down 32 million materials to just 18 candidates for use in batteries in around eighty hours. The fact that the progress in AI is coinciding with the increased automation of experiments by robots is allowing huge volumes of experiments to be undertaken. Presently, the limiting factor is computer processing capacity.
Microsoft’s longer-term vision is to compress to 250 years of chemistry research into the next 25 years using Azure Quantum Elements -combining high-performance computing (HPC), artificial intelligence, and quantum computing. Microsoft claims that ‘Azure Quantum Elements can expand the search space for new materials from thousands to tens of millions of candidates. It also speeds up chemistry simulations by an astounding 500,000 times, compressing a year-long simulation into a single minute’.
Not to be outdone, Google DeepMind have used AI to discover approximately 2.2 million crystal structures. The number of novel materials has been identified by using Google’s AI tool known as GNoME. In a paper published by Nature (see Scaling deep learning for materials discovery, Nov. 23) the authors highlighted that following on from 48,000 stable crystals identified in ongoing studies, AI had enabled the discovery of 2.2 million structures, many of which escaped previous human chemical intuition- an order-of-magnitude expansion in stable materials known to humanity. Of the stable structures, 736 have already been independently experimentally realized.
The model used by Microsoft and Google to combine its capabilities in AI and computer processing coupled to ventures with laboratories is probably how the materials industry will be ‘shaped’ over the next few years. The incumbents will need to ramp up their processing/AI capabilities probably by forming ventures with other Tech giants. Certainly, the ability to file patents quickly will be a defining feature of this industry. One bonus would be that R&D budgets will become not only more focused but also be reduced.
The evolution of the EV battery space is an interesting example of how newly developed materials were brought into mass production quickly. If the above examples hold true, then innovation cycles outside of the auto industry such as in semiconductor or construction materials could be accelerated far more than investors are discounting. Equally, the probability of R&D costs being written off through misconceived ideas should also be dramatically lowered.
The equity market materials sector sits awkwardly within the metals, packaging and building materials sector in benchmark indices. This betrays their growth in innovation that is occurring but also means that they are labelled as ‘cyclicals’. They are considered as responding to macro forces when in fact many of them are monopolies (argon gas, ceramics) in the semiconductor industry, but they also have widespread healthcare applications which mean they are defensive in nature. The applications in aerospace and autos have defined the materials sector to date but battery and electricity storage would be impossible without the use of alloys and highly refined metals.
There is a wide population of stocks to sift through which can be valued on their free cash-flow and tangible book value unlike many of the media and entertainment companies. Moreover, the long duration of investment means that there is a plethora of stocks in the unfashionable European and Japanese markets. The producers of refined and advanced materials are oligopolies which sometimes sit inside larger chemical or mineral companies meaning their book value can be under-represented. Analyst coverage has fallen as the bias has been towards the conventional technology companies – semiconductors (hardware) and software. The disruption from AI and high-performance computing should mean that the incumbents will come under pressure as their previously perceived ‘wide moats’ are bridged by a wave of patent filings usurping their leadership.
