Scottish Mortgage deputy manager review

Hardware

The hardware layer is about making the physical computational devices that enable AI. In this layer, since 2016, we have owned NVIDIA, the leading designer of AI chips. The company has a dominant position, with 90 per cent of all generative AI models trained on its chips. It is the critical enabler of AI.

However, NVIDIA only designs its chips, it needs others to fabricate them. For this, it uses TSMC, the world’s largest integrated circuit manufacturer and a recent addition to the Scottish Mortgage portfolio.

It has a dominant position in an industry where scale matters with the latest foundries, such as the three it is building in Arizona, costing over $65bn. These chip foundries are so critical to supply chains and the economy that they hold geopolitical significance. Consequently, the US government is providing over $10bn in federal grants and loans to ensure they are built in the US.

TSMC can be thought of as a royalty on global computing power, just as NVIDIA can be thought of as a royalty on artificial intelligence. To help make chips, TSMC requires a particularly crucial piece of equipment: lithography machines.

For these, it relies on another of our longstanding holdings, ASML, which has a monopoly position in advanced lithography machines. These rely on the world’s flattest mirrors and one of the most powerful commercial lasers to create an explosion 40 times hotter than the surface of the sun to pattern tiny shapes on silicon that measure just a few nanometres.

This precision is what allows chips to be made containing tens of billions of transistors. To different extents, all three of these hardware companies benefit from the rise of AI while possessing dominant and near-impregnable competitive positions.

The founder of OpenAI, Sam Altman is understandably evangelistic and has gone as far as to say that computing power may become the most precious commodity in the world. This is because the demand for computing power and intelligence could be effectively limitless with demand progressively unlocked by supply at ever lower price points.

Crucially we are seeing a situation where the cost to serve continues to rapidly fall leading these three hardware companies to face a large structural opportunity even if it will likely remain a bumpy cyclical journey along the way.  

Infrastructure

The next layer is infrastructure. Here, we have the cloud service providers that buy NVIDIA’s chips and offer scalable, on-demand access allowing companies to train and deploy AI models without the overhead of building their own infrastructure.

In essence, the cloud service providers democratise access to both computing and AI. There are three dominant cloud service providers. We own Amazon, which operates Amazon Web Services, the largest cloud service provider in the world.

There are also companies that are building large foundational models for AI. Foundational models are trained on broad datasets (ie large swathes of the internet) that can be used to perform a wide variety of tasks. They are also commonly used as the base upon which to build more specialised AI models. These foundational models have become so large and complex that the computing power and energy required to train them is making them increasingly expensive.

The foundational models expected to be released later this year will likely have cost close to $1bn to be trained. That cost is expected to rise to between $5bn and $10bn for the latest models in 2025 and 2026, which is indicative of the growing demand for hardware companies.

A consequence of this is that the ability to create such models is fast becoming the preserve of just a handful of mega-scale companies and those receiving their patronage. We have several holdings developing foundational models, such as Meta Platforms, Amazon and NVIDIA.

Another set of companies providing the infrastructure for AI are database companies. The explosion of data in the world has meant that more data will be created in the next three years than was created in the last thirty years.

Companies that Scottish Mortgage owns, such as Databricks and Snowflake, help businesses store, manage, and use that data in the cloud. That same data is also the lifeblood of AI, with companies increasingly looking to feed their data into foundational models, allowing the creation of powerful new applications specific to their business.

For example, commerce software giant Shopify has combined its proprietary data and merchants’ data with OpenAI’s GPT to create what it calls Sidekick, a conversational assistant that merchants can talk to and ask questions about how to use Shopify’s platform. It can even be asked to accomplish specific tasks, such as compiling reports on bestselling products. AI will allow companies to do more with their data and, in doing so, increase the value of those that offer tools to store, parse and effectively use data.

Applications

The final layer is the application layer. This is about making productive use of AI in the real world. A significant number of Scottish Mortgage’s holdings are making use of it to expand addressable markets, reduce costs and dig deeper competitive moats.

Tesla is using AI to make its cars self-driving and even hopes to leverage those advances to produce humanoid robots. After all, what is a self-driving car if not a robot operating in the physical world? Demonstrative of its progress is that Tesla cars have now driven 1.3bn miles autonomously, and the company is already in conversation with another major carmaker about licensing its self-driving technology.

Recursion Pharmaceuticals is leveraging AI to improve drug discovery by creating a map of human biology that could dramatically cut the cost of developing new drugs. Tempus Labs has built a vast database of over 7 million cancer patients’ clinical records and is applying machine learning to that dataset to enable physicians to make better treatment decisions.

Meta Platforms is using AI to improve advertising targeting across its platforms such as Facebook, Instagram and WhatsApp to powerful effect. Spotify is using it to enhance its personalised song recommendations and has released its AI-powered DJ to much fanfare. AI also enables podcasts on Spotify to be translated into other languages using the actual voice of the podcaster.

The impact of AI across the portfolio is vast because all companies can benefit from the application of intelligence. AI can be thought of as a powerful new set of tools for companies to apply to their business that will, crucially, only get significantly better with time.

The companies that will be best placed to use this toolkit will be those with large amounts of proprietary data, software expertise and a culture of innovation. In each of these dimensions, our portfolio companies should be well-placed.

The opportunities of the application layer in new technology paradigms naturally lag behind those in the hardware and infrastructure layer. Those initial two layers need to scale first in order to support the development of applications. It can also take time and human ingenuity to find ways to leverage and apply powerful new technologies.

We can draw an analogy with the smartphone. When the iPhone was released, it was clearly an impactful piece of hardware. Still, it took time for companies and aspiring founders to build applications to utilise the new device’s potential fully. At its release in 2007, it would have been hard to immediately predict the new business models it would go on to enable, such as ride-hailing, food delivery, mobile payments and short-form video apps such as ByteDance’s TikTok. It even took time to appreciate just how meaningful it would be for existing digital activities such as ecommerce, streaming, and social media, which saw their market opportunities greatly enlarged.

Over time, the benefits of AI are likely to similarly expand to a greater number of companies and lead to new business models. After all, there is no point in investing billions in hardware and infrastructure if there are not a lot of applications to be built.

Conclusion

Despite the excitement surrounding AI, it is still important to remember that progress is rarely a straight line. We cannot rule out that there could be a period of digestion following heavy investments in the hardware and infrastructure layer as companies take longer than expected to work out how to use new capabilities.

Alternatively, we could encounter unexpected limitations to AI models requiring new algorithmic breakthroughs to be made. We are cognisant that the hardware companies, in particular, though currently propelled by insatiable demand, can be viciously cyclical businesses should they hit air pockets of demand.

We continue to expect them to perform well but, partially in recognition of this cyclicality, have been making some mild reductions. This has been the case for our largest holding, NVIDIA, following exceptionally strong performance and having grown its net income by over 500 per cent last year.

Overall, we still believe we are early in experiencing the impact of artificial intelligence. That impact has been most strongly felt in the hardware and infrastructure layers, but it should gradually expand to the application layer. The role of Scottish Mortgage will continue to be to invest in and support progress, and the developments in AI provide robust evidence that progress is continuing at pace.

On behalf of our shareholders, we invest in transformative change. We’re greatly encouraged by the founders of our portfolio companies telling us that, to them, the pace and magnitude of technological-driven change has never appeared greater. The possibility of such change is a crucial enabler of the outlier outcomes we seek, and aim to deliver for our shareholders.

Risk factors

The trust invests in overseas securities. Changes in the rates of exchange may also cause the value of your investment (and any income it may pay) to go down or up.

Unlisted investments such as private companies, in which the Trust has a significant investment, can increase risk. These assets may be more difficult to sell, so changes in their prices may be greater.

The Trust invests in emerging markets, which includes China, where difficulties with market volatility, political and economic instability including the risk of market shutdown, trading, liquidity, settlement, corporate governance, regulation, legislation and taxation could arise, resulting in a negative impact on the value of your investment.