AI: intelligence everywhere

As with any investment, your capital is at risk. This article originally featured in Baillie Gifford's Spring 2024 issue of Trust magazine. Read all the articles here.

In 2018, I spent two days with Prof Brian Arthur, one of the world’s leading thinkers. He literally wrote the book on the nature of technology. During this time, he told me: “As a growth investor, you should be really interested in the developments in AI.”

I replied: “I completely agree. I think it is a new computing paradigm. In other words, the next big thing after the mobile internet era.” Brian looked at me and said: “Well, I think it’s the biggest thing since the Gutenberg printing press in the 15th century.”

In the medieval era, scribes painstakingly copied books by hand, and monasteries chained the works to desks to prevent the theft of the valuable tomes. But Gutenberg’s invention let people share information like never before. It democratised access to knowledge and, in doing so, powered the scientific revolution.

But where the printing press externalised information, AI is externalising intelligence and making it available around the world, instantly and on demand. The impact of intelligence is even greater than that of information. For us, AI is not a new industry but a tool with the potential to rearchitect every industry.

Technology’s trendlines

At Scottish Mortgage, we have long tried to step back from the noise of the financial markets. As Bill Clinton used to say, focus on the trendlines, not the headlines. To us, the trendlines that matter are the deep underlying changes in technology. They ultimately drive economic growth, human progress and, we believe, long-term returns for our shareholders.

The trends in computing technologies have, perhaps, been the most profound of all. At a high level, these have followed Moore’s Law: a rule of thumb that predicts computing power doubles roughly every two years for the same price. That might not sound that impactful because we all know computers get better and faster over time. But a doubling every 24 months adds up.

Within 12 years, it equates to more than a 60-times increase, which then becomes industry and world-changing. It’s why Amazon and other companies that leverage digital technologies keep getting stronger. It’s why Apple’s top iPhone has 2.8 billion times more processing power than Apollo 11’s computer, which took astronauts to the moon.

By following the implication of Moore’s Law, it became clear to us that many digital companies would improve dramatically over the last decade, whether that was Amazon offering cloud computing, MercadoLibre providing digital financial services to millions of Latin Americans or Spotify recommending songs that match your taste.

A new era

Artificial intelligence is accelerating these trends. I began grappling with AI’s potential impact about eight years ago when I met the founder of Baidu, China’s leading search engine.

If I’m honest, at one meeting I felt he was not finding my questions hugely interesting. So I asked: “What are you spending your time on and what interests you these days?” At this point, he lit up and started talking about AI. This was my first real education about its potential in terms of a possible new computing era.

Despite many further conversations and much thinking about the topic over the following years, recent progress has surpassed our expectations.

I met Karim Beguir, co-founder of one of Europe’s leading artificial intelligence startups, last year. He remarked that AI’s rapid advancement was eating Moore’s Law for breakfast thanks to three inputs, all of which are growing exponentially:

• Hardware: NVIDIA’s GPUs – the initials stand for graphics processing units, but you can think of them as computer chips well suited to AI – have been doubling in performance not once every 24 months but once every 12 months.
• Data: the more an AI model has to learn from, the better it gets. About a decade ago, the leading models trained on about 10 billion data points. Today it’s over one trillion. That’s a 100-times increase.
• Improvements in algorithms: the better you design your AI model, the more efficient it becomes. Efficiency is improving so fast that if you have an AI image recognition model – say one identifying different animal species in photos – every nine to 16 months you can achieve the same accuracy in recognising those images with half the ‘computational budget’, ie half the hardware.

NVIDIA chips

What does this mean for Scottish Mortgage?

Firstly, we expect more long-term demand for computing power, which means a lot more hardware will be required. NVIDIA is already experiencing explosive demand. Its revenue guidance for last year’s fourth quarter was $20bn, which compares with $7.6bn just a year ago.

It’s very early to start predicting exactly what AI-driven applications will be the most successful, but we can say that NVIDIA’s chips are likely to train them. Today, its products are involved in an estimated 90 per cent of generative AI, including the creation of text, images and music at human-like levels.

Scottish Mortgage first invested in NVIDIA eight years ago and stuck with it despite contending with a 66 per cent peak-to-trough share price drop in 2022, when many others sold out. I met the firm’s co-founder and chief executive Jensen Huang last year and was struck by his observation that there would have been huge demand for NVIDIA’s hardware to accelerate computing in data centres even if the new AI workloads hadn’t supercharged it.

This will also increasingly benefit key semiconductor industry suppliers, including ASML. It is the only firm capable of making a key machine that enables the manufacture of the latest cutting-edge chips. Then we have those that provide the cloud infrastructure upon which AI applications run, including Amazon Web Services, and those providing related services, such as Snowflake and Databricks.

The road to robots

We also own companies that are developing core business solutions using AI: the most valuable part of Tesla may not be its electric cars but its autonomous driving system. Its vehicles have already driven over 150 million autonomous miles, with US consumers paying Tesla an additional $15,000 for the service. Our valuations don’t account for Tesla’s humanoid robot efforts. Still, I find it telling that AI scientists note that if the company cracks autonomous cars, it has a credible pathway to building all kinds of robots. After all, what is an autonomous car if not a robot operating in the physical world?

Gutenberg’s printing press unlocked immense value by making it easy to disseminate human knowledge, causing a cascade of beneficial second and third-order effects. AI could be the modern equivalent for intelligence, increasing our understanding and capabilities to a degree and scale that are hard to fathom.

As a long-term growth investor, part of our role is to try to imagine different possible futures. However, I struggle to imagine a more impactful or exciting investment opportunity than artificial intelligence is presenting us with.

Investments with exposure to overseas securities can be affected by changing stock market conditions and currency exchange rates.