Valuing scarcity in the age of AI

As with any investment, your capital may be at risk. 

 

Artificial intelligence was the first topic in our Research Agenda last year, and it figures prominently once again. 

In the face of rapid technological advances and their potentially profound implications for economies and societies, how can the long-term stock picker position a portfolio for such an unusually wide range of outcomes?

 

The lens of scarcity

One way of making sense of transformations like this is to consider the opportunities ahead through the lens of scarcity. 

A hallmark of the biggest technology shifts is the transformation of scarcity to abundance. Of taking something that was once expensive, slow or difficult to access, and making it cheap, fast and widely available. This phenomenon has played out in particular across two major themes: information and energy. 

• Information – The printing press made written works abundant. It turned books from rare, hand-copied objects owned by a few into something that could be produced at scale and accessed by millions. That enabled a massive acceleration of knowledge sharing and learning. Communications revolutions have happened many times since, each time enriching interaction and information flow 
  
  
• Energy – More concentrated and efficient energy transmission has prompted similarly tectonic shifts in society, replacing human and animal labour with machines operating at vastly greater speed and scale. Sometimes these shifts are related to the energy sources themselves, such as the move from wood to coal to oil. Sometimes the shifts are related to how we have harnessed dense energy sources. The steam engine, the internal combustion engine and the jet engine all produced step changes in the movement of people and goods 

From abundance to bottlenecks

Each revolution delivered greater abundance to society, while shifting scarcity to new pinch points within the corporate system. Businesses that gatekeep the critical parts of the supply chain where these bottlenecks emerge become the giants of the day. The printing press shifted the bottleneck from book production to distribution, making publishers the new gatekeepers.

We are intrigued that AI abundance at scale will require step changes in both information and energy flow to train and run modern artificial intelligence systems. Large language models (LLMs) process vast amounts of data across thousands of specialised chips called graphics processing units (GPUs), which handle many calculations simultaneously. This creates an enormous demand for data movement, and for the electricity needed to power and cool these systems, binding the digital and physical worlds much more tightly than previous software waves.

We are focused on identifying valuable emerging scarcity. We must also consider which companies’ previously scarce capabilities will be rendered commonplace by the new tools that generative AI is already putting in our hands.

Physical scarcity

We are looking carefully at scarcity in the AI infrastructure buildout. Analysts expect the four major US ‘hyperscalers’ – the largest cloud and datacentre operators – to spend almost $700bn on capital investment in 2026, rising by 60 percent from already huge spending in 2025. Industry projections vary widely, but most expect annual global investment in the trillions of US dollars by 2030. 

The scale of capital spending required is such that only a few companies can afford the bill. We’re owners of these businesses, but are mindful of the impact on cash flow and the use of debt to fund this buildout. The return structures are far from clear, raising the possibility that some current investments may turn out to be poor ones. There may be important emerging cost and benefit differences, such as the in-house chips being developed by Alphabet and Amazon, which could moderate their reliance on expensive chip providers like NVIDIA. 

We are becoming more interested in the recipients of this investment cycle spending. We already own several advantaged suppliers, and we expect more to appear. For example: 

Memory 

Memory chips are likely to be in high demand for the next several years. High-bandwidth memory (HBM) is a type of ultra-fast memory particularly well suited to meeting the speed and power requirements of AI, as it allows large amounts of data to be moved quickly to the chip doing the work. This prevents the system from being held up while complex AI tasks are processed. These chips are in short supply. Chipmakers and buyers are signing longer-term contracts of three to five years (compared with one to two years previously) for memory chips, potentially changing the length of pricing cycles in this cyclical industry. 

We recognise that high prices and long contracts will stimulate new supply, and that innovative companies may find ways to work around supply challenges. But as agentic AI is much more memory-intensive than LLMs, its widespread adoption would outweigh these concerns. There is a step change coming in memory demand, and only a few companies can respond to that.

Processing 

Processing chips, such as NVIDIA’s GPUs and the systems that accompany them, remain in tremendous demand. Rapid advances in performance and energy efficiency currently easily justify short upgrade cycles as customers look to move to the latest technology. There is plenty of room for GPU demand to double again before the 2020s are out. However, investors often overlook the role that other chips will play in the mass deployment of AI that follows on the most intensive forms of model training.

To give an example, we are intrigued by the advent of ‘edge AI’, which means running AI processing on local devices, using locally collected data. Phones, cars and robots are obvious examples, and there could be many more. There are potentially very positive implications for mass-market processors, including application-specific chips (ASICs), central processing units (CPUs) and providers of analogue and power semiconductors.

Supply chain

Rising volumes of complex chips will boost demand at machine makers and materials providers serving the chip producers. The industry depends on highly specialised equipment and materials, and those niches are often occupied by dominant experts. While many industrial supply chains are localising, we think the performance demands of semiconductor equipment will keep this a largely global market, with global winners.

Datacentres 

Datacentres will power the AI revolution, but who powers them? Electricity supply is likely to be a significant bottleneck, even with advances in power efficiency, because each new generation of AI models requires more computing power, and therefore more energy, than the last. We are interested in the companies that can earn an uncapped return from a massive modernisation of electricity supply. This includes material producers, service providers, advantaged installers and many more.

To put this in context, the predicted electricity demand from global datacentres in 2030 is roughly equal to all the electricity produced by Japan in 2024.

 

The end of software scarcity?

Common wisdom predicts that AI may disrupt much of the enterprise software industry. Why pay for a myriad of business software systems when an AI agent with some simple plug-ins can build and maintain those systems for you?

Consumer software businesses, whose success arises from organising and aggregating networks of buyers and sellers, have also found themselves under pressure. Where’s the value in an online superstore, when your AI assistant can tour the shops for you to find the best deal? The online disruptors may become the disrupted.

We own enterprise software companies that have grown by helping customers replace big upfront spending (such as building in-house software and buying servers) with a flexible operating cost that scales with their businesses. We have backed several ecommerce businesses that offer convenience and choice to customers, often accompanied by companion logistics and financial arms. We must be willing to reconsider the value proposition each business offers relative to emerging AI competition.

AI has the potential to turn scarcity into abundance in individual industries. We are already seeing this in software coding, education businesses and creative media. When this happens, users tend to capture most of the value.

In those instances, scarcity may currently be relocating from the software plane to the LLMs. It is not yet clear to us how much of the long-term value LLMs will capture. However, there is a credible argument that a few dominant providers may earn a valuable royalty from this new layer, which both intermediates online knowledge and generates new content.

We think only a select group of businesses will adapt to the new paradigm. The most durable businesses will keep improving what they do to retain scarcity value or benefit from the significant practical barriers to introducing new supply. Our thinking must remain flexible and recognise the value of businesses that will benefit from a slower pace of change than implied by stock market swings.

 

Theses for scarcity

• Trust and accuracy – Businesses that provide verified data and security to their customers may become more valuable in an AI-intermediated world. This applies to companies such as the payments businesses Mastercard and Adyen
• A physical edge – Effective integration with the physical world is a challenge. Businesses such as Samsara, which links physical assets into a monitoring system, or the factory automation business Keyence possess powerful edges. So too do ecommerce platforms that have expanded to fulfilment and logistics 
• Adaptability – Companies that use AI to improve what they do and rethink how they charge customers may unlock significant value. For example, Cloudflare, which began as a website security provider, could evolve into a gatekeeper of online content, helping creators earn returns from AI models that scrape the web 

When considering the portfolio’s most direct exposures to the AI revolution, we see a broadly encouraging picture.

High levels of scarcity and low risk of being bypassed appear to be the most secure growth opportunities, but there will be substantial rewards for companies closest to the end-user that can stay ahead of the technology curve.

There is obvious resilience in the physical parts of the chain, and a wider range of possible outcomes among the companies we own that are closer to the AI layer that sits between users and the internet/software. We will continue to test our assumptions and the alternatives our global stock universe offers. This is among the most exciting opportunities to deliver value to clients that we will encounter in our careers.

 

---

Risk factors

The views expressed should not be considered as advice or a recommendation to buy, sell or hold a particular investment. They reflect opinion and should not be taken as statements of fact nor should any reliance be placed on them when making investment decisions.

This communication was produced and approved in April 2026 and has not been updated subsequently. It represents views held at the time of writing and may not reflect current thinking.

Potential for profit and loss

All investment strategies have the potential for profit and loss, your or your clients’ capital may be at risk. Past performance is not a guide to future returns.

This communication contains information on investments which does not constitute independent research. Accordingly, it is not subject to the protections afforded to independent research, but is classified as advertising under Art 68 of the Financial Services Act (‘FinSA’) and Baillie Gifford and its staff may have dealt in the investments concerned.

All information is sourced from Baillie Gifford & Co and is current unless otherwise stated.

The images used in this communication are for illustrative purposes only.

 

Important Information

Baillie Gifford & Co and Baillie Gifford & Co Limited are authorised and regulated by the Financial Conduct Authority (FCA). Baillie Gifford & Co Limited is an Authorised Corporate Director of OEICs.

Baillie Gifford Overseas Limited provides investment management and advisory services to non-UK Professional/Institutional clients only. Baillie Gifford Overseas Limited is wholly owned by Baillie Gifford & Co. Baillie Gifford & Co and Baillie Gifford Overseas Limited are authorised and regulated by the FCA in the UK.

Persons resident or domiciled outside the UK should consult with their professional advisers as to whether they require any governmental or other consents in order to enable them to invest, and with their tax advisers for advice relevant to their own particular circumstances.

Financial Intermediaries

This communication is suitable for use of financial intermediaries. Financial intermediaries are solely responsible for any further distribution and Baillie Gifford takes no responsibility for the reliance on this document by any other person who did not receive this document directly from Baillie Gifford.

 

193626 10062282