As with any investment, your capital is at risk.
Something profound is happening in the software industry.
Shopify founder and chief executive Tobi Lütke recently had an MRI scan as part of his annual health check. This time, instead of downloading expensive, bespoke viewing software to see the results, he asked Anthropic’s Claude Code to handle the raw MRI data.
With a single prompt, he told the system to find every report and image, convert them into usable formats, organise what mattered into a structured folder and build an index page so he could explore the results in his browser.
Minutes later, he had a web-based viewer.
One more prompt, and it began to annotate what it saw.
“My annual MRI scan gives me a USB stick with the data, but you need this commercial windows software to open it.
— tobi lutke (@tobi) January 11, 2026
Ran Claude on the stick and asked it to make me a html based viewer tool. This looks... way better.”
If a broadly trained AI system can, on request, assemble and interpret a highly specific medical imaging tool, the idea that specialist software enjoys a defensible moat starts to look fragile.
For long-term growth investors, it raises a key question:
Where do value and vulnerability sit in the software industry in the age of AI?
From “software eats the world” to “will AI eat software?”
For the last decade, many investors have been guided by tech entrepreneur and investor Marc Andreessen’s claim that “software is eating the world” – the idea that digital tools, enabled by software-as-a-service (SaaS) and the cloud, would penetrate almost every industry.
Now a new narrative has taken hold: that AI will eat software itself. The share prices of many software companies have lagged both AI beneficiaries and the broader US market in recent years, as investors worry that their products could be replicated by AI systems or undercut by new, AI native competitors.
AI leaders surge while software stocks fall behind (2023–2026, rebased)
The 'AI leaders' refer to the seven largest AI beneficiaries: NVIDIA, Microsoft, Apple, Amazon, Alphabet, Meta and Broadcom. Equal-weighted index monthly rebalancing. Price returns in USD relative to the S&P 500.
Source: Refinitiv Datastream
Capital flows help illustrate what is happening. At one end of the spectrum are the hyperscale cloud platforms, such as Amazon Web Services and Google Cloud. They are committing vast sums to data centres, custom chips and energy to train and run large AI models. A significant share of this spending flows to NVIDIA and other chipmakers, which provide the core computing power and systems on which AI depends.
At the other end are the model builders and applied AI companies, many of which are still private. Since late 2022, they have attracted a growing share of global venture capital.
Our investment in Anthropic, for clients who can hold private companies, is part of this trend. Its Claude models, and particularly Claude Code, are designed to dramatically reduce the time and cost required to turn an idea into working software.
Anthropic’s revenues have grown rapidly, with revenue per employee far ahead of many established software leaders, highlighting the extraordinary efficiency that AI native business models can achieve.
We think, however, that the idea of ‘AI eating software’ is misleading.
Software is not disappearing, but the balance of value within the system is shifting. The key task is to judge whether a software company’s role in its customers’ workflows will remain critical once AI agents sit between users and applications.
The new rules of software
Historically, the big shift in enterprise software was the move from on-premise installations to the cloud. Today, the dividing line is between businesses on the right and wrong side of AI. Three changes look particularly important:
Code is being commoditised
As Claude Code and similar tools improve, the act of writing, testing and deploying software becomes cheaper and more automated. Value shifts from the code itself to how it is applied. Data and integration into customer organisations become the moat.
The scarce skill is no longer simply the ability to ship a feature. It is the ability to define the right problem, apply AI effectively and embed the result deeply into a client’s processes.
Pricing models come under pressure
Much of the software sector grew on seat-based pricing, where revenue scaled with the number of employees using a product. If AI enables clients to achieve the same or better output with fewer staff, that model looks fragile.
We expect a gradual shift towards usage and consumption-based pricing, where customers pay for compute, actions or transactions. That transition is healthy for products with a clear value proposition, but it exposes those whose economics depended on ever-rising headcount.
Fragmented tools are being squeezed
AI agents need clean, well-governed data and reliable scaffolding. A patchwork of surface-level tools, each with siloed data and workflows, becomes a problem. This favours platforms that act as systems of record, sitting at the centre of an organisation’s data and processes, over superficial applications that can be bypassed.
In Tobi Lütke’s MRI example, the Claude model did not care about any particular user interface. It orchestrated files and open-source tools behind the scenes. It is a telling illustration of how AI could interact with other applications: as building blocks rather than using the actual product itself.
Moats in an AI world
In this new paradigm, software moats depend on where a product sits. Enterprise tools embedded in core workflows are hard to replace. Mid-market tools built for speed and simplicity, such as automated mailing or messaging services, are far easier to swap out.
It also matters whether software is horizontal or vertical, and how it charges. Horizontal platforms used across industries will be affected differently from vertical, single-industry tools. And seat, subscription and consumption-based models will each feel the impact of AI in different ways.
We are not complacent, but in our US Growth portfolio, the software holdings often look better placed in an AI world, and typically have one or more of the following:
- Deep systems of record – a single source of truth for business data. Workday in HR and finance or Snowflake in data, for example, hold authoritative versions of critical information. AI agents may bypass their screens, but they still need the trusted data these platforms contain.
- Proprietary, hard-to-replicate data – such as Samsara’s physical sensor network across industrial fleets and sites, or Datadog’s telemetry from complex cloud systems. Generic models cannot easily recreate this context without access to the same streams.
- Infrastructure and orchestration roles – platforms such as Amazon Web Services, Google Cloud and Cloudflare, which provide the compute, tools and edge network that other applications depend on. As AI workloads scale, their centrality to the ecosystem should increase.
- Ecosystem leadership – NVIDIA’s combination of specialised chips, CUDA software and developer loyalty, for example, has made it a default environment for training and serving advanced models.
By contrast, the software companies that appear most vulnerable tend to share different traits. Their revenues are tied to seat counts, their workflows are relatively simple, and they lack proprietary data or high switching costs.
Some will adapt by shifting pricing and deepening integrations; others may see growth, margins and valuations come under sustained pressure.
Where the value shifts
AI is compressing the cost of writing code and challenging many traditional software moats, while simultaneously increasing the importance of the right infrastructure, data, and deeply embedded systems.
For long-term investors, the central task is to distinguish between businesses whose advantages are being eroded and those whose roles in the software stack are becoming more mission-critical.
We remain convinced that companies at the heart of critical workflows, with valuable, hard-to-replicate data and technical teams willing to embrace AI at pace and scale, can continue to offer attractive growth prospects in the years ahead.
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 February 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.
188109 10060821
