A considerable part of my work involves reading, asking questions and speaking with people across different industries. I enjoy that exchange. Research offers one perspective; a conversation with someone dealing with the practical realities of a business offers another. Bringing the two together helps me understand not only what is changing, but why some developments deserve closer attention than others.

Much of what I take from those conversations stays in my notes. Occasionally, however, a set of observations seems worth bringing into a wider conversation rather than keeping to myself. The responses to my recent article, “Confirmed: AI Has Become Inflationary”, encouraged me to do that here. I have been considering which sectors stand to benefit from AI development and the economy forming around it. Several have held my attention, for different reasons. I thought you might find those reasons interesting too.

What follows is my view of nine of them, considered through their place in the AI economy rather than their prospects in the economy generally. I’m sharing this, not as financial advice, nor as prediction of share prices, but as an invitation to look at these sectors with me, consider the reasoning and draw your own conclusions.

CORE IDEA: The stars rate how strongly I believe each sector can retain lasting economic value from AI through 2030. They do not rate the sector’s overall prospects or the amount of investment it may attract. The scale runs from one to three stars, and the sectors discussed here span that full range. The rating is only a summary—the reasoning behind it matters more.

Cybersecurity | ★★★

Cybersecurity is my strongest conviction because the opportunity extends well beyond the companies building AI. As I argued in “Confirmed: AI Has Become Inflationary”, autonomous agents will increasingly interact with systems whose owners may never have chosen to adopt them. Some of those interactions will be legitimate; others will be malicious, as recent reports have repeatedly demonstrated. Large companies, small businesses and individuals will therefore need to prepare often-outdated systems both to process agent-generated requests safely and to withstand AI-enabled attacks. I expect demand for security capabilities to spread across the wider business and consumer landscape, rather than remain confined to the AI industry.

I see opportunities not only for established security providers, but also for businesses that help customers manage autonomous systems safely. As companies delegate more actions to agentic AI, they will need stronger permissions, monitoring, identity controls and containment. Meanwhile, organisations and individuals typically outside the AI sphere will need to invest in cybersecurity as agents become more social: initiating conversations, making requests and attempting transactions across systems they do not own. The challenge will be to enable legitimate interaction without opening the door to manipulation, fraud or intrusion.

Hardware development | ★★½

Hardware development presents a distinct case. The market remains transfixed by silicon innovation, but my conviction here rests on a more basic dynamic: physical scarcity rather than technological novelty. Hardware leaders such as NVIDIA and AMD may benefit from exceptional products, but the wider opportunity lies in the mechanics of expansion. Chips, advanced packaging, memory, cooling and power equipment are all needed to build IT infrastructure at a historic scale, yet their supply cannot grow at software speed.

Software can be distributed almost instantly. Physical infrastructure cannot. Semiconductor plants, electrical substations and supply chains take years to expand. As long as demand grows faster than this capacity, limited supply will support prices and keep the hardware market strong. I do not expect the shortage to last forever: investment in new capacity will eventually reduce these constraints.

Medicine | ★★

Medicine offers a different value proposition. AI can significantly accelerate scientific discovery, improve imaging and documentation, and make better use of scarce clinical capacity. But technical performance alone is not enough. Before reaching the market, it must produce clinically validated outcomes, secure regulatory approval and earn the trust of medical professionals and patients. The hurdles that slow adoption can therefore also protect validated solutions from competition.

Investment banking | ★★

What interests me about investment banking is not primarily its use of AI, but its role in financing the AI build-out. Data centres, chip production, power grids, new generation capacity and industry consolidation all require debt, equity, project finance and advisory services. Banks can therefore benefit from the expansion without owning the underlying infrastructure. The two-star rating reflects that opportunity, but also its cyclical nature: if infrastructure spending slows, financing and advisory activity will slow with it.

Nuclear power | ★½

Can nuclear power move from bespoke mega-projects towards an off-the-shelf product? That is the question that interests me.

My interest centres on small modular reactors (SMRs) and their potential to change the economics of nuclear power. Rather than relying exclusively on bespoke mega-projects, developers could manufacture smaller, standardised units and add capacity as demand grows. X-Energy’s plans for industrial sites and data centres illustrate the model: reactors located alongside large facilities, with additional modules installed when needed. The attraction lies not simply in building a smaller reactor, but in turning nuclear generation into a repeatable industrial product.

China offers another test of that model. Reuters reported in December 2025 that the Linglong One SMR in Hainan was targeting commercial operation in the first half of 2026. Even so, the commercial case remains unproven. Permitting, financing, construction costs and electricity prices will determine whether SMRs can compete at scale. “Off the shelf” describes a possible direction, not an established market. The one and a half star rating recognises the value of existing nuclear assets and the potential of SMRs, while acknowledging that this new commercial model has yet to prove itself.

Foundation models | ★½

Foundation models will remain essential to the AI economy, but that does not make the sector broadly attractive. Developing frontier models requires enormous and continuing investment. At the same time, smaller models, open alternatives and the ability to switch between providers give customers more choice and limit pricing power. The risk is not simply that a model falls behind, but that competing models become good enough to make it interchangeable. A small number of providers may still build exceptional businesses through scale, distribution and deep customer relationships. My caution applies to the sector as a whole: capabilities may keep improving even as the models themselves become harder to differentiate and monetise.

Entertainment | ★½

In entertainment, lower production costs will not automatically produce higher profits. AI can make production, localisation and visual effects faster and cheaper, but it also makes it easier for more creators to enter the market. Content supply may therefore grow much faster than the audience’s capacity to consume it, increasing competition for attention. That dynamic is likely to favour businesses that already control valuable intellectual property, distribution or trusted audience relationships, while putting pressure on undifferentiated producers. Strong franchises may benefit, but their advantage should not be mistaken for attractive economics across the sector as a whole.

Academia and education | ★½

AI could create enormous value in education by expanding access to tutoring, improving feedback, supporting research and helping people with different learning needs. My caution is commercial rather than educational. Schools and universities often have limited budgets, slow procurement processes and strict requirements around governance and assessment integrity. These constraints may delay adoption and limit what institutions can pay, even when the technology delivers clear benefits. Education may therefore generate substantial social value from AI without producing equally strong financial returns for providers. That distinction explains the one-and-a-half-star rating.

Commercial banking | ★

AI has clear uses in commercial banking, from fraud detection and compliance to customer service, document processing and credit assessment. My caution concerns how much of that potential will translate into higher profits. The projected savings from automation often look compelling at first, but diminish once banks account for integration with legacy systems, security, validation and the handling of exceptions. Competition may then force banks to pass much of what remains on to customers through lower prices or better service. AI may therefore become essential for staying competitive without materially improving margins.

An invitation to compare notes

The central question is not whether AI will create value, but who will be able to retain it. Across these sectors, the strongest positions are protected in different ways: by scarce physical capacity, regulatory approval, trusted distribution, valuable intellectual property or direct customer relationships. None is permanent. New supply, failed execution, stronger competition or changing customer behaviour can erode each advantage. What matters is whether that advantage remains defensible as AI becomes more widely available.

These ratings are therefore starting points, not conclusions I expect anyone to adopt. I invite you to test them against your own research and experience, particularly in the sectors you know best. Where does the reasoning hold, and where does it fail? Hearing how others answer those questions is the reason I am sharing these observations.

Would a fresh perspective on your AI strategy be useful?

Get in touch. Whether you are clarifying your priorities or questioning your assumptions, I would welcome the conversation.

Selected References and methodological notes

This report is a strategic synthesis rather than a securities recommendation. It combines published research with my assessment of each sector’s capacity to retain economic value from AI through 2030. The rankings are comparative judgements based on the framework outlined in the reading note, while the sector interpretations and scores reflect my personal assessment. Forecasts are inherently uncertain, and quantitative claims are presented as reported estimates or scenarios rather than guarantees. The selected resources below represent some of the principal sources used in the research.