What's happening

It is hard to find a corner of the market that has not been shaped, in some way, by artificial intelligence over the past 18 months. Hyperscalers are pouring tens of billions of dollars a quarter into data-centre build-outs, electricity Demand projections in the United States have been rewritten, and the share prices of the companies that Supply the picks and shovels of the AI boom have produced eye-watering returns. For UK investors, the question is no longer whether to pay attention but how to think about valuation, concentration and longevity in a theme that has rewritten the rules of the market.

Four names sit at the heart of the AI conversation: Nvidia, Broadcom, AMD and Taiwan Semiconductor Manufacturing Company. Each plays a slightly different role in the value chain, each carries a different risk profile, and each is priced very differently by investors today. The figures from the supplied FT Global 500 sheet provide a useful objective starting point for any analysis.

Markets have become accustomed to AI being the dominant macro narrative. That makes it especially important to understand what the data actually says about these businesses, where they sit relative to their 52-week trading ranges, and what kind of Dividend support, if any, exists when sentiment turns.

The companies in focus

Nvidia is the headline act. The company designs the graphics processing units that power virtually every cutting-edge AI Training cluster, and its data-centre Business has become the single most-watched line item in global technology Earnings. Broadcom plays a more diversified role, supplying networking silicon and custom AI accelerators to the largest cloud providers, alongside a substantial software Franchise. AMD is the long-time challenger to Nvidia in graphics and to Intel in central processors, and its data-centre product roadmap has improved the company's position in AI compute.

Taiwan Semiconductor is the linchpin. Almost every advanced chip used in AI today is fabricated by TSMC, which means that without TSMC's leading-edge foundries the entire AI hardware industry would grind to a halt. The company is therefore not a beneficiary of one customer or another so much as a beneficiary of the whole AI Capital Expenditure cycle.

There is also a wider supporting cast in the supplied figures, including ASML, which makes the lithography machines that TSMC uses, Applied Materials, Analog Devices, Arm Holdings and Micron Technology. Each is exposed to AI in its own way, but our focus here is the four core names.

Why this matters now

The numbers tell the story of why this theme matters. The supplied data sheet shows Nvidia with a Market Value of 4,849.55 billion in local currency, which would put it among the largest companies in history by Market Capitalisation. Broadcom carries a market value of 1,976.39 billion US dollars and AMD 577.94 billion US dollars. Taiwan Semiconductor's market value is 55,365.94 billion New Taiwan dollars. The combined heft of these four names is enough on its own to move global indices.

For UK investors holding S&Amp;P 500 trackers or global Equity funds, this means a meaningful slice of their savings is now riding on the AI build-out. That makes it worth understanding what the figures imply, and where the obvious sensitivities lie.

By the numbers (FT Global 500)

Nvidia is shown at 199.57 in local currency with a 52-week high of 216.82 and a low of 104.08, a daily change of -9.68, a Yield of 0.02 per cent and a price-to-earnings ratio of 40.48. The market value reads 4,849.55 billion. Even after some recent Volatility, the share is much closer to its 52-week high than its low, illustrating just how powerful the rally has been. Note that the country label in the supplied sheet shows the United Arab Emirates, which the data caveat indicates should be ignored — Nvidia is a US-listed company.

Broadcom is at 417.43 US dollars with a 52-week high of 429.31 and a low of 184.02, a yield of 0.62 per cent and a P/E of 78.91, on a market value of 1,976.39 billion dollars. The 52-week range tells investors that Broadcom shares have more than doubled at the top end of their range during the last twelve months. AMD is shown at 354.49 dollars, with a 52-week high of 352.99, low of 91.87, a P/E of 134.79 and a market value of 577.94 billion. The eye-catching figure here is the price relative to the 52-week range, with the recent price effectively at the upper boundary, and the high triple-digit P/E.

Taiwan Semiconductor is at 2,135.00 New Taiwan dollars with a 52-week high of 2,330.00 and a low of 899.00, a yield of 1.03 per cent, a P/E of 47.18 and a local-currency market value of 55,365.94 billion. The wide 52-week range underlines how dynamic the share has been over the period. ASML, the European tooling supplier, is at 1,222.40 euros against a 52-week high of 1,312.80, on a P/E of 61.40 and a market value of 474.47 billion euros. Applied Materials, Micron Technology and Analog Devices, the broader semiconductor supporting cast, also feature in the supplied data and round out the picture.

Two further names worth noting from the same data set are Arm Holdings, at 210.32 US dollars on a P/E of 269.64, and Palantir, at 139.11 in local currency on a P/E of 201.61. Both illustrate how richly the market is willing to value businesses tied to the AI theme on certain metrics.

Growth drivers

The growth case for these companies rests on a single, powerful idea — that the world is at the start of a multi-year capital expenditure cycle to build out AI compute. The largest cloud companies have all signalled that they intend to keep investing aggressively in data centres, networking and accelerators. Sovereign customers, ranging from national AI clouds to defence-related projects, are adding a new layer of demand. Enterprises are gradually moving from experimentation to production deployments of AI tools, which over time should also support demand for inference chips.

Each of the four names benefits in a slightly different way. Nvidia continues to set the pace in training accelerators, with each generation of chip pushing performance further. Broadcom's role in supplying custom silicon to hyperscalers gives it a direct line into customers that want alternatives to off-the-shelf parts. AMD has a credible roadmap in data-centre central processors and graphics accelerators, and its lower starting Market Share gives it room to grow. Taiwan Semiconductor benefits no matter which design wins because almost every advanced chip lands in its fabs.

Beyond AI itself, there are secondary tailwinds. The recovery in PC and smartphone replacement cycles, the ongoing electrification of vehicles, and a steady upgrade cycle in industrial and networking equipment all support semiconductor demand. UK investors should not lose sight of the fact that the chip industry has always been cyclical, but the structural shift towards more compute per workload is a meaningful, long-term tailwind.

The relationship between the four core names is also worth understanding. Nvidia and AMD compete most directly in graphics-accelerated AI training and inference, while Broadcom's role often complements rather than replaces what Nvidia offers, particularly in custom silicon partnerships with hyperscalers. Taiwan Semiconductor sits beneath all of them as the manufacturing foundation that turns designs into chips. Investors who want exposure to the AI build-out without trying to pick a single winning architecture sometimes find that owning the foundry alongside one or more chip designers gives a more balanced expression of the theme.

Risks to watch

Valuation is the most obvious risk. AMD is shown on a price-to-earnings ratio of 134.79 and Broadcom on 78.91. Nvidia at 40.48 looks more reasonable, but its market value is so large that small disappointments translate into very big absolute losses. When valuations are this stretched, even modest changes in interest-rate expectations or growth assumptions can trigger sharp drawdowns.

Concentration risk runs alongside valuation risk. The hyperscaler customers driving today's orders are themselves a small group, and any signal that their capital expenditure is plateauing would weigh heavily on chip names. Geopolitical risk is unavoidable for Taiwan Semiconductor, given its location, and is a slow-burning issue for the entire industry. Export controls between the United States and China have already affected product roadmaps and addressable markets, and any further escalation could change demand patterns quickly.

Execution risk is also real. Producing leading-edge chips at scale is one of the most demanding industrial processes in the world, and any stumble in yields or capacity ramps can be felt across the supply chain. Currency volatility matters as well, especially for sterling-based investors trying to translate gains in dollars or New Taiwan dollars back into pounds.

Finally, there is the inevitable cyclical risk. Semiconductors have always been a cyclical industry. Even if the AI super-cycle continues, periods of digestion are likely. Investors who buy at the top of the range and have to sit through a correction need to be honest about whether they have the time horizon to do so.

Investor takeaway

Looking at the four core names side by side using the supplied figures highlights how differently the market is currently treating them. Nvidia trades on a still-elevated but more digestible 40.48 P/E given its dominance. Broadcom carries a higher 78.91 P/E but offers a small Dividend Yield of 0.62 per cent and exposure to non-AI franchises. AMD's 134.79 P/E reflects the market's expectation of strong earnings growth from a lower base. Taiwan Semiconductor on 47.18 stands out as a relatively measured multiple given its critical position in the value chain.

For UK investors, the key questions are not which name will deliver the best return next quarter but how much exposure to this theme is appropriate, how that exposure is being expressed inside trackers and active funds, and whether currency hedging is worth considering. Position sizing, regular Rebalancing and a long-term horizon are the tools that historically have helped investors stay on the right side of cyclical, theme-driven rallies.

AI is unlikely to disappear as a theme anytime soon, but the path of share prices is rarely a straight line. The figures from the supplied sheet capture a single moment in time, and the gap between today's prices and 52-week lows is a useful reminder of how much volatility this group has already produced. Investors who stay disciplined about valuation are likely to fare better than those who chase momentum at any price.

It is also worth thinking about how AI exposure is held. UK investors who own a global tracker, a US large-cap tracker, a tech-focused fund or even a mainstream pension default fund are likely to already have substantial exposure to Nvidia, Broadcom and Taiwan Semiconductor through those vehicles. Adding individual shares on top of that can quickly tip a portfolio into a heavy single-theme bet without the investor realising. A simple exercise — listing the top ten holdings of every fund held — often reveals a level of overlap that makes adjustments necessary.

Finally, no theme runs in a perfectly smooth line. Even within the AI cohort there will be winners, also-rans and disappointments. Margins, end-customer concentration, geopolitical exposure, manufacturing yields and the speed at which competing architectures emerge will determine which of today's leaders are still leaders in five years. Approaching this group with humility, looking at the figures objectively rather than emotionally, and accepting that some losses along the way are inevitable is more useful than trying to pick the single best name. The supplied data is one input in that process, but it is not the only one.