Nvidia and the AI Sector: How to Read the Investment Thesis Honestly

Few stocks in modern market history have produced numbers like Nvidia in the period between 2023 and 2026. Data centre revenue scaled from tens of billions to over $190 billion in fiscal 2026, market share in AI accelerators settled around 80% of the addressable market, and the company joined a small group of firms whose market capitalization, was presented by commentators  to exceed the GDP of major economies. Behind the headlines is a more complicated story: the thesis is real, the risks are real too, and the narrative around ‘AI as a generational opportunity’ often obscures what investors should actually examine. This article looks at the numbers honestly, explains the bull and bear cases without drama, and offers a framework for thinking about exposure to the sector responsibly.

IMPORTANT DISCLAIMER. This article is provided for general information and educational purposes only. It does not constitute investment advice, a personal recommendation, an offer or a solicitation to buy or sell any financial instrument, or an offer to provide or  enter into any investment service. Nothing in this article should be relied upon as a forecast, projection, or guarantee of future results. Investing in financial instruments involves risk, including the risk of losing part or all of the capital invested. Past performance is not a reliable indicator of future results, and any performance figures or examples presented in this article are provided for illustrative or historical purposes only and do not guarantee future results. Skanestas Investments Limited is regulated by the Cyprus Securities and Exchange Commission (CySEC) under licence number CIF251/14. Independent professional advice should be sought as needed. Any reference to specific products or services is for general informational purposes only; the provision of investment services and the assessment of appropriateness or suitability, where applicable are subject to the applicable regulatory requirements and the firm’s relevant procedures.

 

It is not a recommendation to buy or sell any specific instrument. The framing is informational; this article does not constitute investment, legal, or tax advice, and individual circumstances should be assessed with qualified counsel. This article is a marketing communication, presenting general information and it does not represent the Company’s investment view, price target, or recommendation on Nvidia Corporation or any related instrument.

What the Numbers Actually Show

Nvidia’s fiscal year 2026, which ended in late January 2026, produced reported full-year revenue of approximately $193.7 billion — up 68% year over year (source: Nvidia Corporation FY2026 earnings release, February 2026) — with the data centre segment accounting for the dominant majority of company revenue. Quarterly data centre revenue grew through the year, with Q4 FY2026 alone reported by the company at approximately  $62 billion (source: Nvidia Q4 FY2026 earnings release). The platform shifts driving these figures are accelerated computing and generative AI infrastructure spending by hyperscale cloud providers, sovereign AI initiatives, and large enterprises building dedicated AI compute capacity.

Third-party industry estimates suggest Nvidia held around 80% to 86% of the AI accelerator market depending on how the segment is defined, with training workloads concentrated more heavily in Nvidia silicon than inference workloads where custom chips and CPU competition have a larger presence. The competitive landscape is changing — AMD has expanded its data centre business meaningfully, hyperscalers including Amazon, Google, and Microsoft have invested in custom silicon, and specialty inference players such as Groq have secured licensing arrangements — but Nvidia’s lead in 2026 is regarded as structural rather than cosmetic.

These figures are public, audited, and well-documented. They are not the issue. The issue is what they imply about the future.

The Bull Case in Plain Language

The bullish thesis on the AI sector — and on Nvidia as its most direct beneficiary — typically rests on several pillars.

Demand is structural rather than cyclical. The argument is that AI infrastructure build-out reflects a multi-decade transformation comparable to electrification or the early internet, and the hyperscaler capital expenditure cycle (collectively approaching $600 billion in 2026 by some estimates) has years to run before saturation. The compute intensity of frontier models continues to grow, and inference workloads are scaling faster than training workloads as deployed AI applications proliferate.

Nvidia’s competitive moat is broader than chip performance alone. The argument here is that the company has built a platform — silicon, networking, system architecture, and the CUDA software ecosystem — that compounds advantages over time. Switching costs for large customers are not just hardware-level; they include developer tooling, model libraries, and validated reference implementations that have been refined over fifteen years.

Cash generation supports continued reinvestment. The data centre segment’s gross margins have been described as enabling annual platform refreshes — Hopper, Blackwell, Blackwell Ultra, and the announced Rubin platform — that maintain a performance lead competitors must constantly chase. The combination of revenue growth and high margins produces free cash flow that can fund both return to shareholders and continued R&D.

These arguments are coherent. They are not slogans.

The Bear Case in Plain Language

The bearish concerns about the AI sector and Nvidia’s specific position deserve equally serious treatment, even from investors inclined to the bull view.

Customer concentration is meaningful. Nvidia’s own filings note that revenue is significantly concentrated among a small number of customers — the major hyperscale cloud providers and a handful of large AI infrastructure builders. A change in capital expenditure plans by even one or two of these customers can produce material revenue impact. The hyperscalers are also building their own custom silicon and have direct economic incentive to reduce dependence on Nvidia over time.

AI infrastructure spending is currently driven by capex cycles that may not be sustainable indefinitely at current rates. Approximately $600 billion of hyperscaler capex in 2026 is a number that, even given the scale of the cloud operators, requires confident expectations of return on invested capital. If inference workloads do not generate the revenue the capex anticipates, the next-cycle order book could moderate.

Valuation prices in continued growth at very high rates. Multi-trillion-dollar market capitalisations imply expectations that compound from already-large bases. Even modest disappointment relative to expectations — not absolute decline, just slower-than-expected growth — has historically produced significant share-price reactions in growth stocks at this stage. This is an information feature of valuation mathematics, not a directional view.

Regulatory, geopolitical, and supply-chain factors are real. US export controls on advanced chips to China have constrained part of Nvidia’s addressable market. Manufacturing dependencies on TSMC and on advanced packaging capacity introduce single-vendor concentration that the company is working to mitigate but cannot eliminate quickly. Political dynamics around AI regulation in the EU and elsewhere are evolving and could affect specific use cases or product configurations.

What Neither Side Tells You

Most investor commentary on Nvidia frames the question as bullish or bearish on the company itself. A more honest framing recognises that the question contains several distinct sub-questions that can have different answers.

Is AI as a technology going to produce sustained productivity gains over the coming decade? This is the broadest question, and most thoughtful observers answer something like ‘probably, in some form, with significant uncertainty about magnitude and timing.’

If AI does produce those gains, will the value capture occur primarily at the chip layer, the cloud infrastructure layer, the application layer, or the end-user/enterprise layer? History suggests value capture in technology cycles often migrates over time. Early internet was dominated by infrastructure builders; later value migrated up the stack to platforms and applications. Whether AI follows a similar pattern is uncertain.

Even if value continues to be captured at the chip layer, will Nvidia continue to capture the dominant share, or will competitive pressure reduce its market share over time? The 80%+ share of 2026 is among the highest concentrations seen in major hardware markets in recent decades, according to industry commentary, and historically such shares have been difficult to sustain indefinitely as competition catches up and customers diversify suppliers.

An honest investor weighs all three of these questions separately rather than collapsing them into a single ‘is AI real’ framing.

Sector Versus Single-Stock Exposure

Investors who want exposure to the AI investment theme have several structural choices, each with different risk and return characteristics.

Single-Stock Exposure

Owning Nvidia directly concentrates the bet on a single company. The upside is direct participation if Nvidia continues to outperform the broader sector. The downside is concentration risk — company-specific events (regulation, customer concentration, execution missteps, manufacturing issues) can produce sharp moves regardless of the broader sector trajectory.

Broad Index Exposure

Investors holding S&P 500 or NASDAQ index funds have meaningful indirect exposure to Nvidia and other AI-related names through normal index construction. This is a less concentrated form of participation, but as discussed in the article on Magnificent Seven concentration, the indirect exposure can itself be heavier than investors realise.

Thematic ETFs

AI-themed ETFs hold baskets of companies considered exposed to the AI investment thesis — chips, cloud infrastructure, software, and adjacent areas. They diversify single-stock risk within the theme but introduce thematic concentration: if the AI investment narrative cools, the entire basket can move together.

Diversified Equity Exposure With Awareness

Many sophisticated investors maintain diversified portfolios in which AI exposure is one component among several, sized intentionally relative to the rest of the allocation. This approach captures upside from the theme without making it the dominant determinant of overall portfolio outcomes.

Avoidance

Some investors explicitly choose not to participate in AI sector themes, either because they believe the prices already reflect more than the realistic outcomes warrant, or because they prefer not to concentrate in any single secular theme. This is a defensible position, particularly for investors with shorter time horizons or higher loss aversion.

Behavioural Risks Specific to Hot Themes

Hot investment themes amplify all the behavioural patterns that affect investors. The fear of missing out drives entries at high valuations. Recency bias makes recent strong performance feel like a permanent state. Confirmation bias makes contradicting evidence easier to dismiss. Herd behaviour amplifies position-sizing beyond what the investor would otherwise choose.

These patterns produce predictable outcomes: investors enter the theme late, size positions larger than they should, hold through corrections that become structural drawdowns, and capitulate near the lows after the most disciplined opportunities to add or rebalance have passed. The discipline of pre-committed allocation — deciding what percentage of the portfolio the theme deserves before reacting to recent performance — is one of the most useful protections against this.

How Skanestas Approaches Sector Themes

Skanestas’s portfolio management strategies are constructed within mandate boundaries that include sector and thematic considerations. Certain of Skanestas’ portfolio management strategies,  available to professional clients, can include positions in technology and AI-related names, alongside diversification across other sectors and instruments. Skanestas does not recommend any specific allocation to AI-related names for any individual reader; suitability of any strategy, including exposure to technology or AI-related instruments, is only determined through the firm’s regulatory suitability assessment.  The specific allocation reflects the firm’s investment process and the prevailing market environment, not a permanent thematic conviction. Concentration risk in any single name or theme is one of the variables explicitly considered in portfolio construction, consistent with the firm’s documented risk management framework. The availability and application of such strategies are subject to the relevant mandate, client classification, applicable regulatory requirements and the firm’s authorisation.

FAQ

Is Nvidia overvalued?

There is no single objective answer to this question; it depends on the discount rate applied, the growth assumptions used, and the time horizon. Reasonable analysts in 2026 disagree, with publicly reported  fair-value estimates ranging from significantly above to significantly below current prices. . The disagreement itself is informative — it suggests the price is sensitive to assumptions on which there is no consensus.

Will custom silicon erode Nvidia’s market share?

Probably some, over time. The pace and ultimate equilibrium are uncertain. Hyperscalers’ custom chips are typically designed for specific workloads where they have economic advantage; Nvidia retains lead in flexible high-performance compute. The eventual mix depends on workload evolution and on continued pace of Nvidia’s platform improvements.

How should I think about AI exposure in a balanced portfolio?

One  disciplined approach is to size AI exposure as a deliberate allocation rather than a residual of recent performance – for example, setting a target  percentage, monitoring drift from price moves, and periodically rebalancing toward that target. This applies equally to direct stock holding and to thematic ETFs, though whether and how to apply any such approach depends on individual circumstances.

Are AI-related stocks a bubble?

The label ‘bubble’ is more useful as commentary than as analysis. Some AI-related companies have valuations that price in continued strong execution; others may be reflecting durable economic value. The honest framing is that elevated multiples create elevated risk to disappointment, regardless of whether ‘bubble’ is the right word.

Conclusion

Nvidia’s numbers in 2026 are real, the AI investment cycle is real, and the bullish thesis is coherent. The risks are also real and worth examining seriously. The most useful posture for an investor is not to pick a side of the debate but to hold both views in tension, understand what would have to be true for each to play out, and size exposure deliberately rather than emotionally. Hot themes reward investors who maintain discipline more than those who are most enthusiastic. The investor who decides their allocation in calm conditions and reviews it periodically, is more likely to outperform the one who chases recent strength and capitulates on corrections, though no outcome is guaranteed..

 

About this article: This material is published for general informational purposes by Skanestas Investments Limited, a Cyprus Investment Firm authorised and regulated by the Cyprus Securities and Exchange Commission under licence CIF251/14. The content reflects general industry practice and information available as at the date of publication may be updated from time to time without notice. The article does not establish a client relationship and does not replace the formal contractual, regulatory and client documents  that govern any portfolio management or brokerage relationship with the firm. Risk warning: Investing in financial instruments carries risk and the value of investments may rise or fall. You may receive back less than the amount invested. Please review the firm’s Risk Disclosure Statement and other regulatory documents before engaging with any service. The information provided is general in nature and should not be understood as a statement of the services provided by the firm. Last updated: 2026.