AI Stocks Institutional Research 2026: The Complete Investor's Guide

The AI Investment Landscape in 2026

The artificial intelligence investment cycle has entered a phase of bifurcation. The first wave — characterised by indiscriminate capital flows into any company that could credibly claim AI exposure — has given way to a more discriminating market. Institutional capital is now rotating from AI infrastructure plays into AI application beneficiaries, creating distinct return profiles across the value chain.

This transition mirrors historical technology cycles. The semiconductor buildout that enabled cloud computing (2005–2015) ultimately rewarded infrastructure providers less than the application layer companies that captured enterprise and consumer spending. Investors who extrapolate AI infrastructure demand indefinitely risk repeating the errors of the fibre optic overbuild of the late 1990s.

Three structural dynamics define the current environment:

• Compute commoditisation: As NVIDIA's H100/H200/B200 architecture becomes more widely available and AMD's MI300X gains adoption, the marginal cost of AI inference is declining sharply. This compresses margins for hyperscalers relying on compute resale while benefiting AI-native application companies.
• Agentic AI deployment: The shift from generative AI to agentic AI — systems that can plan, execute, and adapt across multi-step workflows — is creating new software categories. Companies building orchestration layers, tool-use frameworks, and agent management infrastructure represent the next S-curve.
• Geopolitical bifurcation: US export controls on advanced semiconductors have accelerated Chinese domestic development. Investors must now model two separate AI ecosystems: the US-allied stack (NVIDIA, AMD, TSMC, ASML) and the Chinese parallel stack (Huawei Ascend, SMIC, domestic foundries). Return profiles differ materially.

Semiconductor Investment Framework

Tier 1: Foundational Infrastructure

NVIDIA (NVDA) remains the defining company of the AI compute cycle. Its H100/H200 GPU cluster architecture has become the de facto standard for large language model training. The question for investors is no longer whether NVIDIA dominates — it clearly does — but whether its current valuation (trading at ~25x forward revenue at various points in 2025) adequately compensates for the risks of compute commoditisation and Chinese market displacement.

Our base case: NVIDIA's data centre revenue continues to grow at 30–40% annually through 2027, driven by inference demand scaling faster than training demand declines. However, peak multiple expansion has likely passed. The stock rewards patient accumulation on corrections rather than chasing momentum.

TSMC (TSM / 2330.TW) occupies an irreplaceable position in the AI supply chain. Its N3 and N2 process nodes are the only commercially viable options for leading-edge AI chip fabrication. The Taiwan geopolitical risk premium is real but frequently overpriced by markets unfamiliar with the economic deterrence dynamics of semiconductor dependence.

Tier 2: Picks-and-Shovels Plays

ASML (ASML) holds a near-monopoly on extreme ultraviolet (EUV) lithography equipment, the physical bottleneck for sub-7nm chip manufacturing. With a 2-3 year backlog and no credible competitor within a decade, ASML's pricing power is exceptional. Export controls on its most advanced EUV machines to China have modestly impacted volume but improved overall margin quality.

Broadcom (AVGO) has emerged as the dominant custom AI chip (XPU/ASIC) provider for hyperscalers seeking NVIDIA alternatives. Its custom silicon engagements with Google (TPU), Meta, and ByteDance represent a multi-billion dollar annuity stream that the market has been slow to fully price.

Tier 3: Application Layer Compounders

The application layer thesis rests on a simple observation: AI reduces the marginal cost of producing software features to near zero, while the value of software to enterprise customers remains unchanged or increases. Companies with strong distribution networks and existing customer relationships benefit disproportionately.

Microsoft (MSFT) has embedded Copilot across its entire product suite. Azure's AI revenue contribution is growing faster than any other cloud segment. The enterprise sales motion — selling AI capabilities into existing Office 365 and Azure relationships — creates durable switching costs.

Palantir (PLTR) is the institutional investor's contested AI application stock. Bulls cite its AIP platform as the first enterprise-grade AI operating system with demonstrable ROI metrics from deployed customers. Bears cite valuation (consistently 30–50x forward revenue) and customer concentration risk. The stock's behaviour as a sentiment indicator for AI application enthusiasm makes position sizing discipline essential.

Research Coverage: AI & Technology

The following articles represent Solomon Grey Capital's analytical coverage of AI investment themes. Each piece is written for institutional and professional investors who require analytical rigour alongside commercial insight.

Valuation Framework for AI Stocks

Traditional DCF models struggle with AI stocks because the terminal value assumptions — which dominate valuation — require forecasting market structures that do not yet exist. A more useful framework combines three lenses:

1. TAM Capture Analysis

Estimate the total addressable market for the company's primary product category in 5–7 years, then apply a realistic market share range. Sensitivity-test across bull/base/bear scenarios. The discipline here is not the TAM estimate itself — those are always speculative — but the market share assumptions. Companies rarely capture more than 20–30% of large TAMs unless they have monopolistic network effects.

2. Rule of 40 Trajectory

For software and SaaS AI companies, the Rule of 40 (revenue growth rate + free cash flow margin ≥ 40%) remains a useful filter. Companies trading at high multiples need Rule of 40 scores above 60 to justify their valuations. Track this quarterly and compare to peer cohorts.

3. Competitive Moat Assessment

The most important question for any AI investment: what prevents a well-capitalised competitor from replicating this product in 18–24 months? Durable moats in AI typically derive from: proprietary data at scale, switching costs embedded in enterprise workflows, network effects in multi-sided platforms, or physical infrastructure advantages (fabs, data centres, power contracts) that cannot be quickly replicated.

Key Risks to Monitor

Compute commoditisation risk: If AI training becomes significantly cheaper (through algorithmic efficiency, not just hardware), the revenue opportunity for compute providers shrinks. DeepSeek's demonstration of competitive LLM performance at lower compute costs was the first material signal of this risk.

Regulatory risk: The EU AI Act, US executive orders on AI safety, and Chinese AI governance regulations are all moving toward more prescriptive frameworks. Compliance costs will advantage large incumbents over smaller AI-native companies.

Geopolitical risk: Further export control escalation, TSMC forced relocation scenarios, or Taiwan Strait tensions represent tail risks that are difficult to hedge within a pure AI equity portfolio. Gold, energy, and defence allocations provide partial offsets.

Concentration risk: The Magnificent Seven's AI exposure means most institutional equity portfolios already have significant AI risk even without explicit AI stock allocations. Investors should audit their full portfolio AI beta before adding incremental AI-specific positions.

Investment Conclusions

The AI investment opportunity remains structurally intact. The transition from training-dominated to inference-dominated AI workloads, the buildout of agentic AI infrastructure, and the application layer monetisation wave all represent multi-year compounding opportunities for well-positioned investors.

The investment discipline required is selectivity. Not every company claiming AI exposure will benefit. The framework above — TAM capture, Rule of 40 trajectory, competitive moat assessment — provides a systematic basis for distinguishing genuine beneficiaries from AI narrative beneficiaries.

Solomon Grey Capital will continue publishing institutional-quality research on specific AI investment opportunities as the cycle evolves. The research linked above represents our current published coverage. Subscribers receive new analysis as it is published.