The AI Capex Cycle: $725B Hyperscaler Buildout and the Five High-Conviction Positions

McKinsey's research shows global demand for data center capacity could almost triple by 2030, with approximately 70% of that demand driven by AI workloads. Total projected capital expenditure: $6.7 trillion, of which $5.2 trillion is attributable to AI processing loads and $1.5 trillion to traditional IT applications.

The AI capex cycle is defined by the hyperscalers — the companies that build and operate the cloud infrastructure on which AI runs. Amazon, Alphabet, Meta, Microsoft, and Oracle collectively spent $256 billion on capital expenditure in 2024. The Big-5 estimate for 2025 is $443 billion, a 73% YoY increase. Q1 2026 earnings (reported April 29, 2026) confirmed the combined 2026 figure at approximately $725 billion — a ~64% increase over 2025 — with approximately 75% of that spend (~$545 billion) directed at AI-specific infrastructure: GPUs, servers, data center construction, power systems, and cooling. Alphabet raised guidance to $180–190B (partially due to the Intersect acquisition); Meta raised to $125–145B, explicitly citing memory price inflation; Microsoft guided to ~$190B CY2026 ($25B attributed to higher component pricing); Amazon reaffirmed $200B. See Exhibit 1 for the full company-level breakdown. Amazon's 2026 capex of $200 billion alone exceeds the combined annual capex of the entire publicly traded US energy sector. As a share of GDP, AI-related capital formation now sits at approximately 5% — a level last seen during the late-1990s technology boom, but with a structurally different demand foundation (see Section 02).

McKinsey constructed three scenarios ranging from constrained to accelerated demand, shaped by semiconductor supply constraints, enterprise AI adoption rates, efficiency improvements, and regulatory challenges. The base case — $5.2 trillion in AI data center capex — assumes continued growth without runaway acceleration or structural constraints.

The analogy to the late-1990s telecommunications infrastructure bubble is compelling in one dimension — the scale of capital deployment — and misleading in every other. Fiber in the 1990s was built speculatively, with virtually unlimited capacity once laid and zero refresh requirement. Data centers are physically constrained, contractually committed before construction, and subject to accelerated depreciation cycles that naturally absorb any temporary excess. The evidence is visible in vacancy data: North American colocation vacancy has fallen from 9.8% in 2020 to 2.3% in H1 2025 (JLL Research), while the fiber glut post-2001 saw vacancy exceed 20%.

The key structural difference McKinsey identifies is the cost of carrying excess capacity. Fiber, once laid, is nearly free to maintain. Data centers are the opposite: power, cooling, and maintenance are ongoing high costs regardless of utilization. But crucially, AI accelerators have 3–4 year refresh cycles — meaning any overcapacity is rapidly converted into obsolescence, and new workloads pull spare capacity well before it becomes stranded.

The AI capex cycle has introduced a structural change to hyperscaler financial models that was not present in prior technology investment waves: the shift from cash-funded to debt-funded infrastructure. Amazon, Alphabet, Meta, and Microsoft spent a combined $256 billion on capex in 2024. Q1 2026 earnings confirmed the combined 2026 figure at approximately $725 billion. Internal free cash flow cannot scale at that rate. As a result, hyperscalers have collectively turned to debt markets at a scale not seen since the 2001 telecom build-out — but with materially stronger balance sheets underpinning the leverage.

Morgan Stanley and J.P. Morgan project the technology sector will need to issue approximately $1.5 trillion in new debt over the next three years to fund the AI infrastructure build-out. The hyperscalers now spend 45–57% of revenue on capex — ratios previously seen only in capital-intensive industrial utilities and telecommunications companies. Amazon's 2026 capex guidance of $200 billion alone exceeds the combined annual capex of the entire publicly traded US energy sector.

The investment implication is dual-edged. Debt-funded hyperscaler capex raises the structural demand floor for AI infrastructure suppliers — contracts are signed, purchase orders placed, and delivery timelines locked in regardless of short-term sentiment shifts. However, rising leverage also introduces a new risk layer for hyperscaler equity positions themselves: if AI revenue monetisation disappoints, the debt service burden will suppress free cash flow precisely when investor patience is shortest. The debt burden is precisely why Microsoft (MSFT) and Alphabet (GOOGL) carry a Selective rather than High Conviction rating in A.L. Capital Advisory's framework — the infrastructure beneficiaries (NVDA, VRT, EQIX, CEG) capture the upside without carrying the balance sheet risk of the buyers.

In January 2026, OpenAI, SoftBank, and Oracle announced Project Stargate — a $500 billion AI infrastructure programme targeting US deployment, with an initial $100 billion committed within the first four years. Stargate represents a category of demand that does not appear in McKinsey's April 2025 base-case model: sovereign and government-adjacent AI infrastructure, funded at national-strategy scale rather than commercial return logic alone.

Stargate is not an isolated event. Saudi Arabia's Public Investment Fund has committed to a $40 billion AI infrastructure programme. The UAE has established G42 as a sovereign AI entity with data centre commitments across three continents. The European Union's AI Continent Action Plan targets €200 billion in AI investment through 2030. Each of these programmes represents demand outside the hyperscaler-driven model — demand that is contractually committed, politically supported, and insensitive to short-term ROI calculations.

The structural implication for the AI capex cycle is clear: the demand floor is higher than McKinsey's base case assumed, because McKinsey's model was built on commercial hyperscaler logic alone. Sovereign AI programmes add a second, non-correlated demand layer. Goldman Sachs's documented 11 GW US capacity shortfall — growing to 40 GW by 2028 — understates the total demand gap when sovereign programmes are included.

Among the five high-conviction positions, Constellation Energy (CEG) is the most direct Stargate beneficiary: nuclear baseload is the only power source that meets both the carbon-free and uninterruptible specifications that sovereign AI programmes require. Equinix (EQIX) benefits through its hyperscale campus footprint in the geographies where sovereign programmes are concentrating — Virginia, London, Singapore, and the Gulf. NVIDIA (NVDA) benefits from GPU procurement at sovereign scale. Vertiv (VRT) benefits through the thermal management requirements of the dense compute clusters Stargate's architecture demands.

McKinsey's analysis maps the $5.2 trillion AI capex envelope across five distinct investor archetypes. Understanding this architecture is essential: the investment case, risk profile, and return dynamics differ fundamentally across archetypes.

The $5.2–$6.7 trillion capex envelope flows through a defined set of public equities. But raw exposure to the AI theme is not sufficient — the archetype, moat, and balance sheet quality of each company determine whether they capture compounding returns or get crushed in the shake-out.

Every dollar of hyperscaler AI infrastructure capex ultimately flows through a semiconductor. The $660–690 billion committed for 2026 does not build itself — it must be converted into physical chips, packaged onto boards, slotted into racks, and cooled. Understanding the semiconductor supply chain is therefore not a secondary analytical exercise. It is the root constraint that determines whether the AI capex cycle delivers its projected $6.7 trillion in capacity by 2030 or falls short of the McKinsey base case. Three chokepoints define the current supply environment: GPU allocation, advanced packaging capacity at TSMC, and the EUV lithography monopoly held by ASML.

NVIDIA GPU Shortage — Why 90% Market Share Persists Despite Competition

NVIDIA Corporation (NASDAQ: NVDA) enters 2026 in a position with few historical precedents: a near-monopolist in a market growing at 36% annually, constrained not by demand but by its own supply chain. The Blackwell B200 architecture — delivering 2.5× the inference throughput of the H100 at the same power envelope — carries reported order lead times of 12–18 months for hyperscaler allocations as of April 2026. The binding constraint is not TSMC's fab capacity for the GPU die itself, but CoWoS-L (Chip-on-Wafer-on-Substrate with Local) advanced packaging, which stacks the B200 GPU die together with six stacks of HBM3E memory in a single thermally integrated module. TSMC's CoWoS capacity is expanding from approximately 9,000 wafers per month in 2024 to an estimated 30,000 by end of 2026 — but hyperscaler demand is tracking above this build rate.

At Google Cloud Next on April 22, 2026, Google confirmed NVIDIA Vera Rubin GPU instances will be available on Google Cloud in 2027 — the first public confirmation of commercial availability for NVIDIA's next-generation architecture beyond Blackwell. This extends the CUDA and NVIDIA platform advantage through at least the 2027–2028 GPU cycle, reinforcing the structural thesis for NVDA equity at every hyperscaler refresh cycle.

The CUDA software moat is as material as the hardware lead. NVIDIA's Compute Unified Device Architecture — the programming model that underlies virtually every production AI training workload — has been in continuous development since 2007. The model libraries (cuDNN, cuBLAS, TensorRT), the developer tooling, and two decades of academic and commercial code written natively for CUDA constitute a switching cost that Advanced Micro Devices (NASDAQ: AMD) is actively but slowly dismantling with ROCm. A.L. Capital Advisory estimates enterprise migration from CUDA to ROCm at current pace would require 3–5 years for non-latency-sensitive inference workloads — and meaningfully longer for training.

AMD MI300X vs NVIDIA B200 — Deep Technical and Commercial Comparison

AMD's MI300X is the most credible GPU challenger in the AI data center market. The MI300X integrates CPU and GPU chiplets in a unified HBM3 memory pool — 192GB of shared memory versus the H100's 80GB. For very large model inference (70B+ parameter LLMs), the MI300X's memory capacity advantage is architecturally significant: models that require tensor parallelism across 8 H100s can run on 4 MI300X units, reducing interconnect overhead. Microsoft Azure and Meta have both announced MI300X deployments for inference workloads, validating the commercial thesis.

However, the competitive gap remains wide on three dimensions: software maturity (ROCm operator coverage vs CUDA is estimated at 85–90% for inference, but substantially lower for cutting-edge training kernels), supply chain reliability (TSMC allocates CoWoS capacity to NVIDIA first as the larger revenue customer), and ecosystem lock-in (the dominant MLOps toolchain — PyTorch, JAX, TensorFlow — all optimise natively for CUDA). A.L. Capital Advisory's base case: AMD captures 8–12% of the AI accelerator market by 2027, up from approximately 5–6% in 2025. At that share level and current data center GPU ASPs ($25,000–$35,000 per unit wholesale), AMD Data Center revenue could reach $15–20 billion annually by FY2027 — a material but sub-consensus outcome.

Intel, ARM Architecture & the CPU Transition

Intel Corporation (NASDAQ: INTC) is executing a structural pivot from integrated device manufacturer to pure-play foundry (Intel Foundry Services / IFS) while simultaneously defending its CPU franchise against AMD and the ARM architecture wave. In the AI data center context, CPUs play a secondary but non-negligible role: every GPU cluster requires high-performance host CPUs for data ingestion, preprocessing, orchestration, and inference serving. Intel's Xeon Scalable 6th Generation (Granite Rapids) and AMD's EPYC Genoa both compete for this socket. AMD's EPYC has outgrown Intel in data center CPU share for three consecutive years, now estimated at 33–35% of new server deployments versus Intel's 65%.

Arm Holdings plc (NASDAQ: ARM) is the deeper structural story. Arm's v9 architecture — licensed to Apple, Qualcomm (QCOM), Amazon (Graviton), Google (Axion), and NVDA (Grace CPU) — delivers 30–40% better performance-per-watt than x86 at comparable workloads. In the AI inference layer, where power efficiency directly determines cost per token, the x86 architecture's dominance is structurally eroding. Arm's revenue is a royalty on every chip shipped using its architecture — a position of compounding leverage as ARM-based designs proliferate across data centers, edge infrastructure, and AI accelerators.

ASML — The Single-Point-of-Failure in Global AI Chip Supply

ASML Holding N.V. (NASDAQ: ASML) manufactures every extreme ultraviolet (EUV) lithography machine on Earth. There is no second supplier. EUV lithography is required to pattern the sub-7nm transistors that power every leading-edge AI chip — NVIDIA's B200 on TSMC N3E, AMD's MI300X on TSMC N5, Intel's Gaudi 3 on Intel 7. A single EUV tool costs approximately €200 million, weighs 180 tonnes, and requires 40 shipping containers and a dedicated Boeing 747 to transport. ASML ships approximately 50–60 EUV systems per year. Lead times are 18–24 months from order to installation.

The investment case for ASML is the investment case for the AI capex cycle expressed through the supply chain's deepest chokepoint. Every new semiconductor fab built to address AI demand — TSMC Arizona, Samsung Taylor, Intel Ohio — requires ASML EUV machines. The Veldhoven-based company's order book extends through 2027 and includes next-generation High-NA EUV tools (approximately €380 million per unit) required for sub-2nm nodes. No competitor has a functioning EUV tool. The development timeline for a competitor to reach commercial EUV from scratch is estimated by industry analysts at 15–20 years and multiple billions in investment. ASML's geopolitical risk — the Dutch government, under US pressure, has restricted EUV exports to China — removes the largest potential demand overhang and creates a China-exclusion premium that benefits Western fabs.

Custom Silicon — Broadcom & Marvell as the ASIC Layer

The custom silicon trend is one of the most consequential supply-chain developments of the 2026 AI cycle. Google (TPU v5), Amazon (Trainium2, Inferentia3), Meta (MTIA2), and Microsoft (Maia 2) are all deploying custom AI accelerators designed in-house and manufactured at TSMC — explicitly to reduce dependence on NVIDIA and capture gross margin. The hyperscalers cannot design these chips themselves from scratch. They use third-party chip architects: Broadcom Inc. (NASDAQ: AVGO) and Marvell Technology Inc. (NASDAQ: MRVL) are the two dominant custom ASIC designers for AI workloads.

Broadcom has disclosed that its three largest custom AI chip customers will consume a combined addressable market of $60–90 billion per year by FY2027, based on disclosed deployment plans. Marvell's custom AI silicon revenue is smaller but growing rapidly, with design wins at Amazon and Microsoft. Both companies benefit from a structural dynamic that NVIDIA cannot easily disrupt: hyperscalers have a strategic incentive to fund custom silicon as a CUDA countermeasure, regardless of short-term cost premium. The ASIC layer is therefore not a threat to NVIDIA in the near term (custom ASICs are inference-only and cannot match NVIDIA's training flexibility) — but it is a meaningful long-term share shift that AVGO and MRVL are directly positioned to capture.

Server Integration Layer — Super Micro Computer & Dell Technologies

Super Micro Computer Inc. (NASDAQ: SMCI) and Dell Technologies Inc. (NYSE: DELL) occupy the final assembly layer of the AI GPU supply chain — converting raw NVIDIA, AMD, and ASML-derived chips into deployable rack-scale systems. Super Micro's liquid-cooled DGX-H100 and MGX server platforms are NVIDIA's preferred reference design partners; Super Micro has been first to market with Blackwell-based server systems for three consecutive GPU generations. Dell's AI server portfolio (PowerEdge XE9680) competes for the same enterprise and colocation buyer, with the additional distribution advantage of Dell's global direct sales force and ProSupport services.

Both companies have supply chains directly gated by NVIDIA's GPU allocation — when B200 supply is constrained, SMCI and DELL backlog builds and gross margins compress on pre-sold orders. The investment risk is margin: server integration is a low-margin business (5–8% EBIT for SMCI, 4–6% for AI infrastructure servers at Dell), and pricing competition between the two intensifies during GPU shortage periods. The structural opportunity: as the AI capex cycle moves from Phase 1 (Build) to Phase 2 (Deploy), hyperscalers shift from direct NVIDIA procurement to third-party server integrators — expanding SMCI and DELL's total addressable market.

The binding constraint on NVIDIA's Blackwell B200 GPU output in Q1–Q2 2026 is not the GPU die. It is High Bandwidth Memory 3E (HBM3E) — the stacked DRAM that sits beside every AI accelerator and provides the memory bandwidth that separates a usable AI chip from a theoretical one. A single B200 GPU die requires six stacks of HBM3E, each consuming approximately 8GB, for a total of 192GB per chip. At current NVIDIA GPU shipment volumes, the global HBM market must produce and package more advanced memory in 2026 than it produced across all previous years combined. There are exactly three HBM suppliers on Earth: SK Hynix (unlisted in this directory), Samsung (unlisted), and Micron Technology Inc. (NASDAQ: MU). Micron is the smallest of the three — and the most investable for Western investors.

HBM3E — The Architecture of Scarcity

High Bandwidth Memory is not DRAM as conventionally understood. Standard DDR5 DRAM transmits data serially over a narrow bus. HBM stacks 8–12 DRAM dies vertically using through-silicon vias (TSVs), achieving memory bandwidth of 1.2 TB/s per stack versus DDR5's 0.1 TB/s per channel. For AI training workloads — which require feeding terabytes of model weights and activations per second to the GPU's thousands of CUDA cores — HBM bandwidth is the performance ceiling that determines training throughput. The difference between running a 70B parameter model training run on HBM3E and on GDDR6 is roughly 4–5× in training speed, which at hyperscaler compute costs translates directly into tens of millions of dollars per training run.

HBM production requires TSMC (TSM) CoWoS packaging to integrate the HBM stacks with the GPU die — creating a circular dependency in the supply chain: both the GPU and its memory require the same scarce CoWoS capacity. SK Hynix currently holds an estimated 50–55% of the HBM market; Samsung approximately 35–40%; and Micron approximately 8–12%, ramping aggressively. NVIDIA has publicly qualified all three suppliers for HBM3E, but SK Hynix retains a technology generation lead: Hynix began HBM3E production in Q3 2024; Micron qualified HBM3E in Q4 2024; Samsung's HBM3E yield issues delayed their qualification into Q1 2026. This means Micron enters 2026 in an unusual position — a qualified second supplier in a market where the lead supplier cannot meet demand and the third supplier has quality problems.

Micron Technology — The High-Conviction AI Memory Investment Case

Micron Technology Inc. (NASDAQ: MU) is A.L. Capital Advisory's fifth High Conviction position — and the one with the widest gap between current consensus expectations and structural opportunity. Micron's investment case rests on three independent pillars that compound simultaneously through 2028.

First, the HBM revenue inflection. Micron's HBM revenue was negligible in FY2024 (ending August 2024). The company has guided to HBM becoming a multi-billion dollar revenue line in FY2025, with HBM3E production ramping throughout 2025 and into 2026. At SK Hynix's disclosed HBM gross margins (50–55%), HBM is transformative for Micron's blended gross margin profile, which has historically averaged 25–35% across the DRAM/NAND cycle. A single HBM3E 8-Hi stack carries an ASP of approximately $20–25 per GB — versus $3–4 per GB for commodity DDR5 DRAM. The six HBM3E stacks in a single B200 GPU generate more revenue per chip for Micron than an entire DDR5 memory kit for a standard server.

Second, the DRAM pricing cycle. The 2022–2023 DRAM oversupply — driven by consumer PC and smartphone demand collapse — has fully resolved. Industry-wide DRAM bit output growth has been deliberately constrained by all three major producers (Micron, SK Hynix, Samsung), with capital expenditure redirected to HBM production, which is DRAM capacity that becomes unavailable for standard DRAM supply. The structural result: DDR5 server DRAM pricing is rising through 2026 as data center demand (each AI server rack contains $100,000–$400,000 of standard DRAM alongside the GPU) grows faster than supply can respond. Micron is a direct beneficiary of both the volume increase and the ASP uplift.

Third, the NAND recovery. Western investors often model Micron as a pure DRAM company, but NAND flash (used for SSDs, storage arrays, and training data pipelines) represents approximately 35% of Micron's revenue. The NAND cycle bottomed in Q3 2023 at price levels that forced all producers, including Western Digital (NASDAQ: WDC), into EBITDA-negative operations. The recovery is now well established: enterprise SSD pricing has recovered 60–80% from the trough as AI training datasets, model checkpoints, and inference caches drive unprecedented enterprise NVMe demand. Western Digital is the pure-play NAND recovery thesis — its Flash segment directly captures the enterprise SSD pricing cycle without the HBM exposure that makes Micron the more complete AI memory story.

The Memory Wall — AI's Hidden Bottleneck

The "memory wall" is the gap between the growth rate of AI model complexity (parameters, context window, batch size) and the growth rate of memory bandwidth. A GPT-4-scale model requires feeding approximately 140GB of weights to GPU cores during each forward pass. At current HBM3E bandwidth of 1.2 TB/s per stack and 6 stacks per B200, the peak sustainable throughput is approximately 7.2 TB/s per GPU. Projected model sizes for 2026–2027 training runs (estimated 1–10 trillion parameter models) would require 8–12 HBM3E stacks per GPU — beyond the current B200 physical architecture. The implication: every new GPU generation (NVIDIA's Rubin/R100, AMD's MI400) will require more HBM stacks, more advanced packaging, and higher memory bandwidth — a structural demand escalator that benefits Micron, SK Hynix, and the HBM memory ecosystem indefinitely.

The per-rack memory content of an AI server cluster illustrates the scale: a single NVIDIA DGX H100 system (8× H100 GPUs) contains 640GB of HBM2e, plus 2TB of DDR5 system DRAM, plus 30TB of NVMe SSD storage. At blended memory content pricing, the memory stack in a single DGX H100 represents approximately $80,000–$120,000 of the system's total cost — comparable to one H100 GPU. At the 40 GW US capacity shortfall Goldman Sachs documents, each gigawatt of AI data center capacity contains approximately 10,000 racks. The memory content of that capacity: 10,000 × $120,000 = $1.2 billion per gigawatt in memory spend alone. The 40 GW US shortfall implies $48 billion in cumulative incremental memory demand — before Europe, Asia, and Project Stargate are counted.

Geopolitical Risk — China Memory and Export Controls

Micron's most significant risk is regulatory: China's Cyberspace Administration of China (CAC) banned Micron products from "critical information infrastructure" operators in May 2023 — a retaliatory measure following US restrictions on Chinese advanced chip imports. China represented approximately 16% of Micron's FY2023 revenue. The ban's impact has been partially absorbed by Micron redirecting China supply to other markets (particularly India, Southeast Asia, and European data center expansion), but approximately $800M–$1.2B of annualised revenue remains displaced. A full resolution — or further escalation — of the US-China technology trade war is the key binary risk in the Micron investment case.

Chinese domestic memory is a structural headwind, but further than commonly believed from competitive parity. CXMT (ChangXin Memory Technologies), China's domestic DRAM producer, is shipping DDR4 and early DDR5 at estimated 25–35% yield versus industry standard 85–90%. Yangtze Memory Technologies (YMTC), China's NAND producer, reached 232-layer NAND in 2023 but faces ASML EUV import restrictions that will prevent progression to sub-10nm node technology required for next-generation HBM. The Chinese memory industry is not a 2026 threat to Micron's HBM business. Chinese domestic memory represents a 2029–2032 risk for commodity NAND market share.

Portfolio Construction Framework — Five principles for building the AI infrastructure position without getting burned:

The AI capex cycle is not a theme. The AI capex cycle is a decade-long structural reallocation of capital — from consumption to physical infrastructure — at a scale not seen since the electrification of the United States economy in the 1920s. Q1 2026 earnings confirmed the Big-5 hyperscalers will spend approximately $725 billion on AI infrastructure in 2026 alone, nearly tripling the $256 billion deployed in 2024. Goldman Sachs documents a US capacity shortfall already exceeding 11 gigawatts, growing to 40 GW by 2028. North American colocation vacancy has fallen to 2.3% — a level at which pricing power is structural and durable.

This paper has mapped the full capital flow across the cycle's three phases. In Phase 1 (Build, 2024–2026), semiconductor procurement and thermal management are the primary value-capture layer — hence NVIDIA and Vertiv as peak-conviction positions. In Phase 2 (Deploy, 2026–2028), contracted infrastructure operators with physical scarcity moats begin to compound: Equinix's interconnect estate and Constellation Energy's 20-year nuclear power purchase agreements. In Phase 3 (Compound, 2028–2030), all five positions benefit simultaneously as AI-driven cloud revenue scales and contracted revenue compounds across multi-year agreements. A static "AI basket" allocation misses the Phase 2–3 rotation entirely.

The structural bull case rests on three pillars that the bear case cannot dislodge without a fundamental change in physical reality: construction lead times of 18–30 months mean supply cannot respond to short-term sentiment shifts; AI accelerator refresh cycles of 3–4 years mean overcapacity converts to obsolescence faster than it becomes stranded; and the contracts-first structure of the build-out means virtually every dollar of hyperscaler guidance is committed before a shovel enters the ground. These are not financial projections — they are engineering constraints.

The honest risks are equally structural. If enterprise AI deployment fails to materialise at scale by 2027, the demand curve reverts to the constrained scenario ($3.7T vs $5.2T in AI-specific capex). If semiconductor efficiency gains (in the tradition of DeepSeek V3) suppress training demand, NVIDIA's backlog clears faster than consensus models. If hyperscaler ROI discipline breaks down under debt pressure, the capex/revenue ratio remains above 50% indefinitely — compressing free cash flow precisely when patient capital needs a return. These risks are real. The risks are precisely why A.L. Capital Advisory distinguishes High Conviction infrastructure suppliers (NVDA, VRT, EQIX, CEG) from Selective hyperscaler equity positions (MSFT, GOOGL) — infrastructure suppliers are paid regardless of which ROI scenario resolves; hyperscaler equity positions are not.

The monitoring framework is straightforward: vacancy below 3% is healthy; above 6% is the early warning. Capex/revenue ratios declining from 2026 onwards signal the ROI inflection the market is waiting for. Enterprise AI deployment rate — currently in the early phase — is the key leading indicator for whether the McKinsey base case ($6.7T by 2030) is achieved or exceeded. A.L. Capital Advisory updates these readings quarterly.

The conviction ratings in this paper are produced by scoring each security across five criteria, each weighted by its relative importance to long-term AI infrastructure returns. The criteria and weights are stated below. Scoring is on a 1–5 scale (5 = strongest). A score above 19 qualifies for High Conviction; 14–18 for Selective; below 14 for Avoid.

The following scenario tables show how the investment thesis for each high-conviction position varies under different assumptions. The base case is used throughout the main paper. The bull and bear cases are not price targets — they define the range of outcomes that would cause a material re-rating of the conviction.

The AI capex cycle investment thesis is straightforward to track. Five metrics, updated each earnings quarter, determine whether the base case is intact, accelerating, or showing early bear-case signals. A.L. Capital Advisory monitors each figure below against the thresholds defined in the Model Bridge. The most critical single variable is the hyperscaler capex/revenue ratio — when this begins declining, it signals the AI ROI inflection that re-rates cloud infrastructure equities.