What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Do not size AI infrastructure stocks as a single, interchangeable trade. Treat the theme as a chain—chips, platforms, data centers, hardware, and power—and size each position according to its customer concentration, revenue mechanism, financing needs, and sensitivity to a common hyperscaler-spending slowdown. Rising capital expenditure (capex) shows that companies are committing resources; it does not by itself prove that suppliers will earn attractive returns or generate durable cash flow.
The practical method is to map your direct holdings and fund look-through exposure, identify how many positions depend on the same customers or projects, stress-test delays and lower utilization, and then fit the resulting risk inside your entire portfolio. No source establishes one suitable percentage for every investor.
What rising AI capex does—and does not—tell a stock investor
Hyperscalers are spending heavily on computing, networking, facilities and electricity for artificial-intelligence workloads. J.P. Morgan Asset Management reports that AI capex rose from 33% of hyperscalers’ cash flow from operations in 2023 to an estimated 93% in 2026. Stanford Institute for Human-Centered Artificial Intelligence’s AI Index Report 2026 estimates annual hyperscaler capex at roughly $430 billion in 2026; its 2025 and 2026 figures are explicitly estimates. Goldman Sachs discusses a forecast of more than $500 billion of AI-company investment in 2026.
Those figures are not interchangeable. The J.P. Morgan statistic is a capex-to-cash-flow ratio, Stanford’s number is an estimate of annual hyperscaler capex, and Goldman’s figure is a forecast covering AI-company investment. None is a promise of revenue to a particular semiconductor, equipment, data-center or power company.
#1 Best Overall
J.P. Morgan’s investment framework is important: investors are increasingly conditioning rewards for additional capex on evidence of greater final demand. In other words, announced spending must eventually become deployed capacity, paying workloads and cash generation. A large construction program can still produce weak equity returns if projects are delayed, utilization is low, financing costs rise or customers change specifications.
Why the spending chain matters
AI infrastructure exposure spans several business types. Semiconductor companies may sell accelerators, memory or networking silicon. Hyperscalers and other platforms buy and deploy that equipment while monetizing cloud or software services. Data-center owners and operators provide buildings, power and cooling. Technology-hardware suppliers integrate servers, racks and networking systems. Power companies serve the large new loads and may need generation, transmission or grid upgrades.
These businesses share a spending driver but not the same economics. A chip designer may face product-cycle and inventory risk; a data-center operator may carry long-lived assets and debt; a power company may depend on permitting and interconnection; and a platform may spend years before a new capacity block reaches target utilization.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteStart with an exposure map, not a ticker list
Build one inventory containing every relevant individual stock, exchange-traded fund and diversified fund. Record the market value of each holding and the percentage of the fund that is exposed to each company when that information is available. Then assign each exposure to its primary role in the chain.
| Role in the AI infrastructure chain | Typical revenue mechanism | Distinctive risks to examine | Position-sizing question |
|---|---|---|---|
| Semiconductors and components | Sales of accelerators, memory, networking silicon, packaging or related components | Product cycles, customer concentration, inventory corrections, manufacturing capacity and specification changes | How much revenue depends on a few platforms continuing to increase orders, and how quickly could orders be cut? |
| Hyperscalers and platforms | Cloud consumption, software subscriptions, advertising or other services supported by owned infrastructure | Very large upfront capex, uncertain utilization, depreciation, energy availability and slower monetization | Can the company fund the buildout without weakening its balance sheet or other businesses? |
| Data-center owners and operators | Leases, colocation fees, managed capacity or facility operation | Power and grid interconnections, construction delays, tenant concentration, refinancing and stranded capacity | Are contracted revenues firm, and who bears the cost if a project is resized or cancelled? |
| Technology-hardware providers | Servers, racks, networking equipment, cooling systems, storage and integration | Large-project timing, component availability, margin pressure and customer specification shifts | Is demand recurring, or does revenue arrive in lumpy equipment orders? |
| Power and utility businesses | Electricity sales, generation, transmission or infrastructure serving large loads | Permitting, fuel or construction costs, grid constraints, regulatory treatment and customer credit | Does the company have an executable path to serve the load, rather than only a demand forecast? |
Use the role that best describes the economic exposure, not merely the company’s marketing language. A platform that owns a data-center fleet still has different cash-flow drivers from a landlord leasing powered shells. If a business spans roles, split the exposure when segment information permits; otherwise mark the classification as approximate.
Measure look-through concentration
Several differently named holdings can amount to one capex-growth bet. Concentration can arise through a shared customer, the same project, a common component bottleneck or one financing market. Add direct ownership to indirect fund exposure before deciding that a portfolio is diversified.
Customer concentration
For every holding, list the hyperscalers or large enterprises that account for material demand. A semiconductor supplier selling to three major platforms may be more dependent on a small customer group than its ticker count suggests. A data-center operator with several facilities can still be concentrated if one tenant or one cloud expansion supplies most of the growth.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Project and geography concentration
Identify whether multiple holdings depend on the same campus, region, utility connection or construction timetable. An unavailable transmission upgrade can affect a power supplier, a data-center operator and the equipment vendors serving that site at once.
Fund overlap
For each fund, multiply your fund weight by the fund’s holding weight to estimate look-through exposure. Add overlapping positions across funds and direct accounts. When a fund does not disclose a current holding weight, record it as unknown rather than treating the fund as diversified by default. Recheck the calculation after index rebalances, mergers or large price moves.
Compare how each holding turns capex into cash
“AI beneficiary” is not a revenue model. Compare the path from spending announcement to realized cash generation.
Equipment seller
An equipment seller usually recognizes revenue when a customer accepts a product or shipment under the applicable contract. Examine order visibility, backlog quality, cancellation terms, inventory and the proportion of sales tied to a few platforms. A customer can increase its total AI budget while switching specifications or suppliers, so aggregate capex is not a guarantee for every vendor.
Facility owner or operator
A facility business converts construction and power capacity into lease or colocation revenue. Check whether capacity is contracted, when it is scheduled to become operational, who funds construction and what happens if a tenant delays deployment. Debt maturities and interest costs matter because the asset is built before all cash is collected.
Power supplier
A power company benefits only if generation, transmission and interconnection projects can be permitted, financed and completed. A forecast of data-center load may be revised before it becomes delivered electricity. Review regulatory recovery, contract structure, construction timelines and exposure to a single large customer.
Platform or hyperscaler
A platform spends first and seeks returns through cloud services, software, advertising or other products. Assess whether customer demand is growing fast enough to absorb new capacity, how depreciation will affect margins and whether energy or data-center constraints limit deployment. A strong balance sheet can reduce financing risk but does not eliminate utilization risk.
Stress-test the common slowdown before choosing a position size
Use scenarios rather than a single capex forecast. The U.S. Securities and Exchange Commission’s Corgi ETF Trust II registration statement identifies slower AI spending, unavailable power or grid interconnections, and changed customer specifications as reasons projects may be delayed, resized or cancelled. Advanced Micro Devices’ 2025 Form 10-K, filed February 4, 2026, adds that customers may lack data-center capacity, sufficient energy or capital for AI-infrastructure deployments.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →| Scenario | What changes | Questions for each business type |
|---|---|---|
| Growth continues, but more slowly | Capex rises at a lower rate than current estimates | Which companies can remain profitable without a rapid increase in orders or utilization? |
| Project delay | Power, permits, construction or equipment delivery moves to a later quarter | What revenue, cash flow and interest expense are exposed to the timing change? |
| Project resizing or cancellation | A customer reduces capacity or changes technical specifications | Are deposits refundable, can inventory be redirected, and who absorbs stranded construction cost? |
| Lower customer utilization | Installed capacity is available but workloads do not fill it | How do lease rates, cloud margins, depreciation and returns on invested capital respond? |
| Financing or energy constraint | A customer cannot obtain capital or adequate electricity | Does the holding have enough liquidity, and are projects dependent on one interconnection or lender? |
For each holding, write the expected effect on revenue, operating margin, free cash flow, debt service and valuation if the scenario occurs. Do not assume every company experiences the same percentage change. A supplier may lose an order immediately, while a contracted operator may see a delayed start date, and a regulated utility may recover some investment under a different timetable.
Use a risk-budget method instead of a universal percentage
The evidence does not support a standard AI-infrastructure allocation. Your total portfolio, investment horizon, liquidity needs, tax situation, existing concentration and tolerance for loss determine whether a thematic position is appropriate.
A practical worksheet is:
- Set a thematic loss budget. Decide how much of the total portfolio you could tolerate losing from an adverse AI-capex scenario without selling other assets to meet a cash need. This is a personal risk limit, not a market forecast.
- Estimate scenario loss for each holding. Apply your delay, slowdown and cancellation assumptions to the company’s revenue and cash-flow exposures. Use company filings and fund disclosures; label uncertain inputs.
- Aggregate correlated losses. Add losses for holdings that would be affected by the same hyperscaler, project or power bottleneck. Do not reduce the estimate merely because the securities have different names.
- Allocate the budget across roles. If one role dominates the estimated loss, reduce or rebalance before adding another holding with the same dependency. A basket spanning chips, facilities and power is not automatically balanced if all three rely on one expansion cycle.
- Check portfolio-wide impact. Include employer stock, retirement funds, broad-market funds and options. A new position can raise technology or growth exposure even when its label is “utility” or “infrastructure.”
- Define a review trigger. Revisit the map when a major customer changes capex guidance, a project misses an interconnection milestone, a holding reports a material backlog change or fund weights move substantially.
Conceptually, position weight should be no larger than the loss budget divided by the estimated loss from the scenario, adjusted for overlap with other holdings. The calculation is only as credible as its assumptions; it is a discipline for exposing concentration, not a mathematically precise forecast.
What to compare before buying or adding
| Comparison axis | Evidence to collect | Why it changes sizing |
|---|---|---|
| Dependence on hyperscaler customers | Customer disclosures, concentration notes, contract terms and backlog | A small customer group can make the position’s risk much larger than its industry label suggests. |
| Revenue source | Equipment sales, facility operation, construction revenue or power supply | Timing, margins, working capital and cancellation exposure differ by mechanism. |
| Project execution | Permits, interconnection status, construction schedule and committed tenants | Announced capacity can fail to become operating capacity. |
| Demand conversion | Utilization, paid workloads, cloud consumption, lease commencements or delivered megawatt-hours | Realized demand tests whether capex is producing cash rather than only assets. |
| Balance-sheet capacity | Cash, debt maturities, capital commitments and financing terms | Large buildouts can require funding before revenue arrives. |
| Portfolio overlap | Direct and fund holdings, shared customers and common projects | Correlation can turn several small positions into one large thematic bet. |
The reviewed evidence does not provide a comparable, current valuation table for a selected basket of stocks. Do not rank securities by price-to-earnings, enterprise-value-to-sales or another multiple unless you have current, date-stamped figures from consistent filings or market data. A low multiple can reflect higher execution or financing risk, while a high multiple can embed assumptions about future demand; neither conclusion follows from capex totals alone.
Common sizing mistakes and how to correct them
Mistake: treating forecasts as results
Stanford’s 2025 and 2026 capex figures are estimates, and Goldman Sachs’ more-than-$500-billion 2026 figure is a forecast. Label every projection with its source, date and scope. Separate announced budgets from money actually spent and from revenue recognized by a supplier.
Mistake: counting categories as diversification
Owning a chip company, a data-center operator and a power supplier can still be one trade if all depend on the same hyperscaler buildout. Map the customer and project links before declaring the basket diversified.
Mistake: ignoring the customer’s ability to deploy
Demand can be constrained by data-center capacity, energy or customer capital. AMD’s filing makes these constraints explicit. A purchase order or public spending target does not remove the possibility of delayed deployment.
Mistake: sizing from the headline market number
A $430-billion hyperscaler-capex estimate describes spending scale, not the addressable revenue or margin for any one company. Determine what portion of that spending reaches the holding, when it is recognized and what costs are required to serve it.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesMistake: using a fixed thematic percentage
A fixed percentage ignores the rest of the portfolio and the investor’s cash needs. Use a personal loss budget and scenario analysis, then cap the position where correlated downside would exceed that budget.
Mistake: overlooking liquidity and timing
Equipment orders, facility commencements and power projects can be lumpy. A position that is acceptable for a long horizon may be unsuitable if you need the money before a project reaches operation. Match the holding’s cash-flow timeline to your own.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.A repeatable review checklist
- Have all direct stocks and fund look-through holdings been listed at current weights?
- Has each exposure been assigned to chips, platforms, data centers, hardware, power or another clearly defined role?
- Which customers, campuses, utilities or financing sources are shared?
- What evidence shows that announced capex is becoming deployed, paid demand?
- What happens to revenue, margins, cash flow and debt service if a project is delayed, resized or cancelled?
- Could power, grid interconnection, customer capital or specification changes block deployment?
- Who bears construction and stranded-capacity costs?
- Are the spending figures observed results, company guidance, estimates or forecasts?
- Does the combined downside fit your total-portfolio loss budget and liquidity needs?
- What factual event will trigger a new review?
How to interpret the major spending estimates
Use the three widely cited figures as context, not as a target price or allocation signal.
| Figure | What it measures | Status and limitation |
|---|---|---|
| 33% of hyperscalers’ cash flow from operations in 2023 | AI capex as a share of hyperscaler operating cash flow | Historical figure reported by J.P. Morgan Asset Management; it is a ratio, not a supplier-revenue forecast. |
| 93% in 2026 | AI capex as a share of hyperscaler operating cash flow | Estimated 2026 figure reported by J.P. Morgan; it is not realized spending. |
| Roughly $430 billion in 2026 | Annual hyperscaler capex | Stanford AI Index 2026 estimate; 2025 and 2026 values are marked as estimates. |
| More than $500 billion in 2026 | AI-company investment | Goldman Sachs forecast; broader category and not an observed result. |
Differences among these numbers are expected because they use different populations, definitions and measurement bases. The correct use is to frame scale and intensity, then test whether a particular holding can convert that spending into durable cash flow.
Free tools Windows power users keep installed
One-click scans. No signup required.
When to reduce, pause or avoid adding exposure
Consider reducing or postponing an addition when your look-through exposure already depends on the same customer group, when a project has unresolved power or interconnection requirements, or when the holding must raise substantial capital before revenue begins. A pause can also be rational when a forecast has moved faster than evidence of utilization or paid demand.
Best Value
That is not a prediction that AI investment will fail. It is a recognition that spending growth, stock returns and business cash flow are separate outcomes. Requiring evidence of deployment and customer economics helps prevent a headline capex number from dictating a position size that your portfolio cannot absorb.
Bottom line
Size AI infrastructure stocks by dependency and downside, not by the size of the AI-spending headline. Map every holding by role, look through funds, identify shared hyperscaler and project exposure, compare each revenue mechanism, and stress-test delays, cancellations, power constraints, specification changes and lower utilization. Then set the combined position within a personal portfolio risk budget. The available spending estimates explain why the theme matters; they do not establish a universal allocation or guarantee attractive returns for any security.
Frequently Asked Questions
Does higher AI capex automatically make AI infrastructure stocks more attractive?
No. Higher spending can support suppliers, but investors still need evidence that projects become deployed capacity, paid demand and sustainable cash flow. Spending can be delayed, resized or cancelled.
Recommended Free Tools
How can I find overlap between an AI ETF and stocks I already own?
List each fund’s current holdings and weights, multiply them by your fund position, and add those look-through amounts to direct holdings. Then group the total by customer, project and infrastructure role.
Should chips, data centers and power companies receive equal portfolio weights?
There is no evidence-based equal-weight rule. Their customer concentration, contracts, financing, construction and regulatory risks differ, so weights should follow your portfolio risk budget and scenario losses.
Are the 2026 AI spending numbers actual spending?
No. Stanford labels its 2025 and 2026 capex values as estimates, J.P. Morgan’s 93% figure is an estimate, and Goldman Sachs’ more-than-$500-billion figure is a forecast. They should not be presented as realized results.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
