What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Yes—Nvidia is only one major beneficiary of the AI build-out. Spending also reaches rival accelerator and server-CPU suppliers, custom-chip designers, cloud platforms, and data-center operators. Their reported figures measure different things: some are AI-specific revenue, some combine AI with other products, and some are backlogs or capital expenditures rather than sales. That makes a supply-chain view more useful than a simple list of “AI stocks.”
AI is an infrastructure supply chain, not a single-product market
Generative AI services require several layers to work together. Chip designers supply accelerators and server processors; cloud companies turn those chips into rentable computing capacity; networking moves data between processors and storage; and data-center operators provide power, cooling, space and interconnection. A company can benefit from AI demand without selling a branded graphics processor.
As an Amazon Associate I earn from qualifying purchases.
The financial exposure differs at every layer. A semiconductor vendor may report product revenue, a cloud provider may report a service backlog while spending billions on capacity, and a data-center landlord may disclose leasing or partnership activity without breaking out AI-only sales. Revenue growth therefore does not prove that a business is highly profitable, and a large capital-spending program is not revenue at all.
Nvidia remains the benchmark—but its figures are not the whole AI economy
Nvidia is the clearest reference point because its accelerators and networking products are widely used in AI clusters. In its fiscal 2026 annual report, Nvidia reported $215.9 billion of total revenue, up 65% year over year. The same filing reported data-center compute revenue growth of 59% and data-center networking revenue growth of 142%. Those data-center measures include products used for AI, but they are not a complete measure of Nvidia’s AI profit.
#1 Best Overall
You can verify the company’s fiscal-period definitions and segment disclosures in its fiscal 2026 annual report filed with the SEC. Nvidia’s scale sets the comparison point, but the other businesses below participate through different products and economics.
AMD combines AI accelerators with the CPUs that run servers
What AMD sells
AMD supplies both general-purpose server CPUs and Instinct accelerators. CPUs handle operating-system tasks, databases and parts of AI workloads; accelerators perform the matrix calculations used to train and serve models. Cloud operators and enterprises often buy both, so a supplier can gain from AI-server expansion even when the disclosed number is not accelerator-only.
What the reported numbers mean
AMD reported $16.6 billion in data-center net revenue for fiscal 2025, up 32%. The company attributed that performance primarily to demand for fifth-generation EPYC processors and Instinct MI350 GPUs. Because the data-center category includes EPYC CPUs as well as accelerators, it must not be read as pure AI revenue. AMD’s fiscal 2025 annual report provides the company’s definitions.
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 →AMD’s quarterly disclosure shows how quickly the mix can change without becoming a clean AI metric. In the second quarter of fiscal 2026, data-center segment revenue was $6.7 billion, up 107% year over year, driven by EPYC processors and Instinct GPUs. The Q2 fiscal 2026 results release reports that segment figure; it does not isolate accelerator revenue.
Why AMD matters to the supply chain
- Alternative capacity: Customers can diversify accelerator purchases instead of relying on one supplier.
- CPU and accelerator cross-selling: The same data-center build can require EPYC CPUs for host and control functions alongside Instinct GPUs.
- Execution risk: Product launches, software compatibility, packaging capacity and customer qualification affect how much of the announced demand becomes revenue.
Broadcom monetizes custom AI chips and the network around them
Custom accelerators are a different business model
Large cloud companies sometimes design chips tailored to their own workloads. An application-specific accelerator can improve performance, power efficiency or supply control for a particular service, but it generally depends on a few very large customers. Broadcom supplies design and semiconductor solutions for these custom systems and also benefits when AI clusters need high-speed networking.
Rank #2
- Ideal for Gifting
- Ideal for a bookworm
- Compact for travelling
A specifically labeled AI revenue figure
Broadcom reported $16.7 billion of AI semiconductor revenue in its third quarter of fiscal 2026, up 221% year over year. Chief executive Hock Tan said, “Q3 AI semiconductor revenue of $16.7 billion grew 221% year-over-year, and 54% quarter-over-quarter,” in the company’s results announcement. The Q3 fiscal 2026 release identifies demand for custom AI accelerators and AI networking as the drivers.
Broadcom’s label is more AI-specific than AMD’s data-center segment, but it still does not show how much came from each product or customer. Concentration, design-win timing and the cost of developing each custom platform can materially affect future results.
Cloud operators earn from AI usage while funding the infrastructure
AWS: internal chips plus rented compute
Amazon monetizes AI in two connected ways. Its cloud business sells compute, storage and managed machine-learning services, while its semiconductor group develops chips that can lower the cost or improve the availability of that compute. Amazon said the combined annual revenue run rate for its Trainium and Graviton chips exceeded $10 billion. That is an Amazon statement about its own business, not an independently audited measure of AI-only profit.
Amazon also said Trainium2 was fully subscribed, with 1.4 million chips landed, and described use in Bedrock and Project Rainier. “Fully subscribed” indicates customer demand for available capacity; it does not mean every chip has generated the same amount of revenue or that the program has a disclosed margin. The company’s Q4 2025 results contain those disclosures.
The cloud trade-off
Cloud providers must spend ahead of demand. They buy chips, construct or lease facilities, connect regions and reserve electricity before customers’ workloads are fully visible. Utilization, pricing, depreciation and power costs determine whether rising AI demand translates into attractive returns. A cloud operator can therefore report strong AI bookings while its near-term cash flow is pressured by construction and equipment spending.
Alphabet shows why backlog and capital expenditure are not revenue
Alphabet’s Google Cloud is another route for AI spending to reach a company that is not merely selling a standalone accelerator. Cloud customers pay for computing, storage, networking and managed AI tools; Alphabet also designs its own tensor-processing hardware and builds the capacity needed to run those services.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →At the end of Q2 2025, Alphabet reported a $106 billion Google Cloud backlog. In the same dated earnings call, it reported $22.4 billion of capital expenditure for Q2 2025 and guided to approximately $85 billion for full-year 2025. Those figures are historical 2025 disclosures, not current 2026 guidance. The company said, “Our AI infrastructure investments are crucial to meeting the growth in demand from Cloud customers.” The Q2 2025 earnings-call transcript provides the context.
A backlog represents contracted or committed work expected to be recognized over time; it is not the same as current-period revenue. Capital expenditure is money invested in servers, data centers and related assets; it is not sales. Investors should keep those categories separate when comparing Alphabet with chip companies.
Data-center operators provide the scarce physical layer
AI clusters need high-density space, redundant power, advanced cooling, fiber connections and direct links to other networks. Operators that supply those facilities can benefit even when they do not manufacture a processor. Their contracts may be with cloud companies, enterprises, network providers or specialized AI tenants.
Equinix’s Q2 2026 earnings release discusses AI-infrastructure demand and partnership activity, but it does not quantify AI-only revenue. Its Q2 2026 release is therefore evidence of strategic exposure and customer activity, not a directly comparable AI sales figure.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC 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 & 11Rank #4
What to watch in this layer
- Signed capacity versus occupied capacity: A development pipeline can take years to power and fill.
- Power availability: Grid interconnection and local permitting can constrain growth more than customer demand.
- Capital intensity: New halls, substations and cooling systems require substantial upfront spending before rent is collected.
- Contract structure: Lease length, customer concentration and pass-through clauses affect risk and cash flow.
Other suppliers benefit, but current AI-only figures are not established here
Manufacturers that fabricate advanced chips, suppliers of high-bandwidth memory, and companies providing power equipment, cooling and electrical contracting are necessary parts of the AI build-out. However, the available company disclosures do not provide a directly reviewed, current AI-only revenue figure for those categories. It is more accurate to identify them as unquantified supply-chain beneficiaries than to assign a precise market share or growth rate.
The same caution applies to any claim that a particular supplier “gets rich” from AI. A vendor may have strong orders while facing wafer, packaging, memory, labor or electricity constraints. Without a comparable disclosure, strategic importance should not be presented as measured AI revenue.
How to compare AI beneficiaries without mixing incompatible numbers
| Business | What it sells | Disclosed measure | AI specificity | Main comparison risk |
|---|---|---|---|---|
| Nvidia | Accelerators, server systems and networking | $215.9B total revenue in fiscal 2026; data-center compute +59% and networking +142% | Data-center figures are not a complete AI-profit measure | Large segment includes more than one product category |
| AMD | EPYC server CPUs and Instinct accelerators | $16.6B data-center revenue in fiscal 2025; $6.7B segment revenue in Q2 fiscal 2026 | Broader segment, not AI-only | CPU and GPU contributions are combined |
| Broadcom | Custom AI accelerators and AI networking | $16.7B AI semiconductor revenue in Q3 fiscal 2026 | Specifically labeled AI semiconductor revenue | Product and customer contributions are not broken out |
| Amazon Web Services | Cloud compute and managed services; Trainium and Graviton chips | More than $10B combined annual chip revenue run rate | Company-stated chip run rate, not a full AI-profit figure | Cloud capex, utilization and operating costs affect returns |
| Alphabet/Google Cloud | Cloud services and internally designed AI hardware | $106B backlog at Q2 2025; $22.4B Q2 2025 capex | Backlog and capex are not AI revenue | Historical period and mixed cloud workloads |
| Equinix | Data-center space, power, networking and interconnection | Q2 2026 release discusses AI demand and partnerships | No AI-only revenue figure stated | Strategic exposure cannot be converted into a sales percentage |
Use the same fiscal period where possible. Do not compare a quarterly number with an annual number without labeling it, and do not infer shareholder returns from segment growth alone. Revenue, backlog, capital expenditure, bookings and profitability answer different questions.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Key risks behind the AI revenue story
Demand can be concentrated
Custom-chip and accelerator suppliers may depend on a small number of cloud or enterprise buyers. A delayed deployment, redesign or change in a customer’s model strategy can move orders between quarters.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsCapacity can lag demand
Advanced packaging, memory, electricity and powered data-center space cannot always expand as quickly as software demand. A company may have purchase commitments but be unable to recognize revenue until hardware is installed and accepted.
Best Value
- It can be a gift option
- Comes with secure packaging
- Helpful in various ways
Software and switching costs matter
Accelerator performance depends on compilers, libraries, drivers and developer tools. A technically competitive chip may take time to qualify if customers have built workflows around another platform.
High growth can require high spending
Cloud operators and data-center landlords invest before utilization is certain. Depreciation, financing costs, energy prices and maintenance can reduce the cash generated by each additional AI customer.
Fiscal calendars obscure the timing
“Q3 fiscal 2026” does not necessarily cover the same months as another company’s third quarter. Always read the period in the cited filing before drawing a growth comparison.
Free tools Windows power users keep installed
One-click scans. No signup required.
A practical checklist for evaluating an AI beneficiary
- Identify the product: Is the company selling an accelerator, CPU, custom chip, cloud service, networking, or physical capacity?
- Classify the metric: Mark it as AI-specific revenue, a broader segment, backlog, bookings, capital expenditure or a run rate.
- Check the period: Record the fiscal quarter or year, growth rate and whether the number is recurring, historical or guidance.
- Test concentration: Look for dependence on one cloud customer, one design win, one region or one construction project.
- Estimate execution needs: Consider software support, manufacturing and packaging, power, permitting, cooling and staffing.
- Separate sales from returns: Review margins, cash flow and capital intensity rather than assuming that fast revenue growth equals “rich” profits.
What “benefiting from AI” really means
Nvidia captures the largest and most visible share of accelerator spending, but the AI economy is broader than its income statement. AMD participates with CPUs and accelerators; Broadcom with custom silicon and networking; Amazon and Alphabet with cloud capacity and internally designed chips; and data-center operators with the physical infrastructure that lets all of those systems run. The most credible comparison keeps each company’s product, metric and fiscal period intact instead of treating every AI-related number as interchangeable.
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.
