S&P 500 AI Concentration: How Much of the Index Is AI-Linked?
The S&P 500 is concentrated, and most of its top-10 weight sits in a conservative AI-linked cohort.
State Street's September 9, 2026 S&P 500 index holdings show the 10 largest security lines at 38.09% of the index. Grizzly Bulls classifies seven companies with directly evidenced AI-infrastructure supply or large-scale AI/cloud-platform infrastructure economics as a conservative AI-linked cohort. Together they represented 29.46% of the index, or 77.3% of the top-10 weight.
The top 10 security lines were 38.09% of the index
Alphabet appears twice because Class A and Class C are separately weighted securities. That is why the table has 10 security lines but only nine companies.
September 9, 2026 S&P 500 top holdings
What counts as AI-linked in the 29.46% figure?
The screen is deliberately narrower than “companies with an AI product.” Inclusion requires direct current company evidence that a material part of the business or capital-spending cycle is tied to AI infrastructure supply or large-scale AI/cloud platform infrastructure.
Two sides of the same infrastructure cycle
Why Apple is excluded
7.05% of the September 9 index snapshot.
Apple has a material AI strategy, but the current primary financial disclosure used here does not separately quantify AI revenue or an infrastructure-spending channel comparable to the conservative cohort.
Why Tesla is excluded
1.58% of the September 9 index snapshot.
Tesla explicitly describes AI initiatives and AI-driven capex, but its current AI economics are embedded in automotive, robotics and energy rather than the common datacenter/cloud supplier-platform cycle used for the core cohort.
Why these seven companies passed the conservative screen
The classification uses current issuer financial disclosures rather than a thematic label. The metrics below are examples of the evidence supporting inclusion; they are not meant to make unlike business models directly comparable.
NVIDIA
Fiscal 2026 revenue was $215.9 billion. Fourth-quarter Data Center revenue was $62.3 billion, up 75% year over year.
Broadcom
Q3 AI semiconductor revenue was $16.7 billion of $29.591 billion total revenue, or about 56.4%.
Micron
Cloud Memory plus Core Data Center revenue was $25.293 billion in Q3, about 61.0% of total revenue.
Microsoft
Quarterly capital expenditures were $41 billion, with roughly two thirds, primarily CPUs and GPUs. Microsoft described roughly $175 billion of calendar-2026 investment after its lease-accounting shift.
Alphabet
Q2 capital expenditures were $44.9 billion. Google Cloud revenue was $24.8 billion, up 82% year over year.
Amazon
Q2 cash capital expenditures were $53.1 billion and first-half cash capex was $96.3 billion. Amazon said the spending was primarily technology infrastructure, with the majority supporting AWS growth, plus fulfillment capacity.
Meta
Meta guided to $130–$145 billion of 2026 capital expenditures to support AI efforts and the core business.
Today's concentration has a surprisingly close 1965 comparison
S&P Dow Jones Indices identified the mid-1960s as the last comparable concentration era. Its June 1965 top 10 add to 38.24%, almost identical to the 38.09% September 2026 snapshot. The concentration looks similar; the companies do not.
Top-10 concentration: June 1965 vs. September 2026
Concentration is a risk to understand, not proof the index is broken
A cap-weighted index deliberately lets companies become larger portfolio weights as their float-adjusted market values rise. That creates concentration when a small group of winners dominates, but it is also how an index investor can own a future winner before knowing which company will become one.
The 29.46% cohort also does not mean seven companies are one trade. NVIDIA, Broadcom, Micron, Microsoft, Amazon, Alphabet, and Meta have different customers, margins, capital structures, and revenue models. The common point is narrower: a large fraction of index weight has direct evidence of exposure to the same broad AI compute and infrastructure investment cycle.
Equal weighting reduces mega-cap concentration, but it is not a free diversification upgrade. S&P DJI notes that equal weighting changes size, value, and anti-momentum exposure and requires periodic rebalancing. A concentration observation by itself is also not a complete market-timing rule.
Methodology and limitations
- Freeze State Street's published September 9, 2026 S&P 500 top-holdings weights and recompute the top-10 sum from those 10 security lines.
- Keep Alphabet Class A and Class C as separate index security lines but one economic company in unique-company counts.
- Include a company in the conservative AI-linked cohort only when current primary-source financial disclosure directly supports material AI-infrastructure supply or large-scale AI/cloud platform infrastructure economics.
- Recompute the infrastructure-supplier, platform/infrastructure-buyer, combined cohort, and cohort-share-of-top-10 figures from the frozen security-line weights.
- Use S&P DJI's historical research for the June 1965 top-10 comparison and capitalization-weighting interpretation. Do not treat the August 31 official 37.8% snapshot and September 9 State Street 38.09% snapshot as same-date measurements.
This is a dated concentration study, not a forecast of AI returns, a sector-classification system, or an estimate of AI's contribution to index revenue or earnings. The cohort can change as index weights and company disclosures change.
Primary sources
Download the concentration snapshot
The CSV contains the 10 frozen security lines, their weights, and the conservative cohort classification. The JSON adds the aggregate cohort calculation, exclusions, historical comparison, interpretation boundary, and core source URLs.
Citation and reuse
When citing the 29.46% figure, describe it as a conservative Grizzly Bulls company-level AI-infrastructure/platform exposure screen using the September 9, 2026 S&P 500 weight snapshot, not as the share of index revenue attributable to AI.