FSKAX is marketed as owning the whole US market, but it is market-cap weighted, so a handful of names dominate it. These top 15 companies are ~39.7% of the entire fund, and roughly 34.4% of the fund sits in AI-exposed names (about 27.2% in the AI/chip core). That means FSKAX is, to a first approximation, a leveraged bet on the artificial-intelligence capital-expenditure cycle. The purpose here is to pressure-test that bet at the company level and answer one question directly: what happens to these holdings if the AI trade re-concentrates in a few winners, or if the data-center buildout that underwrites it slows - a risk that just became concrete when New York imposed the first statewide moratorium on hyperscale data centers on July 14, 2026.
Ordered by weight in FSKAX (Alphabet's two share classes consolidated). "Rating" is exposure to the data-center buildout and the moratorium wave.
| # | Company | FSKAX wt. | Type | Latest report | Buildout exposure | Note |
|---|---|---|---|---|---|---|
| 1 | NVIDIA NVDA | 6.96% | AI/chip core | Fiscal Q1 FY2027 | Demand-levered | ~92% of revenue is data center; ~54% from just 3 customers |
| 2 | Alphabet GOOGL / GOOG | 5.77% | AI/chip core | Q1 2026 | Direct-builder | builds its own data centers; capacity-constrained |
| 3 | Apple AAPL | 5.71% | AI-exposed | Fiscal Q2 2026 | Minimal | buys partner GPU capacity; not a hyperscale builder |
| 4 | Microsoft MSFT | 4.34% | AI/chip core | Fiscal Q3 FY2026 | Direct-builder | ~$190B 2026 capex; no disclosed NY project |
| 5 | Amazon AMZN | 3.71% | AI/chip core | Q1 2026 | Direct-builder | ~$200B 2026 capex, mostly AWS AI |
| 6 | Broadcom AVGO | 2.84% | AI/chip core | Fiscal Q2 FY2026 | Demand-levered | 6 core custom-silicon customers; high concentration |
| 7 | Meta Platforms META | 1.92% | AI/chip core | Q1 2026 | Direct-builder | mega-sites in OH and LA; capex raised to $125-145B |
| 8 | Tesla TSLA | 1.54% | AI-exposed | Q1 2026 | Minimal | captive compute at Giga Texas; not a hyperscale builder |
| 9 | Berkshire Hathaway BRK.B | 1.25% | Non-AI | Q1 2026 | Minimal | AI immaterial; indirect via Apple stake |
| 10 | JPMorgan Chase JPM | 1.21% | Non-AI | Q2 2026 | Minimal | AI a productivity lever, not yet material |
| 11 | Eli Lilly LLY | 1.06% | Non-AI | Q1 2026 | Minimal | AI in drug discovery; not yet material |
| 12 | ExxonMobil XOM | 0.92% | Non-AI | Q1 2026 | Beneficiary | wants to SELL gas-fired power to data centers |
| 13 | Micron Technology MU | 0.83% | AI/chip core | Fiscal Q3 FY2026 | Demand-levered | ~61% data-center revenue; but HBM sold out through 2027 |
| 14 | Walmart WMT | 0.83% | Non-AI | Fiscal Q1 FY2027 | Minimal | AI a margin/productivity lever |
| 15 | Advanced Micro Devices AMD | 0.83% | AI/chip core | Q1 2026 | Demand-levered | GPU demand rides on Meta / OpenAI / Oracle |
Each card reads from the company's most recent quarterly report (primary sources). For the five non-AI names, the deep AI/chip and data-center dimensions are replaced by a short AI-exposure note, per scope.
Revenue $81.6B, +85% YoY. GAAP net income $58.3B; operating income $53.5B. Gross margin 74.9%, operating margin ~65.6%, net margin ~71.4%. No China H20 charge this quarter; Q2 guide assumes zero China data-center compute.
Data Center revenue $75.2B, +92% YoY = ~92% of total; within it, compute $60.4B (+77%) on the Blackwell 300 ramp plus InfiniBand / Spectrum-X / NVLink networking. Edge Computing (gaming, pro-viz, auto) $6.4B (+29%). The moat is the CUDA software ecosystem plus a full rack-scale system.
Q2 FY2027 guide: revenue $91.0B ±2%, gross margin ~74.9%. Blackwell shipping at scale with 'extended' lead times; Vera Rubin begins shipping fiscal Q3 FY2027. CEO Huang calls the AI-factory buildout 'the largest infrastructure expansion in human history.'
NVIDIA's revenue is everyone else's AI capex, so it is the purest read on the trade. Hyperscalers are ~50% of data-center revenue (~$37.9B, +115% YoY). Concentration risk (10-Q): three direct customers = 21%, 17%, 16% of revenue (~54% combined). If the buildout pauses, the hit is fast and concentrated here.
A leveraged bet on continued data-center construction. NY's July 14, 2026 moratorium on facilities ≥50 MW, plus any spreading permit-pause trend, would lengthen already-extended lead times and could turn deferrals into cancellations. NVIDIA discloses no state-level exposure (UNCONFIRMED), and its forward guide is still accelerating, so no slowdown is visible yet.
Revenue $109.9B, +22% YoY. Operating income $39.7B; operating margin 36.1%. Net income $62.6B, +81% (boosted by ~$36.9B of pretax equity-security gains, so operating income is the cleaner read). Google Cloud $20.0B, +63%; Cloud operating margin 32.9% (vs 17.8% a year ago).
Search advertising is the engine: Google Services revenue $89.6B (+16%), segment operating income $40.6B (Search & other $60.4B +19%; YouTube ads $9.9B +11%). Cloud is the second lever, with operating income tripling year over year.
8th-generation TPU (TPU 8t = 3x the processing of 7th-gen Ironwood; TPU 8i = 80% better performance per dollar). Gemini 3.1 Pro; Gemma 4 downloaded 50M+ times. Will be among the first to offer NVIDIA's Vera Rubin NVL72. First-party models process 16B tokens/minute via API, up from 10B.
Revenue from gen-AI products grew ~800% YoY; Cloud backlog ~$460B (more than doubled QoQ); Gemini Enterprise paid users +40% QoQ. Q1 capex $35.7B; full-year 2026 capex guidance raised to $180-190B, with 2027 to 'increase meaningfully.'
Management said Cloud 'revenue would have been higher with more capacity' - a real near-term constraint. Specific data-center sites and any NY/Northeast projects are not disclosed (UNCONFIRMED). The $180-190B capex is a bet that demand keeps outrunning supply.
Revenue $111.2B, +17% YoY. Net income $29.6B; EPS $2.01 (+22%). Operating margin 32.3%; gross margin 49.3%. Products gross margin 38.7% vs Services gross margin 76.7%.
iPhone leads the top line (March-quarter record $57.0B), but the margin engine is Services (76.7% gross margin): ~28% of revenue yet ~43% of gross profit. The moat is pricing power over a locked-in installed base.
Apple runs its AI servers on its own Apple Silicon (not NVIDIA/AMD) for Private Cloud Compute. Apple Intelligence is now active on 80%+ of eligible devices; a ~$1B Google Gemini deal will power a revamped Siri targeting a fall iOS 19 launch. R&D jumped ~34% YoY to $11.4B.
No standalone AI revenue line; management frames AI as 'incremental' spend, not a separate P&L. Capex stays light at ~$14.3B for 2026 (estimate) versus the five hyperscalers' collective ~$660-690B. Monetization is indirect (device upgrades, Services attach) and not yet separately quantified (UNCONFIRMED).
Not materially relevant. Apple runs limited first-party data centers and buys partner GPU capacity rather than building ≥50 MW campuses, so NY's moratorium does not target its footprint. Any exposure is indirect (partner-hosted compute cost/latency). UNCONFIRMED that Apple has any ≥50 MW NY project.
Revenue $82.9B, +18% YoY. Operating income $38.4B (+20%); net income $31.8B (+23%); EPS $4.27. Operating margin ~46.3%. Intelligent Cloud $34.7B (+30%); Azure +40%. Commercial backlog (RPO) $627B, +99%.
Two engines: Productivity & Business Processes $35.0B (+17%, Microsoft 365 Commercial Cloud +19%) and Intelligent Cloud $34.7B (+30%, Azure the driver). More Personal Computing $13.2B (-1%) is the soft spot.
AI business run rate reached $37B annualized, +123% YoY. Custom silicon (Maia 200, Cobalt) is deployed; added ~1 GW of capacity in the quarter. Multi-model strategy (OpenAI + Anthropic usage doubled QoQ). Explicitly 'capacity-constrained through 2026.'
Signal: $37B AI run rate and Azure +40% (above ~35% consensus). Cost: capex plus finance leases $31.9B, +49% YoY; forward guidance >$40B next quarter and ~$190B for calendar 2026 (+~61%, including ~$25B from higher memory pricing). Management frames the assets as supporting 'monetization over the next 15 years.'
~$190B of 2026 capex with no pullback; a Cheyenne, Wyoming expansion was announced in April. No Microsoft-owned NY hyperscale project is disclosed (the flagged Somerset, NY site belongs to TeraWulf, not Microsoft), so direct NY exposure is limited (UNCONFIRMED); the moratorium reads as a broader Northeast permitting-risk signal.
Net sales $181.5B, +17% YoY. Operating income $23.9B (13.1% margin, a record). Net income $30.3B includes a $16.8B non-cash Anthropic gain, so operating income is the truer read. AWS $37.6B (+28%, fastest in 15 quarters); AWS operating income $14.2B, ~37.8% margin.
AWS produced $14.2B of the $23.9B total operating income (~59%) on just 21% of sales. Retail runs thin (North America ~8% margin, International ~3.5%). AWS is the profit engine.
Amazon's chip business is now a >$20B revenue run rate, growing triple digits; Trainium2 is fully committed and Trainium3 nearly subscribed. It is also deploying 1M+ NVIDIA GPUs (dual-chip strategy). Bedrock customer spend +170% QoQ; a new Anthropic deal is worth >$100B.
AWS AI revenue run rate >$15B in ~3 years. Q1 capex ~$43-44B (single-quarter figure varies by source, UNCONFIRMED); full-year 2026 capex ~$200B (CEO commentary, not formal guidance). ROI language: 'customer commitments for a substantial portion.' Backlog $364B.
~$200B of 2026 capex aimed mostly at AWS AI data centers, with growth 'limited by power' and memory costs that 'skyrocketed.' No AWS NY/Northeast project was disclosed (direct exposure UNCONFIRMED); NY's queue held ~25 proposals totaling ~9,340 MW when the moratorium landed.
Revenue $22.2B, +48% YoY. GAAP net income $9.3B; operating income $10.8B. Gross margin 69.4%; operating margin 48.6%; free cash flow $10.3B (46% of revenue).
A dual engine: Semiconductor Solutions $15.0B (68% of revenue, custom AI silicon + networking) and high-margin, recurring Infrastructure Software / VMware $7.2B (32%).
AI semiconductor revenue $10.8B, +143% YoY from custom accelerators (XPUs) and AI networking. Six core custom-silicon customers, named to include Anthropic, Google, Meta and OpenAI. Tomahawk 6 (100-terabit) switch shipping; Jericho fabric. Q3 guide: total ~$29.4B (+84%), AI semi $16.0B (>200% YoY).
AI is ~49% of revenue and climbing fast (Q1 $8.4B to Q2 $10.8B to Q3 guide $16.0B). CEO Hock Tan reiterated AI semiconductor revenue 'in excess of $100 billion in fiscal 2027.'
XPU and networking demand is a derivative of hyperscaler buildout, and the six-customer concentration means a pullback by even one large buyer is material. NY's moratorium plus a possible multi-state trend could defer ≥50 MW capacity; Broadcom gave no moratorium-related guidance cut, and magnitude of any exposure is UNCONFIRMED.
Revenue $56.3B, +33% YoY. Operating income $22.9B; operating margin 41%. Net income $26.8B / EPS $10.44 includes a ~$8.0B one-time tax benefit; without it, ~$18.7B / $7.31. Family of Apps operating income $26.9B; Reality Labs operating loss $(4.0)B.
Advertising funds everything: Family of Apps revenue $55.9B produced $26.9B of operating income, covering the $4.0B Reality Labs loss. Ad impressions +19% YoY; average price per ad +12%. Family daily active people 3.56B (+4%).
Rolling out >1 GW of its own custom silicon (MTIA) co-developed with Broadcom, plus 'significant amounts of AMD chips to complement the new NVIDIA systems.' Meta Superintelligence Labs shipped its first model, 'Muse Spark.' Specific GPU counts not quantified (UNCONFIRMED).
Revenue signal is AI-driven engagement and ad-targeting gains. Q1 capex $19.8B; full-year 2026 capex guidance raised to $125-145B (from $115-135B), citing higher memory pricing and data-center costs. Total 2026 expenses guided $162-169B.
Named mega-sites: Prometheus (New Albany, Ohio, ~1 GW, 2026) and Hyperion (Richland Parish, Louisiana, scaling toward 5 GW, >$50B). Disclosed builds are in OH/LA, not NY, so direct NY-moratorium exposure appears limited (UNCONFIRMED, based on absence of disclosure).
Revenue $22.4B, +16% YoY. GAAP net income $477M (+17%); operating income $941M (+136%); operating margin 4.2%. Total gross margin 21.1% (up from 16.3%); automotive gross margin ex-credits 19.2%. Q2 2026 results were not yet out as of the research date (due July 22, 2026).
The profit lever was automotive gross-margin recovery (16.2% to 21.1%), plus high-margin energy storage (~39.5%) and +42% services. Pressure points: operating expense +37% YoY (AI/R&D and stock comp), regulatory credits down ~36%, storage revenue -12%. Underlying auto strength (19.2% ex-credits) is real.
In-house training clusters Cortex 1 (>100k H100-equivalent) and Cortex 2 (>130k, ramping). FSD approved in the Netherlands; Robotaxi launched in Dallas/Houston in April 2026. The AI5 inference chip taped out April 15, 2026 (~5x AI4, volume 2027); Tesla is adding in-house semiconductor fabrication. Optimus targets volume production in 2026.
Not materially yet. Robotaxi is in limited early operation and FSD revenue is small against the spend; 2026 capex is guided >$25B. Management frames AI, Robotaxi and Optimus as future engines, not current earnings drivers. (Q1 free cash flow was +$1.44B per the SEC filing.)
Tesla's compute (Cortex 1 and 2) sits at Giga Texas, a captive AI cluster rather than a merchant data center, and NY's ≥50 MW moratorium targets utility-interconnected facilities. Not materially relevant on the current footprint; a broad multi-state wave is a future siting watch-item (UNCONFIRMED, no NY plans disclosed).
Operating earnings $11.35B, +~18% YoY (missed ~$11.56B consensus). Net earnings $10.1B. Total revenues $93.7B. Cash and Treasuries ~$397B ($51.5B cash + $339.3B T-bills). First quarter under new CEO Greg Abel.
A diversified holding company: insurance underwriting $1.72B (+~28%), BNSF railroad $1.38B, and Berkshire Hathaway Energy $1.11B (all after-tax), plus float-funded investment income and a large equity portfolio (Apple is the biggest holding). OxyChem was acquired January 2, 2026 for ~$9.5B.
AI angle: the 10-Q shows no material AI initiative and no standalone AI revenue; AI is not financially material to Berkshire. Indirect exposure comes through the Apple stake (Apple's on-device AI and Services). The exact Apple position value this quarter was not itemized (UNCONFIRMED).
Managed net revenue $58.0B, +27% YoY. Net income $21.2B (+41%); excluding significant items $16.9B (+13%). EPS $7.70 (ex-items $6.14). ROTCE 29% reported / 23% ex-items. Net interest income $25.6B (+10%). One-offs: a $4.6B Visa-share gain (+$1.27 EPS) and $1.0B of equity gains.
Scale across four engines: Consumer & Community Banking ($5.3B net income), the Corporate & Investment Bank ($9.7B; Markets revenue $12.1B +35%, Equity +86%), Asset & Wealth Management ($2.0B, AUM $5.1T +18%), and Corporate (Visa/equity gains).
AI angle: the internal LLM Suite has reached ~200,000 users with cited 30-40% efficiency gains, part of an ~$18B tech budget and ~$2B/year of claimed AI savings (secondary sources, UNCONFIRMED against filings). Real but not yet separately material in reported results.
Revenue $19.8B, +56% YoY (65% volume, -13% price). Net income $7.4B (+168%). Gross margin 81.9%; operating income $8.9B (+141%). Non-GAAP EPS $8.55 (beat ~$6.66). Full-year 2026 guidance raised to $82-85B revenue.
The incretin franchise: Mounjaro (diabetes) $8.66B worldwide (+125%) and Zepbound (obesity) $4.16B US (+79%), together ~$12.8B and roughly $6.7B of the year-over-year growth. Lilly has committed >$55B to production capacity since 2020 (aggregate UNCONFIRMED).
AI angle: in January 2026 Lilly and NVIDIA launched an AI drug-discovery lab (up to $1B over five years on NVIDIA BioNeMo), and its TuneLab platform licenses proprietary models to biotechs. Strategic but not a current earnings contributor; no AI revenue is broken out.
Revenue ~$85.1B (secondary source, UNCONFIRMED). GAAP net income $4.18B (EPS $1.00), down ~46% YoY, driven by $3.9B of unfavorable derivative timing plus a $0.7B hedge item; earnings excluding those were $8.8B ($2.09/sh), up from $7.6B. Upstream $5.74B. Production 4.6M boe/d.
Integrated production growth (Guyana set a record >900k gross bbl/d; Permian and Guyana are the advantaged barrels), plus $15.6B of structural cost savings since 2019. Operating cash flow $8.7B; $9.2B of shareholder distributions; debt-to-capital 15.4%.
AI angle (inverse exposure): unlike the others, Exxon could benefit from data-center power demand. It is developing behind-the-meter gas-fired power for hyperscalers (Permian gas + carbon capture removing >90% of CO2), with a December 2025 NextEra partnership on a ~1.2-1.5 GW plant. Pre-contract and not yet material, but strategic optionality that grows if grid constraints (the very thing driving NY's moratorium) push data centers toward dedicated generation.
Revenue $41.46B, +346% YoY (beat $35.82B consensus). GAAP net income $28.2B; EPS $24.67. Non-GAAP gross margin 84.9%, a record, up from 39.0%. Analyst caveat: ~85% gross margin is historically extraordinary for a memory maker (prior peaks ~45-60%); this is cycle-peak territory, and memory is deeply cyclical, so peak margins carry peak-cycle downside risk.
Pricing and mix in an AI-memory super-cycle. Segment gross margins: Cloud Memory 83%, Core Data Center 87%, Mobile & Client 87%, Auto & Embedded 79%. HBM4 (High Bandwidth Memory) is shipping >$1B and ramping ~2x the pace of HBM3E.
HBM3E and HBM4 are fully booked through calendar 2027, with demand extending into 2028 and HBM4E already oversubscribed. Management sees the HBM total addressable market 'easily crossing $100B in 2027.' Q4 FY2026 guide: revenue $50.0B ±$1.0B, gross margin ~86%.
Decisively yes: data-center revenue ~$25.3B (~61% of total), enterprise SSD ~$5B, and demand 'far above supply.' Q3 net capex $7.1B; adjusted free cash flow $18.3B.
The ~61% data-center concentration makes Micron directly geared to the buildout, so a broad slowdown would eventually soften HBM pricing. Near-term insulation is real: HBM is contractually sold out through 2027. Note the NY tie-in: Micron's Clay, NY megafab (up to four DRAM fabs, ground broken January 2026) is semiconductor manufacturing, not data-center load, so the ≥50 MW data-center moratorium does not directly restrict it (analysis, UNCONFIRMED). The real exposure is demand-side, post-2027.
Revenue $177.8B, +7.3% YoY. Operating income $7.5B (+5.0%, pressured ~250 bps by fuel/fulfillment). Net income $5.33B; EPS $0.67. Gross margin 24.3% - thin retail economics (operating income ~4.2% of revenue). Global e-commerce +26%.
Scale plus grocery-led traffic (Walmart US comps +4.1%) funding a shift to higher-margin income: advertising +37% (Walmart Connect), marketplace ads +50%, and membership fees +17.4% - all growing far faster than the ~7% top line.
AI angle: Walmart is deploying AI across the supply chain, associate tools, and a shopping agent ('Sparky') credited with digital growth. No AI revenue is broken out; it functions as a margin and productivity lever, not yet financially material (UNCONFIRMED).
Revenue $10.25B, +38% YoY. GAAP gross margin 53%, operating income $1.48B (+83%), net income $1.38B (+95%), EPS $0.84 (non-GAAP $1.37). Data Center revenue $5.78B (+57%), segment operating income $1.60B (~27.7% margin).
Data Center is now the engine (56% of revenue, ~64% of segment operating income), driven by EPYC server CPUs and ramping Instinct GPUs. CEO Lisa Su: 'Data Center now the primary driver of our revenue and earnings growth.'
Instinct MI355X posted leadership MLPerf results; the MI400 series launched (MI455X: 432 GB HBM4, up to 40 PFLOPS FP4), with the Helios rack shipping in 2026. ROCm 7.x is the open answer to NVIDIA's CUDA but still narrower. Named commitments: Meta up to 6 GW, OpenAI 6 GW, Oracle ~50,000 MI450.
Yes and accelerating: Q2 2026 guide ~$11.2B (~46% YoY). Su guided a path to 'tens of billions of dollars in annual Data Center AI revenue in 2027.' NVIDIA still holds the software-ecosystem lead.
Direct NY exposure is modest, but Instinct/EPYC demand rides on a few hyperscalers (Meta, OpenAI, Oracle, Microsoft), so multi-GW slips at any one materially move Data Center revenue. The real risk is a spreading permit-pause slowing the power-and-siting pipeline that gates GW-scale deployments (NY-specific exposure UNCONFIRMED).
Putting the 15 together, the picture is coherent and a little unnerving. The profits are genuine and, right now, growing fast: cloud, advertising, and memory are throwing off record operating income, and every builder is spending into the demand rather than pulling back. If you own FSKAX, that engine has been carrying your returns.
The fragility is that the engine has very few moving parts. The same six or so hyperscalers are simultaneously the customers of NVIDIA, Broadcom, AMD and Micron and the builders whose capex is the demand. If AI spending re-concentrates in the strongest one or two players, the arms suppliers with customer concentration (NVIDIA, Broadcom, AMD, Micron) feel it first and hardest, while the surviving hyperscalers may actually strengthen. That is why "re-concentration" is not simply bad for FSKAX; it reshuffles which of these 15 win, and FSKAX owns them in cap-weighted proportion, so the mega-caps that win would still dominate your result.
The data-center moratorium is the cleanest early-warning variable to watch. Today it is one state and no megacap discloses a New York project, so it is not in the numbers. But it is the first physical constraint - power and permits - on a buildout that has so far been limited only by chips and money. If the pause spreads, the transmission runs demand-side: slower campus energization, then softer chip and memory orders, then the cyclical names (Micron especially, once its 2027 HBM bookings roll off) feeling it before the advertising and software engines do. The counter-position is ExxonMobil, which profits precisely because that power constraint is real.
Bottom line: FSKAX gives you the AI winners and the AI risk in one wrapper, weighted toward the biggest builders. The buildout is still accelerating, so the thesis is intact today; the things that would break it are customer re-concentration and a spreading power/permit constraint, both of which are now observable rather than hypothetical. This is the same "what could reverse it" question from the small-cap analysis, viewed from the inside of the megacaps themselves.