Report Board
Big Four AI CapEx receiver chain
Break the Big Four hyperscaler capex wave into company share, spend pools, and first-order revenue receivers.
A Reviewable Logic Chain
Each card stays open and maps one transmission node without collapsible controls or pseudo-precise scores.
$200B
The Big Four 2026E capex midpoint is about $710B; Amazon, Microsoft and Alphabet together are close to 81%.
$190B
AI compute is the largest pool, but physical data center and power/cooling capacity are the hardest second layer.
$185B
The higher the expected revenue share, the larger the equity elasticity, but the more it needs order and capacity proof.
$135B
The Big Four 2026E capex midpoint is about $710B; Amazon, Microsoft and Alphabet together are close to 81%.
Scope first
This report uses the 2026-04-29 earnings cycle and later capex summaries as the base: Amazon around $200B, Microsoft around $190B, Alphabet/Google at a $185B midpoint, and Meta at the $135B midpoint of its $125B-$145B range.
The $710B midpoint is not all GPUs. The working split is roughly $360B for AI compute systems, $200B for physical data center infrastructure, $60B for networking/storage, $50B for leases/land/other infrastructure, and $40B for Amazon non-AWS logistics/satellite spend.
Why receiver chains matter
CapEx becomes supplier revenue. First-order receivers are GPU/ASIC/HBM/advanced packaging/server systems; second-order receivers are networking, optics and connectivity; third-order receivers are power, cooling, construction, REITs and PPAs.
The useful question is how large that revenue can be relative to each supplier's future revenue base.
Priority order
The first priority remains NVDA, AVGO, TSM, SK Hynix and MU because they sit closest to compute, HBM and packaging constraints.
The second priority is the physical infrastructure chain represented by VRT, ETN, PWR, EME and FIX, where data center capacity constraints are turning into order visibility.
The third priority is networking and optical interconnect exposure such as ANET, COHR, LITE and ALAB. Elasticity is high, but the proof must come from customer concentration, capacity and order language.
Use boundary
Each row is a revenue-opportunity estimate, not company guidance and not a price target. The future revenue share is a rough elasticity lens.
AVGO and MRVL appear in multiple rows because different revenue layers are being separated; those rows should not be added together.
Direction / spend pool / ticker / expected revenue
Future revenue share = estimated annual revenue opportunity from Big Four AI/data center CapEx divided by a rough future annual revenue base for the supplier. Supply-chain layers overlap, so rows such as AVGO and MRVL should not be double counted.
| Direction | Spend pool | Ticker | Expected revenue | Future revenue share | Note |
|---|---|---|---|---|---|
| GPU / AI accelerators | $180B-$260B | NVDA | $120B-$180B | 30%-45% | GB200/GB300, CUDA and system supply; share uses a rough future revenue base near $400B. |
| GPU / AI accelerators | $180B-$260B | AMD | $8B-$20B | 7%-16% | MI300/MI350/MI450 share-gain case; rough future revenue base near $125B. |
| AI server systems | $60B-$100B | DELL | $25B-$45B | 18%-32% | Racks, servers and storage integration; rough future revenue base near $140B. |
| AI server systems | $60B-$100B | SMCI | $8B-$18B | 20%-45% | High-density AI servers and liquid cooling; rough future revenue base near $40B. |
| AI server systems | $60B-$100B | HPE | $2B-$6B | 5%-14% | HPC/Cray and enterprise AI clusters; rough future revenue base near $43B. |
| Custom ASIC / TPU / MTIA | $60B-$100B | AVGO | $35B-$55B | 35%-55% | Google TPU, custom ASIC and packaging pull-through; rough future revenue base near $100B. |
| Custom ASIC / high-speed interconnect | $60B-$100B | MRVL | $5B-$12B | 24%-57% | Custom ASIC, DSP and DPU/interconnect read-through; rough future revenue base near $21B. |
| HBM / DRAM / NAND | $170B-$220B | SK Hynix | $55B-$80B | 45%-65% | HBM leadership and DRAM upcycle; rough future revenue base near $123B. |
| HBM / DRAM / NAND | $170B-$220B | Samsung Electronics | $45B-$70B | 10%-16% | HBM, DRAM, NAND and foundry mix; denominator uses total company revenue near $430B. |
| HBM / DRAM / NAND | $170B-$220B | MU | $25B-$40B | 36%-58% | HBM3E, DRAM and NAND AI server attach; rough future revenue base near $69B. |
| Storage / HDD / SSD | $15B-$30B | WDC | $4B-$8B | 13%-27% | Nearline HDD and enterprise SSD; rough future revenue base near $30B. |
| Storage / HDD | $15B-$30B | STX | $3B-$7B | 13%-29% | Nearline HDD capacity growth; rough future revenue base near $24B. |
| Advanced node / CoWoS | $80B-$130B | TSM | $45B-$75B | 28%-47% | N3/N2, CoWoS and advanced packaging; rough future revenue base near $160B. |
| Semicap equipment | $40B-$70B | ASML | $8B-$15B | 19%-36% | EUV/High-NA and advanced process expansion; rough future revenue base near $42B. |
| Semicap equipment | $40B-$70B | AMAT | $6B-$12B | 18%-35% | Deposition, etch and packaging equipment; rough future revenue base near $34B. |
| Semicap equipment | $40B-$70B | LRCX | $5B-$10B | 22%-43% | Memory and advanced logic equipment; rough future revenue base near $23B. |
| Semicap inspection | $40B-$70B | KLAC | $3B-$7B | 18%-41% | Yield inspection and process control; rough future revenue base near $17B. |
| AI networking switches | $35B-$60B | ANET | $7B-$11B | 52%-81% | Ethernet switching and cloud titans; rough future revenue base near $13.5B. |
| Networking / Ethernet / ASIC | $35B-$60B | AVGO | $10B-$18B | 10%-18% | AVGO networking split; already included in the AVGO ASIC exposure and should not be double counted. |
| Networking / optical interconnect | $35B-$60B | MRVL | $3B-$6B | 14%-29% | MRVL interconnect split; already included in the MRVL ASIC/interconnect exposure. |
| Networking equipment | $35B-$60B | CSCO | $4B-$8B | 6%-12% | Data center switching, routing and enterprise AI networks; rough future revenue base near $67B. |
| Optical modules / components | $20B-$35B | COHR | $3B-$6B | 24%-48% | 800G/1.6T modules and components; rough future revenue base near $12.5B. |
| Optical modules / components | $20B-$35B | LITE | $2B-$5B | 29%-71% | High-speed optical modules and components; rough future revenue base near $7B. |
| PCIe / CXL / connectivity chips | $8B-$15B | ALAB | $1B-$3B | 31%-94% | PCIe retimers, CXL and rack-scale connectivity; rough future revenue base near $3.2B. |
| Power / UPS / liquid cooling | $80B-$130B | VRT | $6B-$10B | 42%-70% | UPS, power, thermal management and liquid cooling; rough future revenue base near $14.2B. |
| Power equipment / distribution | $80B-$130B | ETN | $6B-$11B | 18%-32% | Distribution, switchgear and data center electrical systems; rough future revenue base near $34B. |
| Power equipment / automation | $80B-$130B | Schneider Electric | $8B-$14B | 15%-26% | Electrical systems, controls and energy management; rough future revenue base near $54B. |
| Electrical equipment / automation | $80B-$130B | ABB | $3B-$7B | 8%-19% | Electrical automation, distribution and energy management; rough future revenue base near $37B. |
| Grid / generation equipment | $80B-$130B | GEV | $3B-$8B | 6%-16% | Generation, grid equipment, transformers and capacity upgrades; rough future revenue base near $50B. |
| Data center electrical construction | $60B-$110B | PWR | $4B-$8B | 11%-23% | Transmission, distribution and data center electrical work; rough future revenue base near $35B. |
| Data center MEP construction | $60B-$110B | EME | $3B-$7B | 12%-28% | Mechanical and electrical construction; rough future revenue base near $25B. |
| Data center HVAC / modular | $60B-$110B | FIX | $4B-$8B | 29%-57% | HVAC, mechanical systems and modular builds; rough future revenue base near $14B. |
| Data center construction / communications | $60B-$110B | MTZ | $2B-$5B | 7%-18% | Communications, power and data center construction; rough future revenue base near $28B. |
| Data center REIT / colocation | $20B-$50B | EQIX | $2B-$5B | 17%-42% | Colocation, interconnect and leasing expansion; rough future revenue base near $12B. |
| Data center REIT / colocation | $20B-$50B | DLR | $2B-$5B | 22%-56% | Large-scale data center leasing and interconnect; rough future revenue base near $9B. |
| Power supply / PPA | $20B-$50B | CEG | $2B-$6B | 4%-12% | Nuclear power and long-term PPAs; rough future revenue base near $50B. |
| Power supply / PPA | $20B-$50B | VST | $1B-$4B | 4%-16% | Generation capacity and wholesale power exposure; rough future revenue base near $25B. |
| Power supply / PPA | $20B-$50B | NRG | $1B-$3B | 3%-8% | Retail power and capacity resources; rough future revenue base near $38B. |
Meta / Microsoft / Google / Amazon · 2026E CapEx
Earnings releases, announcements, filings, estimate tables, and reviewable sources.
- Core signal
- 2026E capex guidance, AI server mix, GPU/ASIC/HBM/power construction orders, supplier revenue share
- Current read
- Compute equipment remains first-order; power, cooling and construction are the second major constraint.
- Next question
- Which suppliers receive the capex first, and which names are only second-order valuation read-throughs?
The Big Four 2026E capex midpoint is about $710B; Amazon, Microsoft and Alphabet together are close to 81%.
AI compute is the largest pool, but physical data center and power/cooling capacity are the hardest second layer.
The higher the expected revenue share, the larger the equity elasticity, but the more it needs order and capacity proof.
Whether the four companies keep raising capex guidance or shift more spend into leases
Hyperscaler concentration and delivery cadence at NVDA/AMD/AVGO/MRVL
Whether TSM CoWoS, HBM supply and advanced packaging remain bottlenecks
Whether VRT/ETN/PWR/EME/FIX backlog keeps being pulled by data centers
Customer concentration, margin and order wording at ANET/COHR/LITE/ALAB