Broadcom guides AI revenue to reach $230 billion by 2028

Broadcom guides AI revenue to reach $230 billion by 2028: the analysis positions AVGO, TSM and MU to benefit and NVDA, AMD and MRVL under pressure; confidence high.

Analyzed 2026-09-16 — point-in-time, never revised. Outcomes are scored against SPY and published either way.

Ranked exposures

  1. ▲ AVGO Broadcom — impact +96 · confidence high · horizon immediate · 1st order
    Broadcom expands high-margin semiconductor revenue as hyperscalers scale custom AI accelerators and Ethernet networking deployments.
    Chain: Hyperscalers award multi-year custom accelerator programs → Broadcom books larger XPU and networking backlog → AI semiconductor revenue guided toward 230 billion → Segment mix shifts to high-margin silicon
    What would break the call: A lead hyperscaler cancels or delays its XPU program, cutting the disclosed AI backlog.
  2. ▲ TSM TSMC — impact +78 · confidence high · horizon quarters · 2nd order
    TSMC fills leading-edge wafer and advanced-packaging capacity as hyperscalers scale Broadcom custom AI accelerator deployments.
    Chain: Broadcom accelerator orders rise sharply → Wafer starts booked at leading-edge nodes → CoWoS advanced packaging capacity fills → TSMC revenue and utilization increase
    What would break the call: Broadcom shifts meaningful accelerator volume to Intel Foundry or Samsung, reducing TSMC wafer share.
  3. ▲ MU Micron — impact +72 · confidence high · horizon quarters · 2nd order
    Micron captures stronger high-bandwidth-memory demand as custom AI accelerator deployments intensify data requirements.
    Chain: Custom accelerators require stacked memory bandwidth → HBM content per accelerator rises → Micron HBM capacity sells out forward → Memory pricing and margins expand
    What would break the call: Samsung qualifies HBM4 broadly and floods supply, collapsing memory contract pricing.
  4. ▲ ALAB Astera Labs — impact +70 · confidence low · horizon weeks · 2nd order
    Astera Labs sells connectivity chips into custom accelerator racks, raising unit shipments alongside Broadcom silicon deployments.
    Chain: Accelerator racks scale in node count → PCIe and Ethernet signal integrity degrades → Retimers and fabric switches added per rack → Astera Labs unit shipments rise
    What would break the call: Broadcom integrates retiming and fabric switching into its own silicon, displacing Astera content.
  5. ▲ CRDO Credo — impact +68 · confidence low · horizon weeks · 2nd order
    Credo supplies active electrical cables and retimers for the dense back end networks these accelerator clusters require.
    Chain: Back end networks densify around accelerators → Copper links need retiming at 800G → Active electrical cables replace passive copper → Credo product revenue accelerates
    What would break the call: A single large hyperscaler customer insources AEC design, removing most of Credo revenue concentration.
  6. ▲ CLS Celestica — impact +66 · confidence low · horizon quarters · 2nd order
    Celestica builds switches and accelerator racks for hyperscalers, expanding connectivity segment volume and operating margin.
    Chain: Hyperscalers order more Ethernet switches → Custom accelerator racks need integration → Celestica connectivity segment volume rises → Operating margin expands on scale
    What would break the call: A top hyperscaler moves rack integration to a rival ODM, cutting Celestica program volume.
  7. ▼ NVDA Nvidia — impact -65 · confidence medium · horizon weeks · 1st order
    Nvidia loses accelerator demand as hyperscalers redirect budgets toward cheaper Broadcom custom silicon.
    Chain: Hyperscaler AI budgets are finite → Custom silicon absorbs a larger share → Merchant GPU order growth slows → Nvidia pricing power and share erode
    What would break the call: Total AI capital spending expands fast enough that Nvidia data center revenue keeps compounding regardless.
  8. ▲ ANET Arista Networks — impact +61 · confidence low · horizon weeks · 2nd order
    Arista Networks sells more high speed Ethernet systems as Broadcom based AI clusters require expanding scale out connectivity.
    Chain: Custom clusters use Ethernet fabrics → Broadcom switch silicon powers Arista systems → Port counts scale with accelerator nodes → Arista data center revenue grows
    What would break the call: Hyperscalers adopt white box switches at scale, displacing Arista branded systems.
  9. ▲ VRT Vertiv — impact +60 · confidence medium · horizon quarters · 3rd order
    Vertiv expands power and liquid-cooling orders as hyperscalers build additional facilities for Broadcom accelerator clusters.
    Chain: Accelerator clusters raise rack power density → New data center capacity is built → Power distribution and liquid cooling ordered → Vertiv backlog and revenue increase
    What would break the call: Hyperscalers pause data center construction on power permitting delays, freezing Vertiv order intake.
  10. ▲ AMKR Amkor — impact +60 · confidence low · horizon quarters · 2nd order
    Amkor gains outsourced advanced packaging and test volume as custom accelerator units multiply beyond internal foundry capacity.
    Chain: Advanced packaging demand exceeds foundry capacity → Overflow assembly outsourced to OSAT vendors → Amkor test and packaging volume rises → Advanced packaging revenue mix improves
    What would break the call: TSMC expands CoWoS capacity enough to retain all packaging in house.
  11. ▲ COHR Coherent — impact +58 · confidence medium · horizon quarters · 2nd order
    Coherent grows datacom revenue as AI cluster expansion increases optical interconnect bandwidth requirements.
    Chain: Cluster scale-out increases optical link count → 800G and 1.6T transceiver demand rises → Coherent datacom shipments and lasers grow → Networking segment revenue expands
    What would break the call: Co-packaged optics adoption bypasses pluggable transceivers, shrinking Coherent datacom volume.
  12. ▲ FN Fabrinet — impact +58 · confidence low · horizon quarters · 3rd order
    Fabrinet assembles those transceivers under contract, so its optical manufacturing volume and revenue rise with accelerator deployments.
    Chain: Optical transceiver demand rises with clusters → Module vendors outsource contract assembly → Fabrinet optical manufacturing volume increases → Revenue and utilization improve
    What would break the call: A key module customer shifts assembly in house or to a rival contract manufacturer.
  13. ▲ MSFT Microsoft — impact +53 · confidence low · horizon quarters · 3rd order
    Microsoft realizes higher subscription revenue after corporations deploy more artificial intelligence tools.
    Chain: Cheaper custom accelerators lower inference cost → Azure AI capacity scales more economically → Enterprises deploy more Copilot seats → Microsoft subscription revenue rises
    What would break the call: Enterprise AI seat adoption stalls, leaving Copilot revenue flat despite cheaper compute.
  14. ▲ ASML ASML — impact +52 · confidence low · horizon quarters · 3rd order
    Advanced chip manufacturing investments lift ASML revenue from extreme ultraviolet lithography systems.
    Chain: Foundries add leading-edge capacity for accelerators → Capital budgets shift to EUV tools → ASML bookings and backlog rise → System revenue grows with shipments
    What would break the call: Foundry capital spending is cut on memory oversupply, pushing EUV bookings lower.
  15. ▼ AMD AMD — impact -51 · confidence medium · horizon weeks · 1st order
    AMD loses data-center accelerator share as hyperscalers commit more volume to Broadcom custom designs.
    Chain: Hyperscalers allocate budgets to custom silicon → Merchant accelerator tenders shrink → AMD Instinct volume growth slows → Data center segment revenue disappoints
    What would break the call: A large sovereign or neocloud order lifts Instinct volume independently of hyperscaler budgets.
  16. ▼ MRVL Marvell — impact -51 · confidence medium · horizon weeks · 1st order
    Marvell loses custom accelerator design wins as hyperscale customers award larger ASIC programs to Broadcom.
    Chain: Custom ASIC programs consolidate with Broadcom → Marvell loses next generation design sockets → Data center custom silicon revenue slips → Guidance and multiple compress
    What would break the call: Marvell announces a major new custom accelerator win with a top hyperscaler.
  17. ▲ AMAT Applied Materials — impact +49 · confidence low · horizon quarters · 3rd order
    Applied Materials gains semiconductor equipment demand from foundry capacity additions supporting advanced logic and packaging production.
    Chain: Foundry adds logic and packaging capacity → Deposition and etch tool orders rise → Applied Materials system revenue grows → Advanced packaging segment expands
    What would break the call: China export restrictions tighten further, cutting a large slice of Applied Materials system sales.
  18. ▲ AMZN Amazon — impact +48 · confidence low · horizon quarters · 3rd order
    Enterprises ramp up AI workloads and therefore Amazon grows its AWS cloud revenue.
    Chain: Custom silicon lowers cost per inference → AWS prices AI capacity aggressively → Enterprise AI workloads migrate to AWS → AWS revenue growth accelerates
    What would break the call: AWS AI capacity remains supply constrained, capping revenue growth despite cheaper silicon.
  19. ▲ LITE Lumentum — impact +48 · confidence low · horizon quarters · 2nd order
    Cluster scale-out expands 800G transceiver and laser demand, raising Lumentum cloud and networking revenue.
    Chain: Accelerator clusters need more optical links → 800G and 1.6T module demand rises → Lumentum ships lasers and transceivers → Cloud and networking revenue grows
    What would break the call: Co-packaged optics or silicon photonics from Broadcom bypasses pluggable modules at scale.
  20. ▲ ARM Arm — impact +47 · confidence low · horizon quarters · 3rd order
    License agreements for AI chip architectures increase Arm royalty income over time.
    Chain: Custom accelerators pair with Arm CPU cores → Data center Arm licensing expands → Royalty per chip rises with content → Arm royalty revenue grows
    What would break the call: Accelerator designs adopt RISC-V control cores instead, removing Arm royalty content.
  21. ▼ IBM IBM — impact -45 · confidence low · horizon quarters · 3rd order
    IBM experiences slower legacy hardware revenue growth as AI shifts spending to cloud platforms.
    Chain: Enterprise budgets shift toward AI cloud → On-premise hardware refresh cycles stretch → IBM infrastructure revenue growth slows → Mix pressure weighs on segment margin
    What would break the call: A strong new mainframe cycle lifts IBM infrastructure revenue despite AI spending shifts.
  22. ▲ KLAC KLA — impact +45 · confidence low · horizon quarters
    Process control equipment demand rises for KLA since AI chips require precise fabrication.
  23. ▲ MPWR Monolithic Power Systems — impact +45 · confidence low · horizon quarters · 2nd order
    Higher accelerator rack density raises power management content per board, lifting Monolithic Power data center revenue.
    Chain: Accelerator racks draw more current → Vertical power delivery modules added per board → Monolithic Power content per socket rises → Enterprise data center revenue grows
    What would break the call: Broadcom customers standardize on Infineon or Renesas power stages, displacing Monolithic Power content.
  24. ▲ ETN Eaton — impact +42 · confidence low · horizon quarters · 3rd order
    Accelerator clusters raise facility power demand, expanding Eaton electrical backlog for data center switchgear and busway.
    Chain: Cluster deployments need new power capacity → Data center electrical buildout accelerates → Eaton switchgear and busway orders rise → Electrical Americas backlog and margin grow
    What would break the call: Utility interconnection queues stall data center starts, flattening Eaton electrical order growth.

Sources

What happened

HorizonVerdictTop pickExcess vs SPY
1 trading daysRightAVGO+0.5% vs SPY
5 trading daysRightAVGO+5.3% vs SPY

More

Related scenarios

Stocks in this scenario

AVGO · TSM · MU · ALAB · CRDO · CLS · NVDA · ANET · VRT · AMKR · COHR · FN · MSFT · ASML · AMD · MRVL · AMAT · AMZN · LITE · ARM · IBM · KLAC · MPWR · ETN · QCOM · ONTO · CSCO · INTC · TXN · SMCI · CRWV · AAPL · DELL · ORCL · HPQ

News events · All scenarios · Pricing · Track record · Methodology · Map your own scenario