MRTNZ

MRTNZ Investment Office · Theme 360

The AI Trade

Own the bottlenecks. Trade the applications selectively.

The MRTNZ callScarcity still earns first.

Compute, networking, power, and cooling remain the cleanest exposures. The next leg depends on software proving that AI usage can become durable margin.

23Liquid AI names
+35.6%Median Street upside
−23.0%Median drawdown
July 28, 2026Market snapshot

The 360 view

The trade is broadening, but the economics are not equal.

Being right about AI adoption is not enough. The investable question is who captures the rent, who funds the build, and which valuations already assume a perfect handoff from infrastructure to applications.

The bottleneck map

Demand is abundant. Throughput is scarce.

AI value must pass through a physical system with long lead times. The narrowest point—not the loudest product—sets the economics of the current trade.

Broad streams of AI application demand converging through chips, networking, power, cooling, and data-center capacity before emerging as concentrated output.
Conceptual flow of AI demand through the physical infrastructure constraint—not a quantitative market-size chart.
Illustrative scenarioBase build

Supply improves, but power and commissioning remain slow-moving constraints.

DemandModels + agents + applications
System throughput41= lowest capacity index
Current limiterTime-to-capacity
Selected constraintTime-to-capacity

The slowest physical dependency sets the commissioning date, even when every component has already been ordered.

Public-market read-throughDELL · HPE · VRT
Opportunity sequenceWhere the rent is captured now—and where it moves next
Now · Scarcity capture

Own the physical choke point

Accelerators, HBM, networking, power, and cooling monetize the build before downstream returns are fully visible.

MU · NVDA · ANET · VRT · ETN · GEV · CEGStay while: lead times, power access, and thermal capacity remain constrained.
Next · Platform proof

Promote scaled distribution

Cloud and data platforms move forward when AI revenue begins outrunning depreciation and capital intensity.

ORCL · MSFT · AMZN · GOOGL · METARotate when: utilization, backlog conversion, and AI revenue growth improve together.
Then · Application proof

Buy evidence, not demos

Applications earn larger weights company by company when agents produce durable growth, retention, or margin.

PLTR · NOW · CRM · CRWDPromote when: monetization survives bundling pressure and shows measurable customer ROI.

Capacity scores are illustrative MRTNZ scenario indices from 0–100. They demonstrate bottleneck mechanics; they are not measured industry utilization, supply forecasts, or security price targets.

Capital stack

Where the money lands

Four layers, four different return profiles. Counts reflect this curated 23-name universe.

018 names

Compute

Silicon, memory, foundry capacity, lithography, and networking remain the first rent collectors.

Cycle, policy, and customer concentration.
025 names

Platforms

Hyperscalers fund the build and own distribution. The question is whether revenue outruns depreciation.

Capital intensity and price competition.
036 names

Infrastructure

Power, cooling, grid equipment, and servers monetize physical scarcity one layer below the model race.

Long-cycle execution and theme premiums.
044 names

Applications

The largest long-run pool, but today’s most selective trade: usage must become durable revenue and margin.

Bundling, weak ROI, and multiple compression.

Opportunity board

Best setup ≠ best company

The score combines 35% direct AI exposure, 25% business quality, 20% Street setup, and 20% price dislocation.

23 names · ranked by MRTNZ opportunity score

Rank / companyRoleScorePriceFrom highStreet upsideNext earnings
01MUMicron Technology, Inc.ComputeHigh-bandwidth memory91.2Fat pitch$821−34.6%+91.3%Not listed 02PLTRPalantir Technologies Inc. Class AApplicationsOperational AI software91.1Fat pitch$124−40.5%+46.7%Aug 3 03ORCLOracle CorporationPlatformsAI cloud capacity89.5Fat pitch$120−65.3%+116.5%Sep 8 04NVDANVIDIA CorporationComputeAccelerators and AI systems88.1Fat pitch$197−16.7%+57.3%Aug 26 05MSFTMicrosoft CorporationPlatformsCloud, models, and copilots87.4Fat pitch$393−29.2%+40.4%Jul 29 06MRVLMarvell Technology, Inc.ComputeConnectivity and custom compute85.5Strong watch$174−47.1%+55.5%Aug 27 07AVGOBroadcom Inc.ComputeCustom silicon and networking85.1Strong watch$381−23.0%+34.6%Sep 3 08TSMTaiwan Semiconductor Manufacturing Company Ltd.ComputeLeading-edge foundry84.5Strong watch$392−18.1%+33.8%Not listed 09ASMLASML Holding N.V. New York Registry SharesComputeAdvanced lithography83.6Strong watch$1,583−20.9%+51.1%Not listed 10NOWServiceNow, Inc.ApplicationsWorkflow agents81.2Strong watch$111−45.0%+26.4%Not listed 11METAMeta Platforms, Inc. Class APlatformsOpen models and recommendation infrastructure80.9Strong watch$593−25.5%+36.7%Jul 29 12VRTVertiv Holdings, LLC Class AInfrastructurePower and cooling79.8Strong watch$270−29.1%+45.2%Not listed 13GOOGLAlphabet Inc. Class APlatformsModels, cloud, and custom accelerators75.9Quality / wait$334−18.3%+27.8%Not listed 14AMDAdvanced Micro Devices, Inc.ComputeAccelerator challenger74.8Quality / wait$455−22.3%+27.8%Aug 4 15ANETArista Networks, Inc.ComputeAI-cluster networking74.7Quality / wait$170−10.6%+13.8%Aug 4 16CRMSalesforce, Inc.ApplicationsEnterprise agents and data73.3Quality / wait$182−33.8%+31.8%Sep 2 17CEGConstellation Energy Corporation Common StockInfrastructureFirm power supply72.4Quality / wait$260−37.0%+35.6%Aug 6 18AMZNAmazon.com, Inc.PlatformsCloud and custom silicon70.8Quality / wait$231−17.1%+37.9%Jul 30 19HPEHewlett Packard Enterprise CompanyInfrastructureEnterprise systems and networking63.6Pricey / prove it$45.59−29.0%+49.2%Sep 2 20GEVGE Vernova Inc.InfrastructureGrid equipment and generation61.8Pricey / prove it$943−21.1%+34.3%Not listed 21DELLDell Technologies Inc. Class C Common StockInfrastructureEnterprise AI servers59.2Pricey / prove it$392−16.5%+20.1%Aug 27 22ETNEaton Corporation, PLC Ordinary SharesInfrastructureElectrical distribution59.2Pricey / prove it$386−11.6%+19.3%Jul 31 23CRWDCrowdStrike Holdings, Inc. Class AApplicationsAI-native security operations57.1Pricey / prove it$182−16.4%+4.9%Aug 26

Top five now

The current fat pitches

Rankings are a research queue, not an instruction to buy. Each needs price discipline and an earnings check.

01 · Compute

MU

91.2

High-bandwidth memory. Latest annual revenue growth: +48.9%.

Price
$821
Street upside
+91.3%
Next check
Every session
Invalidation: Memory remains cyclical even when AI demand is structurally strong.
02 · Applications

PLTR

91.1

Operational AI software. Latest annual revenue growth: +56.2%.

Price
$124
Street upside
+46.7%
Next check
August 3, 2026
Invalidation: Exceptional expectations make multiple compression the central risk.
03 · Platforms

ORCL

89.5

AI cloud capacity. Latest annual revenue growth: +17.3%.

Price
$120
Street upside
+116.5%
Next check
September 8, 2026
Invalidation: Financing intensity, customer concentration, and execution at scale.
04 · Compute

NVDA

88.1

Accelerators and AI systems. Latest annual revenue growth: +65.5%.

Price
$197
Street upside
+57.3%
Next check
August 26, 2026
Invalidation: Capex concentration, export controls, and rapid product-cycle expectations.
05 · Platforms

MSFT

87.4

Cloud, models, and copilots. Latest annual revenue growth: +14.9%.

Price
$393
Street upside
+40.4%
Next check
July 29, 2026
Invalidation: Returns on capital may lag the pace of infrastructure spending.

Rotation playbook

What changes the leadership

The key transition is from scarcity economics to measurable enterprise returns.

  1. NowOwn scarcity

    Accelerators, memory, networking, power, and cooling capture the immediate buildout.

  2. NextDemand platform proof

    Cloud and model revenue must widen faster than depreciation and capital intensity.

  3. ThenPromote applications

    Software earns a larger weight only when agents lift growth, retention, or margin.

  4. RiskPrepare for commoditization

    Lower compute prices help adopters but can compress the first wave of infrastructure winners.

12–24 month framework

Three ways the handoff unfolds

These are MRTNZ scenarios, not Street forecasts or price targets.

Bull

Diffusion wins

Infrastructure demand stays constrained while platforms and applications convert usage into expanding revenue and cash flow.

Base

Narrow leadership

Compute and power keep earning; hyperscaler returns improve slowly; application winners remain company-specific.

Bear

Capex air pocket

Utilization disappoints, accelerator pricing falls, and the market reprices suppliers before end-demand catches up.

Method

Facts first. Judgment labeled.

Market facts come from the current Nasdaq snapshot and reported annual financials.

Street setup uses published target and rating aggregation, not MRTNZ valuation.

AI exposure and opportunity scores are MRTNZ judgments. They rank the research queue; they do not forecast returns.

Refresh cadence follows the Market 360 data pipeline at open, midday, and close.

Read the AI infrastructure capital-cycle field note →