Market map

The AI value chain

Every layer of the AI stack — from agents at the apex to lithography at the base — rated on moat, growth, margin, cost to run, and security exposure. Read it as a ledger, or drill into it as a map.

31 components · 101 sub-components · 8 tiers + 3 cross-cutting forces
View
Lens
Moat / substitutabilitychokepointstrongcontestedcommodityHow hard is this to replace?
Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Brand + distribution moat; switching cost is one login away.
Forensic read

The apex layer captures the user, not always the margin. Prices compress as base-model capability converges. Winners defend with distribution, memory, and workflow, not model access, which everyone rents.

4 sub-components
Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Consumer default with the widest funnel; defensibility is habit, not tech.
Forensic read

Category-defining reach, but every rival ships comparable capability within months.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Enterprise and developer lean; safety brand as differentiator.
Forensic read

Strong in code and regulated settings where reliability outweighs raw novelty.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Bundled into Workspace, Android, and Search; distribution is the moat.
Forensic read

The one horizontal assistant with a captive billion-user surface to push into.

Moat
contested
Growth
hyper
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
Answer engine adjacent to search; ad and citation model unproven.
Forensic read

Sharp product, structurally exposed to Google and to model-vendor encroachment.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Sticky via repo context and habit; model-swappable underneath.
Forensic read

Fastest-monetizing AI category. Real moat is codebase context and developer habit, but every entrant rides the same foundation models, a thin layer over someone else’s compute.

4 sub-components
Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
IDE fork with deep repo context; growth is extraordinary, moat is habit.
Forensic read

Proof the application layer can move fastest, and that it rents its intelligence.

Moat
strong
Growth
high
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Incumbent distribution through GitHub; default for enterprises.
Forensic read

Doesn’t need to be best; it ships where developers already are.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Agentic IDE competing on autonomy; crowded, undifferentiated underneath.
Forensic read

A feature race where the model does the work and the wrapper takes the margin.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Terminal-native agents; power-user wedge into workflows.
Forensic read

Distribution shifts to whoever the developer already trusts for the model.

Moat
strong
Growth
high
Margin
fat
Cost to run
moderate
Security
critical
Substitutability verdict
Workflow + proprietary data + compliance lock-in. Hard to rip out.
Forensic read

Where durable application value concentrates: regulated data, embedded workflow, and switching pain buy pricing power the horizontal assistants lack.

4 sub-components
Moat
strong
Growth
high
Margin
fat
Cost to run
moderate
Security
critical
Substitutability verdict
Legal workflow + privileged data; switching cost is high.
Forensic read

The template for defensible vertical AI: regulated domain, deep integration.

Moat
strong
Growth
high
Margin
fat
Cost to run
moderate
Security
critical
Substitutability verdict
Embedded in EHR and clinician workflow; compliance moat.
Forensic read

Health-system integration and reimbursement alignment make it sticky.

Moat
strong
Growth
high
Margin
fat
Cost to run
moderate
Security
high
Substitutability verdict
Outcome-based pricing on customer-service automation.
Forensic read

Charging for resolved outcomes, not tokens, is the defensible business model.

Moat
strong
Growth
high
Margin
fat
Cost to run
moderate
Security
high
Substitutability verdict
Hundreds of unbuilt niches with the same defensibility recipe.
Forensic read

The layer where fat margin hides, if the domain has data and switching pain.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Distribution is the moat; AI bolted onto an installed base.
Forensic read

Incumbents don’t need the best model; they own the seat. AI is a feature that raises contract value on software customers already locked in.

3 sub-components
Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Per-seat upsell across Office’s installed base.
Forensic read

The purest distribution play in AI: monetize AI on seats you already own.

Moat
strong
Growth
high
Margin
fat
Cost to run
moderate
Security
high
Substitutability verdict
Agents on top of CRM data and workflow.
Forensic read

Data gravity and workflow lock-in convert to AI pricing power.

Moat
strong
Growth
high
Margin
fat
Cost to run
moderate
Security
high
Substitutability verdict
Workflow platform embedding AI across IT and ops.
Forensic read

Owns the system of action, which is where agents get deployed.

Moat
strong
Growth
hyper
Margin
thin
Cost to run
extreme
Security
critical
Substitutability verdict
Differentiated only at the frontier; the lead is measured in months.
Forensic read

Capex-heavy race with thin per-token economics. Moat exists at the capability frontier and erodes the moment open weights close the gap. Value migrates to whoever monetizes distribution on top.

4 sub-components
Moat
strong
Growth
hyper
Margin
thin
Cost to run
extreme
Security
critical
Substitutability verdict
Capability plus consumer distribution; enormous burn.
Forensic read

Ahead on reach, structurally dependent on Azure compute and continuous capex.

Moat
strong
Growth
hyper
Margin
thin
Cost to run
extreme
Security
critical
Substitutability verdict
Frontier capability with enterprise and safety positioning.
Forensic read

Competes at the frontier while leaning into reliability-sensitive buyers.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
critical
Substitutability verdict
Only fully vertically integrated stack: TPU, data, distribution.
Forensic read

Structurally advantaged: owns compute, data, and the surfaces to ship into.

Moat
contested
Growth
hyper
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
Fast follower with aggressive compute buildout.
Forensic read

Capital and talent can buy proximity to the frontier, not a durable moat.

Moat
commodity
Growth
hyper
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
Commoditization is the strategy; free weights collapse the price floor.
Forensic read

The single most important substitution force in the stack. Open weights turn "the model" into infrastructure, pushing durable value up to apps and down to compute. Watch this line to price every layer above it.

4 sub-components
Moat
commodity
Growth
hyper
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
Strategic commoditizer; Meta gives weights away to deny rivals a moat.
Forensic read

A deliberate scorched-earth play that resets the entire base-model economics.

Moat
commodity
Growth
hyper
Margin
thin
Cost to run
extreme
Security
moderate
Substitutability verdict
European open-weight champion; sovereignty angle.
Forensic read

Credible open models with a regulatory tailwind, thin economics.

Moat
commodity
Growth
hyper
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
Cost-efficiency shock; showed frontier-adjacent results far cheaper.
Forensic read

The clearest signal that training-cost moats are softer than assumed.

Moat
commodity
Growth
hyper
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
Strong open multilingual family with wide adoption.
Forensic read

Global open-weight competition keeps the price floor falling.

Moat
contested
Growth
high
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Cost and latency edge, but frontier models keep absorbing the niche.
Forensic read

Economically rational for narrow tasks. Fragile long-term: general models get cheaper faster than specialized ones can defend a niche.

3 sub-components
Moat
contested
Growth
high
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Cheap, fast, edge-deployable; capability rising quickly.
Forensic read

Good-enough small models erode the case for paying for frontier calls.

Moat
contested
Growth
high
Margin
moderate
Cost to run
moderate
Security
high
Substitutability verdict
Vertical fine-tunes on proprietary data.
Forensic read

Defensible only where the data is proprietary and the domain is narrow.

Moat
contested
Growth
high
Margin
moderate
Cost to run
heavy
Security
low
Substitutability verdict
Privacy and latency via local NPUs.
Forensic read

Shifts some inference off the cloud, capping a slice of datacenter demand.

Moat
commodity
Growth
high
Margin
thin
Cost to run
light
Security
moderate
Substitutability verdict
Open-source with near-zero switching cost. Easily replaced or rebuilt.
Forensic read

Glue code. Real but hard to monetize; model vendors keep absorbing orchestration into their own APIs, hollowing the layer.

3 sub-components
Moat
commodity
Growth
high
Margin
thin
Cost to run
light
Security
moderate
Substitutability verdict
Popular glue; being commoditized from above and below.
Forensic read

Ubiquitous and hard to charge for as agent APIs go native.

Moat
commodity
Growth
high
Margin
thin
Cost to run
light
Security
moderate
Substitutability verdict
Retrieval-centric orchestration.
Forensic read

Useful primitive, thin as a standalone business.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
light
Security
moderate
Substitutability verdict
OpenAI/Anthropic absorbing orchestration into the model layer.
Forensic read

The vendors are eating this layer; that is the whole risk to it.

Moat
contested
Growth
high
Margin
moderate
Cost to run
moderate
Security
moderate
Substitutability verdict
pgvector inside Postgres commoditizes the standalone pure-plays.
Forensic read

Retrieval matters; the standalone database is contestable. When a checkbox in Postgres does most of the job, the specialist has to win on scale and ops, not existence.

3 sub-components
Moat
contested
Growth
high
Margin
moderate
Cost to run
moderate
Security
moderate
Substitutability verdict
Managed vector DB; premium on ops and scale.
Forensic read

Has to out-operate a free Postgres extension to justify itself.

Moat
commodity
Growth
high
Margin
thin
Cost to run
light
Security
moderate
Substitutability verdict
Open-source vector search.
Forensic read

Capable and free, which is exactly the pricing problem.

Moat
commodity
Growth
high
Margin
thin
Cost to run
light
Security
moderate
Substitutability verdict
A checkbox in the database everyone already runs.
Forensic read

The commoditizer: good-enough retrieval with zero new vendor.

Moat
contested
Growth
high
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
Price war on tokens-per-dollar; open-source serving stacks underneath.
Forensic read

A margin-thin arbitrage on GPU rental. Differentiation is speed and cost, both of which erode as open engines and hardware improve.

3 sub-components
Moat
contested
Growth
high
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
Hosted open-model inference at aggressive prices.
Forensic read

Competing on price for a service the open stack can self-host.

Moat
contested
Growth
high
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
Model deployment and hosting.
Forensic read

Convenience layer over commoditizing compute.

Moat
commodity
Growth
high
Margin
thin
Cost to run
light
Security
moderate
Substitutability verdict
Open serving engines that set the performance bar.
Forensic read

Free, fast, and the reason hosted-inference margins stay thin.

Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
moderate
Substitutability verdict
Becomes sticky once wired into CI; still early and fragmented.
Forensic read

Underbuilt and structurally necessary as agents ship to production. The layer most likely to consolidate into a defensible category; worth watching for entrants.

3 sub-components
Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
moderate
Substitutability verdict
Eval plus tracing wired into dev pipelines.
Forensic read

Gets sticky once it sits inside CI; early leader position is open.

Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
moderate
Substitutability verdict
ML observability extending into LLMs.
Forensic read

Production monitoring is a real need as agents fail in subtle ways.

Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
high
Substitutability verdict
Input/output policy enforcement.
Forensic read

Necessary for regulated deployment; likely bundled over time.

Moat
strong
Growth
high
Margin
fat
Cost to run
extreme
Security
high
Substitutability verdict
Capital scale + enterprise lock-in. Not easily replaced at the top end.
Forensic read

The landlords. They set compute price, vertically integrate into silicon to escape NVIDIA, and capture recurring enterprise spend. Structurally advantaged, but their own custom chips are the tell on where GPU pricing power is headed.

4 sub-components
Moat
strong
Growth
high
Margin
fat
Cost to run
extreme
Security
high
Substitutability verdict
Scale leader plus in-house Trainium/Inferentia.
Forensic read

Building its own silicon is the clearest signal on NVIDIA dependence.

Moat
strong
Growth
high
Margin
fat
Cost to run
extreme
Security
high
Substitutability verdict
OpenAI-aligned; enterprise distribution engine.
Forensic read

Owns the highest-profile model tenant and the enterprise relationships.

Moat
strong
Growth
high
Margin
fat
Cost to run
extreme
Security
high
Substitutability verdict
Most vertically integrated via TPU and DeepMind.
Forensic read

The only hyperscaler that controls its own accelerator at scale.

Moat
contested
Growth
hyper
Margin
fat
Cost to run
extreme
Security
high
Substitutability verdict
Aggressive GPU capacity and large AI contracts.
Forensic read

Late but buying share with capacity commitments and concentrated deals.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
moderate
Substitutability verdict
Proves raw GPU rental is contestable, but it is a capex treadmill.
Forensic read

Explosive revenue, brutal economics: debt-funded GPU fleets that depreciate on NVIDIA’s release cadence. The forensic risk sits in financing structure and customer concentration, not demand.

3 sub-components
Moat
contested
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
moderate
Substitutability verdict
GPU-native cloud; debt-funded fleet, concentrated customers.
Forensic read

The purest neocloud test case: demand is real, the balance sheet is the risk.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
moderate
Substitutability verdict
Full-stack neocloud with vertical ambitions.
Forensic read

Same treadmill: fleet value tracks NVIDIA’s next launch.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
moderate
Substitutability verdict
GPU rental; Crusoe uses stranded energy.
Forensic read

Differentiates on power access, still exposed to depreciation cycles.

Moat
contested
Growth
high
Margin
moderate
Cost to run
extreme
Security
critical
Substitutability verdict
Moat is regulatory, not technical; data residency mandates.
Forensic read

Growth driven by policy rather than product. Durable only where law forces it; commoditizes wherever it doesn’t.

3 sub-components
Moat
contested
Growth
high
Margin
moderate
Cost to run
extreme
Security
critical
Substitutability verdict
Data-residency-mandated regional capacity.
Forensic read

Exists because regulation requires it, not because it competes on merit.

Moat
strong
Growth
high
Margin
moderate
Cost to run
extreme
Security
critical
Substitutability verdict
Security-cleared, air-gapped environments.
Forensic read

A genuinely defensible niche gated by clearance and compliance.

Moat
contested
Growth
high
Margin
moderate
Cost to run
extreme
Security
critical
Substitutability verdict
State-backed AI infrastructure (e.g., Gulf, Asia).
Forensic read

Policy and capital create instant scale with uncertain economics.

Moat
contested
Growth
high
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
Integration and assembly; replaceable, thin margin, GPU-gated.
Forensic read

High revenue, low margin: passing NVIDIA silicon through to buyers. Value is captured up (chips) and down (cloud), squeezing the integrator in the middle.

3 sub-components
Moat
contested
Growth
high
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
Fast reference systems; thin margin, governance scrutiny.
Forensic read

Revenue tracks GPU allocation; it earns an assembly spread, not a moat.

Moat
contested
Growth
high
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
Enterprise integration and support.
Forensic read

Distribution and service, squeezed between silicon and cloud.

Moat
commodity
Growth
high
Margin
thin
Cost to run
heavy
Security
moderate
Substitutability verdict
White-box manufacturing for hyperscalers.
Forensic read

Pure contract build; the definition of a replaceable middle.

Moat
strong
Growth
high
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
At GPU-cluster scale the fabric is hard to swap; real lock-in.
Forensic read

The interconnect is a quiet chokepoint: training clusters are bottlenecked by the network, and NVIDIA owns much of that fabric. Under-appreciated leverage.

3 sub-components
Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Ethernet scaling for AI back-end networks.
Forensic read

Positioned as the open alternative to NVIDIA’s proprietary fabric.

Moat
strong
Growth
hyper
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Bundled high-performance fabric with the GPUs.
Forensic read

Extends the GPU moat into the network the cluster depends on.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Merchant switching silicon for hyperscalers.
Forensic read

The silicon under most Ethernet fabrics, sold to everyone.

Moat
strong
Growth
high
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Scarce components (lasers, transceivers) with real supply constraints.
Forensic read

A physical bottleneck as clusters scale. Optics scarcity is a genuine constraint, not a narrative; supply lead times gate buildout.

3 sub-components
Moat
strong
Growth
high
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Lasers and transceivers in tight supply.
Forensic read

Component scarcity gives real, if cyclical, pricing leverage.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
DSPs that drive high-speed optical links.
Forensic read

High-value silicon at the center of the interconnect scaling problem.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Emerging integration of optics onto the package.
Forensic read

Next-gen bottleneck-breaker; unproven at volume, high option value.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Fast becoming the binding constraint on the entire buildout.
Forensic read

You can’t rack GPUs you can’t power or cool. This has flipped from afterthought to gating factor, one of the most defensible growth lines in the chain.

3 sub-components
Moat
strong
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Thermal and power management for datacenters.
Forensic read

Direct leverage on every megawatt of AI capacity built.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Direct-to-chip cooling for dense GPU racks.
Forensic read

Air cooling fails at Blackwell densities; this becomes mandatory.

Moat
strong
Growth
high
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Grid-to-rack power distribution.
Forensic read

The unglamorous electrical layer now gating datacenter timelines.

Moat
chokepoint
Growth
hyper
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
CUDA lock-in + performance lead. The hardest moat above the fab line.
Forensic read

The apex predator of silicon. CUDA is a decade-deep software moat, not just faster chips. Every layer above pays the NVIDIA tax, and every hyperscaler is spending billions to escape it. That escape effort is the one crack to monitor.

3 sub-components
Moat
chokepoint
Growth
hyper
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Current-generation apex; demand outruns supply.
Forensic read

Near-total pricing power this cycle; the chip everyone queues for.

Moat
chokepoint
Growth
high
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Massive installed base still in heavy demand.
Forensic read

The workhorse; depreciation risk if Blackwell supply catches up.

Moat
chokepoint
Growth
high
Margin
fat
Cost to run
light
Security
high
Substitutability verdict
The actual moat: a decade of tooling and lock-in.
Forensic read

Rivals can match silicon faster than they can match the software estate.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
The deliberate substitution for NVIDIA, funded by its biggest customers.
Forensic read

Where NVIDIA’s moat gets tested. Hyperscaler in-house silicon and merchant challengers won’t dethrone CUDA soon, but they cap pricing power and re-rate the whole compute layer over time.

5 sub-components
Moat
strong
Growth
hyper
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Most mature NVIDIA alternative; internal plus Cloud.
Forensic read

The proof that a credible non-CUDA accelerator can run at scale.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Amazon’s in-house training and inference silicon.
Forensic read

A direct attempt to cut the NVIDIA tax on its own fleet.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Leading merchant challenger; software gap closing.
Forensic read

The only merchant GPU credibly chasing NVIDIA; ROCm is the hurdle.

Moat
strong
Growth
hyper
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Co-designed accelerators for hyperscalers.
Forensic read

Arms the buyers building their way off NVIDIA; a picks-and-shovels win.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
moderate
Substitutability verdict
Inference-specialized architectures.
Forensic read

Interesting performance angles, unproven durability against incumbents.

Moat
contested
Growth
declining
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
INTEL CASE: former x86 chokepoint, now substitutable in the growth vector.
Forensic read

The clearest "valuable or replaced?" verdict in the map. AI training value migrated to GPUs, where Intel is a non-factor. Arm (Graviton, Apple) is eating general compute on efficiency, and AMD took server share. The x86 franchise still prints cash, but it sits in the one vector that is flat-to-declining while every high-growth vector routes around it. Former chokepoint, increasingly substitutable.

4 sub-components
Moat
contested
Growth
declining
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Legacy franchise in the one flat-to-declining vector.
Forensic read

Still cash-generative, but the AI buildout routes around it.

Moat
strong
Growth
high
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Sustained server share gains against Intel.
Forensic read

Executed the share shift x86 leadership failed to defend.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Efficiency-driven disruption of general compute.
Forensic read

Perf-per-watt wins in the cloud where power is the constraint.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
low
Substitutability verdict
Vertically integrated on-device compute.
Forensic read

Owns its stack end to end; irrelevant to datacenter, potent on device.

Moat
chokepoint
Growth
hyper
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Supply-constrained and technically brutal; a hard bottleneck on GPUs.
Forensic read

The under-covered chokepoint. High-bandwidth memory gates GPU output; supply is tight and sold out ahead. A three-player oligopoly with genuine pricing power, arguably a cleaner scarcity story than the GPUs it feeds.

3 sub-components
Moat
chokepoint
Growth
hyper
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
HBM technology and volume leader.
Forensic read

Sets the pace on the memory that gates every top GPU.

Moat
strong
Growth
hyper
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Scaling HBM to close the gap.
Forensic read

Capable challenger; qualification timing is the swing factor.

Moat
strong
Growth
hyper
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
US-based third source ramping HBM.
Forensic read

The domestic supply-chain hedge in a two-player Korean race.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Also the custom-ASIC partner of choice; dual exposure, hard to displace.
Forensic read

Broadcom in particular sits at two chokepoints at once: switching silicon and co-designed accelerators for hyperscalers. Under-narrated leverage on the AI buildout.

2 sub-components
Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Switching silicon plus custom accelerator design.
Forensic read

Double exposure to the buildout; wins whether the fabric is Ethernet or the chip is custom.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Optical DSP and custom compute silicon.
Forensic read

High-value silicon across interconnect and bespoke accelerators.

Moat
chokepoint
Growth
high
Margin
fat
Cost to run
extreme
Security
critical
Substitutability verdict
Near-monopoly on advanced-node manufacturing. Effectively irreplaceable now.
Forensic read

The physical bottleneck under all AI silicon. NVIDIA, Apple, AMD, and the hyperscaler ASICs all fabricate here. The moat is measured in years and hundreds of billions of capex. The tail risk is geopolitical, not competitive.

2 sub-components
Moat
chokepoint
Growth
high
Margin
fat
Cost to run
extreme
Security
critical
Substitutability verdict
Sole reliable source of leading-edge logic.
Forensic read

If any single company is load-bearing for AI, this is it.

Moat
chokepoint
Growth
hyper
Margin
fat
Cost to run
extreme
Security
critical
Substitutability verdict
Packaging capacity that gates GPU output.
Forensic read

The real near-term bottleneck within the foundry; advanced packaging is sold out.

Moat
contested
Growth
high
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
INTEL CASE: aspiring second source to TSMC. Challenger, not yet a chokepoint.
Forensic read

The bull case for Intel is here, not in CPUs: a credible Western leading-edge foundry with policy tailwinds. But it is unproven at volume yield versus TSMC and burning extreme capex to prove it. Optionality with real execution risk, the opposite profile from its legacy franchise.

2 sub-components
Moat
contested
Growth
high
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
Leading-edge process aiming to reach parity.
Forensic read

The whole thesis rides on yield at volume, still unproven.

Moat
contested
Growth
high
Margin
thin
Cost to run
extreme
Security
high
Substitutability verdict
Winning third-party fab clients as validation.
Forensic read

Marquee customers would prove the model; absence would indict it.

Moat
contested
Growth
steady
Margin
moderate
Cost to run
extreme
Security
high
Substitutability verdict
Perennial number-two on leading edge; yield gap keeps it contested.
Forensic read

Structurally capable, persistently behind TSMC on advanced-node yield. Matters as a second source, but hasn’t converted capability into leadership.

1 sub-components
Moat
contested
Growth
steady
Margin
moderate
Cost to run
extreme
Security
high
Substitutability verdict
Capable leading-edge fab trailing on yield.
Forensic read

The standing second source that never quite catches the leader.

Moat
chokepoint
Growth
high
Margin
fat
Cost to run
heavy
Security
critical
Substitutability verdict
Sole supplier of EUV. The single hardest bottleneck in the entire chain.
Forensic read

If any node is truly irreplaceable, it is this one. No leading-edge chip exists without ASML’s EUV machines, and there is no second source. A monopoly at the base of everything, and the ultimate geopolitical control point.

2 sub-components
Moat
chokepoint
Growth
high
Margin
fat
Cost to run
heavy
Security
critical
Substitutability verdict
The only EUV lithography machines in existence.
Forensic read

A literal monopoly; export policy on these machines shapes the whole industry.

Moat
chokepoint
Growth
hyper
Margin
fat
Cost to run
extreme
Security
critical
Substitutability verdict
Next-generation lithography, also sole-source.
Forensic read

Extends the monopoly forward another node generation.

Moat
chokepoint
Growth
high
Margin
fat
Cost to run
light
Security
high
Substitutability verdict
Duopoly no advanced chip can be designed without. Deeply entrenched.
Forensic read

The hidden chokepoint most AI narratives skip. You cannot design a modern chip without this duopoly’s tools, a software toll booth in front of all silicon, with software-grade margins and switching costs measured in years.

2 sub-components
Moat
chokepoint
Growth
high
Margin
fat
Cost to run
light
Security
high
Substitutability verdict
Design tools plus a vast IP library.
Forensic read

Half of an unavoidable duopoly with recurring, high-margin revenue.

Moat
chokepoint
Growth
high
Margin
fat
Cost to run
light
Security
high
Substitutability verdict
Design, simulation, and verification tools.
Forensic read

The other half; every chip company is effectively a customer.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Few suppliers for process-critical tools; concentrated and technical.
Forensic read

The pick-and-shovel layer beneath the foundries. Concentrated, high-margin, and export-controlled, durable exposure to every fab that gets built.

3 sub-components
Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Broadest process-equipment portfolio.
Forensic read

Sells to every fab regardless of which chip wins.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Etch and deposition leadership.
Forensic read

Process-critical tools with few substitutes at the leading edge.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
high
Substitutability verdict
Major Japanese process-equipment supplier.
Forensic read

Concentrated supply base; export controls add leverage and risk.

Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Dominant in yield-critical inspection; hard to displace at leading edge.
Forensic read

Yield is everything at advanced nodes, and KLA effectively owns the inspection that protects it. A quiet, high-margin near-monopoly in its niche.

1 sub-components
Moat
strong
Growth
high
Margin
fat
Cost to run
heavy
Security
moderate
Substitutability verdict
Near-monopoly in yield-critical inspection.
Forensic read

Advanced-node economics depend on the tools it dominates.

Moat
strong
Growth
steady
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Concentrated supply of physical inputs; slow-growth but essential.
Forensic read

The literal raw material. Unglamorous, concentrated, and impossible to skip, the kind of supply-chain base that only gets attention when it breaks.

2 sub-components
Moat
strong
Growth
steady
Margin
moderate
Cost to run
heavy
Security
moderate
Substitutability verdict
Dominant silicon-wafer suppliers.
Forensic read

A quiet duopoly on the substrate every chip starts from.

Moat
strong
Growth
steady
Margin
moderate
Cost to run
heavy
Security
high
Substitutability verdict
Concentrated chemical inputs (JSR, TOK).
Forensic read

Choke risk hides here; a few suppliers, export-sensitive.

Cross-cutting forces · thread through every tier
Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
high
Substitutability verdict
The emergent moat; proprietary and rights-cleared data is the new scarcity.
Forensic read

As models converge, data becomes the differentiator. Rights-cleared proprietary corpora and quality labeling increasingly separate winners, but the vendor layer (labeling shops) is itself contestable as models self-generate training signal.

3 sub-components
Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
high
Substitutability verdict
Labeling incumbent moving up into data ops.
Forensic read

Real business, exposed as models increasingly self-label.

Moat
strong
Growth
high
Margin
fat
Cost to run
light
Security
high
Substitutability verdict
Owned, rights-cleared, hard-to-replicate data.
Forensic read

The actual durable moat as base models commoditize.

Moat
contested
Growth
hyper
Margin
moderate
Cost to run
light
Security
moderate
Substitutability verdict
Model-generated training signal.
Forensic read

Could erode the labeling vendors it competes with.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
critical
Substitutability verdict
The binding physical constraint on the entire chain; increasingly the ceiling.
Forensic read

Compute is now gated by electrons, not just chips. Power availability is re-rating from macro backdrop to primary investable theme: the buildout is capped by what the grid and generation can deliver.

3 sub-components
Moat
strong
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
critical
Substitutability verdict
Interconnection queues gating datacenter timelines.
Forensic read

The literal ceiling on how fast capacity can come online.

Moat
strong
Growth
hyper
Margin
moderate
Cost to run
extreme
Security
high
Substitutability verdict
Long-dated firm power for AI campuses.
Forensic read

High-conviction long-term supply, slow to build, heavily hyped.

Moat
strong
Growth
high
Margin
moderate
Cost to run
extreme
Security
high
Substitutability verdict
Bridge power co-located with datacenters.
Forensic read

The pragmatic near-term answer to grid constraints.

Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
critical
Substitutability verdict
Structurally necessary, still underbuilt; a category forming in real time.
Forensic read

Cuts through every layer: model theft, prompt injection, supply-chain integrity, data leakage. Under-provisioned relative to how fast AI is shipping to production, a whitespace category, and a systemic risk when it fails.

3 sub-components
Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
critical
Substitutability verdict
Theft, extraction, and prompt-injection defense.
Forensic read

Nascent tooling for a threat surface expanding faster than the defenses.

Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
critical
Substitutability verdict
Provenance across models, data, and dependencies.
Forensic read

Hard, unsexy, and increasingly mandatory for regulated buyers.

Moat
contested
Growth
high
Margin
moderate
Cost to run
light
Security
critical
Substitutability verdict
Runtime data leakage and access control.
Forensic read

Where breaches will actually happen as agents touch live systems.

Method note. Ratings are ordinal analytical judgments (chokepoint / strong / contested / commodity), not fitted statistics. The map is a reasoning surface, not a data feed. No market-share percentages are asserted anywhere. Tickers mark publicly listed exposure.