Anthropic

PrivateFree featured Strategy Map

AI safety and research company developing large language models and AI assistants, including Claude, with a focus on responsible AI development.

Anthropicis private — you can’t buy shares. This map shows what you can buy that moves with it, and what moves against it, with dated evidence on every connection.

www.anthropic.comUpdated 2026-07-16research currentmethodology 2026.07.2Relationship mapping, not a recommendation.
Refresh research
Export

Decisive takeaway

Strongest public comparable: Global X Artificial Intelligence & Technology ETF (AIQ) etf holding overlap. Concentrated exposure to cloud infrastructure providers (Google, Microsoft, AWS parent Amazon) and NVIDIA GPUs that Anthropic depends on for training and inference.

What just changed

Broader market connections7

Similar public companies, funds, commodities and bonds tied to the same market — a looser connection, clearly labeled.

Global X Artificial Intelligence & Technology ETFAIQ · NASDAQ

ETF holding overlapetfPrice could riseDiversified equity fundindirectsource 2026-06-30inference
18conn
What it captures
Concentrated exposure to cloud infrastructure providers (Google, Microsoft, AWS parent Amazon) and NVIDIA GPUs that Anthropic depends on for training and inference.
What it misses
No direct exposure to Anthropic itself; captures enablers rather than the private AI application layer where Anthropic competes.
Why the price could fall
May be harmed if cloud providers shift capacity to proprietary AI models, reducing third-party API customers' access and demonstrating vertical integration risk.

Materiality: medium · Confidence: medium · Medium term (1–3 years)

Expense ratio 0.68% Holding weights are shown only when published by the issuer — the exposure may be material or incidental; check the fund’s current holdings.

View in IBKRverified 2026-07-16

Natural Gas FuturesNG · NYMEX

Commodity inputfuturePrice could fall as input costs risehigh riskLeveraged derivativeindirectinference
15conn
What it captures
Natural gas is the primary marginal fuel for US data center electricity generation, directly affecting operational costs for compute infrastructure.
What it misses
Does not capture renewable energy contracts, long-term fixed-price power purchase agreements, or regional electricity price divergence from gas prices.
Why the price could rise
May benefit if natural gas prices decline — the same driver that reduces electricity costs for cloud providers and lowers per-token inference costs at scale.
Why the price could fall
May be harmed if natural gas prices spike — the same driver that increases data center electricity costs and compresses gross margins on compute-intensive model training and inference.

Materiality: medium · Confidence: medium · Medium term (1–3 years)

Advanced instrument — structure-specific risks apply. Disclosures

Root NG · months All months · multiplier 10000 · physical delivery

View in IBKRverified 2026-07-16

United States Natural Gas FundUNG · NYSE Arca

Commodity inputetfPrice could fall as input costs riseDiversified equity fundindirectinference
9conn
What it captures
Equity ETF tracking NYMEX natural gas futures providing liquid exposure to the primary fuel driving US data center electricity costs.
What it misses
Contango/backwardation roll costs in futures-based ETF structure degrade long-term tracking; misses renewable energy hedging and fixed PPAs.
Why the price could rise
May benefit if natural gas futures decline — the same driver that lowers electricity input costs for compute infrastructure used in model training and inference.
Why the price could fall
May be harmed if natural gas futures rally — the same driver that increases power costs for cloud providers and pressures gross margins on compute-intensive AI workloads.

Materiality: medium · Confidence: medium · Medium term (1–3 years)

Expense ratio 0.012% Holding weights are shown only when published by the issuer — the exposure may be material or incidental; check the fund’s current holdings.

View in IBKRverified 2026-07-16

Roundhill Generative AI & Technology ETFCHAT · NASDAQ

Thematic fundetfPrice could rise on disruptionDiversified equity fundspeculativesource 2026-06-30inference
8conn
What it captures
Public companies integrating generative AI into products and distribution channels — potential strategic partners or acquirers of Anthropic's technology.
What it misses
Private AI startups including Anthropic itself; focuses on publicly traded beneficiaries rather than underlying model developers.
Why the price could fall
May be harmed if concentration among a few dominant LLM providers reduces pricing power for public software companies reselling AI capabilities.

Materiality: low · Confidence: low · Long term (3+ years)

Expense ratio 0.75% Holding weights are shown only when published by the issuer — the exposure may be material or incidental; check the fund’s current holdings.

View in IBKRverified 2026-07-16

ARK Autonomous Technology & Robotics ETFARKQ · CBOE

Thematic fundetfDiluted — small moves either wayDiversified equity fundspeculativesource 2026-06-30inference
4conn
What it captures
Exposure to AI-enabled automation and robotics companies that may integrate LLM reasoning for autonomous systems planning and natural language control.
What it misses
Direct enterprise software and API customers; ARKQ emphasizes physical automation over knowledge work applications where Claude is strongest.
Why the price could fall
May be harmed if robotics sector consolidates around proprietary end-to-end models, reducing demand for third-party LLM APIs like Claude.

Materiality: low · Confidence: low · Long term (3+ years)

Expense ratio 0.75% Holding weights are shown only when published by the issuer — the exposure may be material or incidental; check the fund’s current holdings.

View in IBKRverified 2026-07-16

Brent Crude Oil FuturesBZ · ICE

Commodity inputfuturePrice could fall as input costs risehigh riskLeveraged derivativespeculativeinference
3conn
What it captures
Oil prices influence global diesel and fuel oil costs for backup generators and emergency power systems in data centers, plus indirect transportation costs.
What it misses
Excludes primary grid electricity costs, which dominate total energy spend; backup power is only a small fraction of data center energy consumption.
Why the price could rise
May benefit marginally if oil prices collapse — the same driver that reduces diesel fuel costs for data center backup generators and lowers logistics expenses.
Why the price could fall
May be harmed marginally if oil prices surge — the same driver that raises standby power fuel costs, though these represent <1% of total data center OpEx.

Materiality: low · Confidence: low · Short term (0–12 months)

Advanced instrument — structure-specific risks apply. Disclosures

Root BZ · months All 12 months plus quarterly extends · multiplier 1000 · physical delivery

View in IBKRverified 2026-07-16

iShares iBoxx $ High Yield Corporate Bond ETFHYG · NYSE

Bond / creditbond etfBond — income and credit risk, not equity upsideCredit incomespeculativeinference
2conn
What it captures
Broad high-yield corporate credit exposure including lower-rated technology and infrastructure issuers that may serve AI industry supply chain.
What it misses
Anthropic is private and has no public debt; holds minimal direct AI company exposure; primarily captures broad credit market risk, not Anthropic-specific relationships.
Why the price could fall
May be harmed if broader credit market tightening or recession reduces access to capital for speculative-grade issuers, overwhelming any AI sector tailwinds.

Materiality: low · Confidence: low · Medium term (1–3 years)

Expense ratio 0.49% Holding weights are shown only when published by the issuer — the exposure may be material or incidental; check the fund’s current holdings.

View in IBKRverified 2026-07-16

Long / short mechanisms

Relationship mapping, not a recommendation.

GPU allocation tightening from helium-driven fab constraints

Semiconductor shortageconfidence: highinference

Strait of Hormuz disruption halts Qatari helium exports, forcing semiconductor fabs onto allocation and constraining NVIDIA GPU production. NVIDIA H100/A100 delivery timelines extend 6-12 months, limiting Anthropic's ability to scale training and inference capacity for new Claude models.

  • Global X Artificial Intelligence & Technology ETF (AIQ)price could falldirect · medium term · materiality high

    AIQ's AI infrastructure holdings face margin compression and delayed product launches as GPU scarcity drives allocation-based pricing and extended delivery timelines.

    Caveats: ETF diversification across AI value chain may partially offset GPU-dependent holdings Holdings with existing GPU inventory or long-term supply contracts less affected

  • Roundhill Generative AI & Technology ETF (CHAT)price could fallindirect · medium term · materiality medium

    CHAT's generative AI holdings with secured GPU capacity gain competitive moat as new entrants and smaller players face multi-quarter deployment delays.

    Caveats: Benefit accrues only to holdings with pre-existing supply agreements or cloud commitments Customer willingness to pay premium for scarce AI capacity uncertain ETF includes both GPU-constrained victims and supply-secured beneficiaries

  • ARK Autonomous Technology & Robotics ETF (ARKQ)price could fallindirect · medium term · materiality medium

    ARKQ's autonomous systems and robotics holdings face delayed product cycles as AI chip shortages extend development timelines for perception and planning modules.

    Caveats: Robotics compute requirements differ from LLM training (more inference-focused, lower total chip count) Some holdings may use alternative chip architectures (FPGAs, custom ASICs) less affected by helium shortage ARK's portfolio turnover may rotate toward less semiconductor-dependent positions

source: standing macro context - Strait of Hormuz transit disrupted by the Iran conflict (2026) Anthropic's existing GPU inventory and hyperscaler partnerships (Google Cloud, AWS) may provide buffer through pre-committed capacity Cloud providers may prioritize existing customers over new capacity requests Helium shortage impacts all advanced node production, not GPU-specific

Recurring export controls on frontier AI models

Export restrictionconfidence: high

US government export controls on Claude Fable 5 and Mythos 5 (lifted June 30, 2026) establish precedent for future restrictions on advanced models. Regulatory unpredictability may fragment Anthropic's addressable market and require costly nationality-verification infrastructure for global deployments.

  • Global X Artificial Intelligence & Technology ETF (AIQ)price could fallindirect · medium term · materiality medium

    AIQ's AI software and platform holdings face revenue volatility and market fragmentation as export controls create unpredictable access to frontier models for international customers.

    Caveats: Domestic US-focused AI companies in AIQ portfolio unaffected by export restrictions Non-frontier AI applications may continue serving international markets without controls Industry-wide jailbreak framework may reduce future control likelihood

  • Roundhill Generative AI & Technology ETF (CHAT)price could falldirect · short term · materiality medium

    CHAT's generative AI application holdings lose international revenue opportunities and partnership momentum when access to underlying frontier models becomes subject to sudden nationality-based restrictions.

    Caveats: Holdings relying on non-US or open-weight models face less exposure Enterprise contracts may include regulatory force majeure clauses 18-day control duration suggests limited long-term revenue impact if pattern holds

  • iShares iBoxx $ High Yield Corporate Bond ETF (HYG)price could fallindirect · long term · materiality low

    HYG's technology sector holdings face elevated refinancing risk as export control uncertainty pressures AI company cash flows and extends path to profitability for pre-revenue issuers.

    Caveats: HYG technology exposure is small relative to total portfolio Most AI companies are privately held or investment-grade, not in HYG universe Export controls affect only frontier models, not broader technology sector Venture-backed AI companies typically equity-financed, not high-yield bond issuers

source: https://www.anthropic.com/news/redeploying-fable-5 Controls were lifted after 18 days, suggesting framework may stabilize with clearer advance notice Glasswing partnership framework with Amazon, Microsoft, Google may create compliant distribution channel Enterprise customers in allied jurisdictions (EU, UK, Canada) may receive exemptions

Data center power costs surge from Iran conflict energy disruption

Energy-price spikeconfidence: mediuminference

Strait of Hormuz transit disruption from Iran conflict drives sustained energy price elevation, increasing electricity costs for hyperscale data centers. Anthropic's inference serving costs rise 25-40% as AWS and Google Cloud pass through power surcharges, compressing margins on API subscriptions.

  • Brent Crude Oil Futures (BZ)price could risedirect · short term · materiality high

    Brent crude futures rise 20-35% as Strait of Hormuz disruption removes 17-20% of global oil supply, tightening physical markets and elevating prompt-month prices.

    Caveats: Strategic petroleum reserve releases may dampen price spike Duration depends on conflict resolution timeline Demand destruction from high prices may limit sustained elevation

  • Natural Gas Futures (NG)price could riseindirect · short term · materiality medium

    Natural gas futures gain 15-25% as oil price spike drives fuel-switching demand for natural gas in power generation and industrial applications, tightening North American gas balances.

    Caveats: US natural gas prices primarily domestic-driven, less directly exposed to Middle East disruption LNG export capacity constraints limit price transmission from global markets Seasonal demand patterns (summer vs winter) affect magnitude of fuel-switching US shale production may respond quickly to higher prices, capping upside

  • United States Natural Gas Fund (UNG)price could risedirect · short term · materiality medium

    UNG tracks front-month natural gas futures which benefit from oil-linked fuel-switching demand, though contango roll costs may erode 30-40% of spot price gains.

    Caveats: UNG structural contango drag historically captures only 60-70% of spot price moves Fund rebalancing and roll execution may further reduce realized gains Natural gas storage levels and weather patterns create basis risk vs crude oil prices

source: standing macro context - Strait of Hormuz transit disrupted by the Iran conflict (2026) US data centers primarily powered by domestic natural gas and renewables, less exposed to Middle East oil disruption Hyperscalers may absorb costs short-term to protect customer relationships Long-term cloud contracts may include energy cost adjustment clauses with lag

Catalysts

What breaks the thesis

Business definition

Anthropic is an artificial intelligence safety startup founded in 2021 by former OpenAI executives, including Dario and Daniela Amodei. The company develops advanced large language models and AI systems with an emphasis on AI safety, interpretability, and alignment with human values. Its flagship product is Claude, a family of AI assistants available through API access and direct consumer interfaces, serving enterprises and developers with natural language processing capabilities.

Industries
Artificial IntelligenceMachine LearningEnterprise SoftwareAI Safety Research
Products & services
Claude AI AssistantClaude APILarge Language ModelsAI Safety ResearchConstitutional AI Technology
Customers
Enterprise BusinessesSoftware DevelopersTechnology CompaniesResearch InstitutionsIndividual Consumers
Business model
API subscription and enterprise licensing

Value-chain decomposition

  1. 01

    Compute Infrastructure & Data Centers

    Cloud computing resources, GPU/TPU clusters, and data center capacity required for training and serving large language models at scale.

  2. 02

    Training Data & Datasets

    Curated text corpora, licensed content, web data, and specialized datasets used to train foundation models and ensure quality outputs.

  3. 03

    Model Development & Research

    Core AI research, model architecture design, training pipelines, Constitutional AI methodology, and safety/alignment techniques development.

  4. 04

    Model Deployment & Inference Infrastructure

    Production systems for serving trained models, API infrastructure, latency optimization, and scaling capabilities for real-time responses.

  5. 05

    Product & Application Layer

    Claude AI assistant variants, conversational interfaces, enterprise features, and developer tools packaged for different use cases.

  6. 06

    API & Integration Platform

    Developer APIs, SDKs, documentation, authentication systems, and integration tools enabling third-party applications to access Claude models.

  7. 07

    Enterprise Solutions & Services

    Custom deployments, enterprise licensing, dedicated support, compliance features, and professional services for large organizational customers.

  8. 08

    Distribution & Customer Channels

    Direct sales teams, partnerships with cloud providers, developer community platforms, and consumer-facing interfaces for reaching end users.

Strategic dependencies

What this business materially depends on, upstream and downstream. Inferred dependencies are labeled — they are analytical hypotheses, not confirmed disclosures.

Upstream (inputs & infrastructure)

  • NVIDIA AI GPUssemiconductor · materiality high

    High-performance GPUs (H100, A100) required for training and inference of large language models at scale. NVIDIA chip design in US, manufacturing primarily TSMC Taiwan. Known constraint: Industry-wide GPU supply constraints for AI workloads, allocation-based distribution.

  • Cloud compute infrastructuredata center · materiality high

    Massive distributed compute capacity for model training runs and serving billions of inference requests. Primarily US-based hyperscale data centers. Known constraint: Power availability and cooling capacity limits in key regions.

  • Electricity for computeenergy · materiality high · inference

    Sustained multi-megawatt power draw for training runs lasting weeks to months and continuous inference serving. Known constraint: Data center power capacity constraints in key AI hubs.

  • Venture capital fundingfinancing · materiality high

    Multi-billion dollar funding rounds necessary to sustain capital-intensive model development and infrastructure costs before profitability. US venture capital markets.

  • AI research talentlabor · materiality high · inference

    Elite machine learning researchers and engineers with expertise in transformer architectures, reinforcement learning, and AI safety. Concentrated in San Francisco Bay Area, some remote. Known constraint: Limited global pool of top-tier AI researchers, intense competition from other AI labs.

  • Internet infrastructurecommunications · materiality medium · inference

    Global network connectivity to deliver API responses and web-based Claude interface with sub-second latency requirements.

  • AI governance frameworksregulatory · materiality medium

    Evolving AI safety standards, export controls on advanced models, and data privacy regulations affecting product deployment and capabilities. US federal (Executive Orders, NIST), EU AI Act, UK frameworks. Known constraint: Uncertain regulatory trajectory for frontier AI systems.

Downstream (customers & demand)

  • Enterprise API customerscustomer concentration · materiality high · inference

    Revenue concentration among large enterprise customers integrating Claude into products and workflows via API contracts. Primarily North American enterprises.

Omitted categories

Sources

Research completed 2026-07-16 07:58 UTC

Related companies and themes

These maps share public instruments with Anthropic — useful for comparing exposure or spotting concentration.

Assumptions and limitations

Change history

Nobody subscribes for a report.

You subscribe so you never have to re-do the report. Watch Anthropic and get an email when this map changes — new routes, broken theses, score moves.