Each layer constrains and enables the layers above it. Power flows upward, while constraints flow downward. Track which layers are active in each day's signal.
GPU, chips, data centers, power grids. The physical foundation that constrains all AI above. Who controls compute controls the ceiling for all other layers.
What Controls Here
NVIDIA, AMD, TSMC dominate chip supply. Nuclear and renewable energy determine data center capacity. Sovereign compute becomes geopolitical tool.
Why It Matters
L1 decisions cascade upward. Energy shortages constrain model training (L2). Chip supply shapes which companies can build which models. Export controls fragment the layer into US/China/EU silos.
Key Players
NVIDIA, TSMC, Intel, AMD, Google (TPU), Meta (custom silicon), Huawei, energy grid operators, nuclear/renewable developers
Signal Types to Watch
Data center power constraints
Sovereign computing initiatives
↓
GPT, Claude, Gemini, open-source alternatives. Who ships the base model shapes everything above. Closed vs open, reasoning capabilities, multimodal maturity — these define competitive moats.
What Controls Here
OpenAI (GPT), Anthropic (Claude), Google (Gemini), Meta (Llama), China players (Qwen, Baichuan) determine which models dominate training and inference workloads.
Why It Matters
L2 models are the gateway to L3+ layers. If OpenAI ships faster reasoning, all L3-L4 platforms must adapt. If open-source models become production-ready, the competitive landscape fragments.
Key Players
OpenAI, Anthropic, Google DeepMind, Meta, xAI (Grok), Microsoft Research, Mistral, Hugging Face, Chinese labs (Tsinghua, Baidu, Alibaba)
Signal Types to Watch
Model capability breakthroughs
Training cost improvements
Inference efficiency gains
↓
Agent orchestration, protocols (MCP, OpenAI APIs), embeddings, context windows. The hidden lock-in layer — where switching costs rise sharply once adopted.
What Controls Here
Anthropic (MCP), OpenAI (APIs), Hugging Face (ecosystems), LangChain, LlamaIndex shape how developers integrate AI. Data pipeline layers determine cost of training and inference.
Why It Matters
L3 appears technical but creates market power. If 10,000 startups build on MCP, switching to a different protocol becomes expensive — even if L2 models improve elsewhere. This is often invisible to outsiders but controls L4+ distribution.
Key Players
Anthropic, OpenAI, Hugging Face, LangChain, LlamaIndex, Ray, together with data infrastructure (Databricks, Pinecone, Weaviate)
Signal Types to Watch
Protocol or standard adoption
Lock-in changes in pipeline layers
Agent frameworks consolidation
Embedding quality breakthroughs
↓
ChatGPT, Claude.ai, Copilot, agentic OS. The consumer/enterprise gateway — where users meet AI. Platform control determines distribution and pricing power.
What Controls Here
OpenAI (ChatGPT), Anthropic (Claude), Microsoft (Copilot), Google (Gemini), Apple (on-device), Samsung define which interfaces capture users and use data.
Why It Matters
L4 is where the money flows directly from users. If ChatGPT captures 100M+ users, OpenAI owns the relationship — regardless of which L2 model powers it. This drives enterprise ARR and consumer stickiness.
Key Players
OpenAI, Anthropic, Microsoft, Google, Apple, Meta, Perplexity, Together AI, Replicate, Hugging Face, edge AI companies
Signal Types to Watch
Platform user growth/retention
Pricing changes or tier launches
↓
Consumer apps (design, writing, coding tools), enterprise SaaS (sales automation, customer service, finance). Revenue layer where AI creates ARR and proves ROI.
What Controls Here
Startups (Figma, Canva, NotebookLM), established SaaS (Salesforce, HubSpot, Microsoft 365), consumer apps (Runway, Midjourney, Cursor) monetize AI at the user level.
Why It Matters
L5 proves that L2-L4 have real-world value. When Figma ships AI features and keeps users sticky, it validates the entire stack. Failed L5 applications signal that L2-L4 promise is not yet real.
Key Players
Midjourney, Runway, Cursor, Figma, Canva, Notion, NotebookLM, Together AI, Hugging Face Spaces, thousands of indie startups
Signal Types to Watch
Founder funding and exits
Churn and retention shifts
↓
Physical AI, robotics, enterprise automation, industry-specific deployments. Where AI translates from digital to real-world ROI in specific verticals (manufacturing, healthcare, finance).
What Controls Here
Boston Dynamics, Tesla (Optimus), Figure, scale AI integrators track how deeply AI penetrates manufacturing, healthcare, logistics, finance, and defense.
Why It Matters
L6 ROI determines whether the entire L1-L5 stack survives. If AI can't prove physical productivity gains in key verticals, capital dries up. L6 success compounds back to drive L7 capital inflows.
Key Players
Boston Dynamics, Tesla, Figure, Sanctuary, Scale AI, ABB, KUKA, industry integrators, enterprise software (SAP, Oracle, Salesforce) embedding AI
Signal Types to Watch
Robotics deployment scale-up
Enterprise ROI announcements
Factory/facility automation wins
Workforce displacement signals
↓
Funding rounds, M&A, IPOs, valuation signals. The acceleration layer — capital flows respond to L1-L6 signals and fuel the next cycle of investment.
What Controls Here
Venture capital (a16z, Sequoia, Benchmark), corporate venture (Google Ventures, Microsoft Ventures), sovereign wealth funds, hedge funds determine where capital flows and at what valuation.
Why It Matters
L7 responds to L1-L6 but also drives L1 investment. When $200B flows into AI, chip makers get signal to expand fabs. When AI valuations crash, L1 capex slows. This is the feedback loop that accelerates or dampens the entire stack.
Key Players
a16z, Sequoia, Greylock, Bessemer, corporate venture arms of GAFAM, sovereign wealth funds, public markets (NASDAQ, NYSE), Bloomberg and data providers
Signal Types to Watch
Mega-round funding announcements
M&A and acquisition patterns
Valuation multiples shifts
↓
EU AI Act, US-China chip decoupling, CFIUS reviews, bloc formation (US, EU, China, emerging markets). The fragmentation layer — where politics redraw market boundaries.
What Controls Here
Governments (US, EU, China, India, UK), regulatory bodies, trade bodies determine which companies can operate where, who can export what, and how AI deployment is constrained.
Why It Matters
L8 fragments the global AI stack into geopolitical zones. A startup that works in the US may be banned in the EU. Chinese models may not function in the US. L8 shapes whether the market is one or many.
Key Players
US government (CFIUS, Department of Commerce), EU (EC, Parliament), China, UK, India, Japan, sovereign regulators, trade organizations (USMCA, CPTPP)
Signal Types to Watch
Export control announcements
Regulatory changes (AI Act, etc)
↓
Alignment, environmental pressure, deepfake risk, systemic AI failure modes. The constraint layer — where safety and existential risk set hard limits on deployment.
What Controls Here
Alignment researchers (Anthropic, OpenAI, Redwood Research, DeepMind), safety researchers, environmental engineers, policy thinkers determine acceptable risk thresholds.
Why It Matters
L9 does not accelerate growth — it limits it. When alignment risk becomes visible or environmental cost is too high, regulation follows fast. L9 is where L1-L8 hit their ultimate constraint.
Key Players
Anthropic, OpenAI, DeepMind, Redwood Research, Stuart Russell (UC Berkeley), Paul Christiano, MIRI, academic researchers, environmental organizations
Signal Types to Watch
Alignment research breakthroughs
Safety incidents or near-misses
Environmental impact reports
Deepfake or misuse incidents
↓
Labor markets, education, wealth concentration, culture. The terminal effect layer — where AI reshapes human society. The ultimate feedback loops flow from L10 back to L8 as political backlash.
What Controls Here
Labor economists, sociologists, policymakers, educators track workforce displacement, income inequality, education disruption, cultural shifts caused by L1-L9.
Why It Matters
L10 is not a power layer — it's an impact measure. If L1-L9 create mass unemployment or extreme inequality, political backlash hits L8 (regulation accelerates). If L10 shows broadly shared benefits, regulation stays permissive.
Key Players
Labor departments, education ministries, researchers (Erik Brynjolfsson, Daron Acemoglu), media, public discourse, workers' organizations, inequality monitors
Signal Types to Watch
Workforce displacement reports
Education system disruption
Cultural and media shifts