Users, Enterprises & Distribution
Demand and economics · Who actually pays for all of this
- Why it matters
- Demand is what makes the whole build out economic. Adoption that stalls at pilot stage would eventually be visible in accelerator orders and data centre commitments.
- The bottleneck
- Organisational change. Skills, process redesign, procurement, risk approval and the simple difficulty of getting people to work differently.
- Who captures value?
- Businesses with distribution into an existing workflow, proprietary data about how a task is really done, and domain expertise that is not present in public training data.
- What could change?
- Measurable returns are still uneven. A period of visible disappointment would slow spending, while a clear productivity result in one industry tends to pull the rest along quickly.
Technology creates economic value only when people and organisations use it. Demand runs across every layer rather than sitting on top of one, which is why it is treated here as a cross-cutting question.
Every gigawatt, every accelerator and every training run is a bet that someone downstream will pay for the result. So far the clearest paying groups are developers, who adopted AI tooling faster than any other profession, and large enterprises buying assistants and agents inside software they already own. Consumer subscriptions are real but smaller. Whether the current build out is justified is, in the end, a question about demand rather than about any single layer below it.
Tap or hover a box to see what it does
Enterprise adoption and integration
Large organisations rarely deploy AI unaided. Integrators and core business systems are where model capability gets connected to how a company actually runs.
Accenture
Delivers consulting and systems integration, including a large book of generative AI implementation work.
Most enterprises cannot deploy AI without help. Integrators convert model capability into changed processes, and they bill for the gap.
Discretionary spending is cyclical, and better tooling could eventually shrink the implementation gap they monetise.
SAP
Provides enterprise resource planning software that runs core finance, supply chain and HR processes for many large firms.
The transactional record of how a business actually operates lives here. That context is hard to reproduce and expensive to move.
Long migration cycles, and customers who resist paying more for AI features on top of existing licences.
Developers and workflow tools
Developers adopted AI assistance faster than any other professional group, which is why coding tools became the first genuinely large paid category.
GitHub (Microsoft)
Hosts code and ships Copilot, the coding assistant most developers encountered first.
Developers are the fastest adopting user group in the entire ecosystem, and the tool that sits in the editor has an unusually strong habit advantage.
Coding assistants are now a crowded category with capable rivals and rapid switching.
Microsoft
Runs Azure, distributes AI through Microsoft 365 and Copilot, partners commercially with OpenAI, and designs its own Maia accelerators.
Few businesses touch as many parts of the ecosystem at once. It owns infrastructure, a commercial route to frontier models, and the enterprise distribution to sell the result.
Very heavy capital spending against uncertain payback, dependence on partner model roadmaps, and buyers questioning per-seat AI pricing.
Consumers and small business
Assistance built into products people already pay for, where the proprietary data and the trusted workflow do more competitive work than the underlying model.
Intuit
Sells financial software to consumers and small businesses, with AI assistance built into tax, accounting and payments products.
A useful example of the moat that comes from proprietary data plus a workflow customers already trust with their money.
Competition from lower cost tools and regulatory attention on parts of the consumer tax market.
Alphabet (Google)
Builds Gemini models through Google DeepMind, designs TPU accelerators, operates Google Cloud, and distributes AI through Search, Workspace and Android.
It is one of the few organisations that designs the chip, runs the data centre, trains the model and owns the consumer surface it ships on.
AI answers can cannibalise the advertising business that funds everything else, and antitrust proceedings could reshape distribution.
How to think about the economics
- GrowthHigh
How quickly demand in this part of the ecosystem is expanding.
- Capital intensityVery low
How much money has to be spent up front before revenue arrives.
- Competitive moatHigh
How difficult it is for a credible new entrant to take the business.
- Customer concentrationLow
How much revenue depends on a small number of buyers.
- Disruption riskModerate
How exposed the layer is to a technical or commercial shift.
This is a framework for thinking about the economics of a layer, not a recommendation. An important AI company, a strategically advantaged company, an investable security and an attractively valued security are four different things.
Where value may accrue
Distribution, switching costs and proprietary process knowledge. An application with an ordinary model and excellent distribution generally beats an excellent model with none.
Related questions
Is this just another tech bubble?
Parts of it may be. Some valuations clearly assume everything goes right. But the infrastructure spending is happening regardless of which application company wins, which is the argument for owning the suppliers rather than guessing the winners. The honest position is that the ecosystem contains both durable businesses and speculative ones, and the market is not currently doing a careful job of distinguishing them. That is a reason to read the economics of each layer rather than to treat AI as one trade.
What is agentic AI?
Most AI use is still reactive: you ask, it answers. An agent is given a goal instead, and works out the intermediate steps, calling tools, reading data and looping until it decides the task is finished. That shift is why permissions, audit trails and identity suddenly matter so much. A chatbot that is wrong wastes your time. An agent that is wrong can take an action on your behalf, which is a different category of problem.
What is the difference between training and inference?
Training is not a single event. Pre-training is the long, expensive phase where a model learns general patterns from very large amounts of data. Post-training shapes that raw model into something useful and better behaved, often using human feedback and reinforcement learning. Fine-tuning adapts an existing model to a narrower task or domain, and can be done repeatedly and comparatively cheaply. Inference is the model actually answering, which happens billions of times a day. Much of the current demand for chips, power and data centres is driven by inference at scale rather than by headline training runs alone.
What is a foundation model?
A large, general purpose model trained on a very broad dataset, which can then be adapted to particular tasks. The GPT, Claude, Gemini and Llama families are all foundation models. They are called foundations because applications, tools and agents get built on top of them. Training one is an expensive bet that sitting at the base of the stack is where lasting value accrues, which is precisely the question the rest of this site is trying to help you think about.
Sources and further reading (2)+
- 1Business Trends and Outlook Survey, AI use by firms · US Census Bureau
- 2Annual Developer Survey, AI tooling adoption · Stack Overflow
The AI ecosystem changes rapidly. Company positions, technologies and market data reflect information available at the date above.