Identifying Profit Pools in Industry Value Chains

Hello. Last lesson separated a company’s revenue models from the categories it reports. That distinction matters because revenue is not the same as retained profit: a business can process enormous transaction volume yet earn a modest fee, or generate substantial sales while passing much of the value to suppliers.

This lesson asks where profit actually accumulates across a Software & Platforms value chain. You will build a profit-pool map using revenue, margins, growth, and bargaining-power evidence, then use it to distinguish a profitable layer from an attractive opportunity. Allow about 40 minutes, including the readings and short video excerpts.

1. Define the pool before looking for winners

A profit pool is the aggregate profit earned by businesses in a defined part of an industry over a defined period. The unit of analysis might be a value-chain layer, such as cloud infrastructure, or a narrower layer-and-segment cell, such as managed data platforms sold to large enterprises.

That is a different question from “Which company has the most revenue?” It is also different from “Which company has the highest margin?” A small business at a 40% margin may contribute less to the industry’s total profit than a much larger business at a 15% margin. At the simplest level:

The approximation becomes useful only after you specify what counts as the layer, whose revenue belongs in it, and which profit measure you are using.

Read Umbrex’s Profit Pool Mapping Framework for the underlying mapping method. Its useful move is to cross value-chain layers with economically distinct customer or product slices, rather than treating an entire industry as one market.

The image below makes the resulting map concrete. Each cell estimates profit for one layer and one slice; darker cells and larger bars indicate where more aggregate profit is estimated to reside. Its rows and columns are illustrative, not a prescribed taxonomy for technology industries.

For software and platforms, useful layers might include semiconductor inputs, data-center capacity, cloud infrastructure, managed platforms, applications, distribution, and implementation services. Useful slices could be enterprise versus smaller customers, direct versus partner-led sales, or distinct workload types. Choose slices because their economics differ, not merely because they appear in a reporting table.

One caution from the prior lesson carries over: if a company reports a blended cloud segment, its segment profit is not automatically the profit of the cloud-infrastructure layer. The segment may also contain managed services, and its costs may be shared with other businesses. A map must allocate both revenue and relevant costs without counting the same activity twice.

2. Calculate absolute profit, then test capital intensity

Consider the following invented, annual illustration of an AI-serving ecosystem. These are not estimates of the actual market. Assume each row represents a distinct set of external businesses or allocated activities, with no company counted twice.

The largest current EBIT pool here is cloud infrastructure at $\$8.0$ billion, despite its margin being lower than the margins of accelerator suppliers and managed platforms. The fastest-growing row—AI applications—contributes only $\$2.0$ billion of current EBIT. Thus, largest pool, highest margin, and fastest growth name three different positions.

The revenue figures across rows should not be summed and called final-customer spending or market size. A chip supplier’s sale can become part of a cloud operator’s cost, and the operator’s sale can become part of an application provider’s cost. The rows represent successive businesses’ revenues, not independent purchases by end users. Profits can be aggregated across distinct actors within the stated boundary, provided internal transfers and vertically integrated activities are treated consistently.

EBIT is a practical common measure, but it does not tell the whole story when one layer requires far more capital. Suppose, still entirely illustratively, that the data-center operators employ $\$60$ billion of capital and the managed-platform businesses employ $\$15$ billion. With an assumed 25% tax rate and a 10% capital charge:

For data-center operators, that yields $\$4.8\text{bn}\times 0.75-\$60\text{bn}\times 0.10=-\$2.4$ billion. For managed platforms, it yields $\$6.6\text{bn}\times 0.75-\$15\text{bn}\times 0.10=\$3.45$ billion. The assumptions are deliberately simplified, but the comparison shows why positive operating profit need not imply an attractive return after the cost of capital. Leases, shared assets, and capitalized software development would require careful treatment in a real analysis.

This suggests a sensible reporting convention: show aggregate EBIT to locate today’s operating-profit pools, and add economic profit or returns on invested capital when layers have materially different capital requirements. Do not mix EBITDA for one row, EBIT for another, and net income for a third.

3. Explain the pattern with bargaining power—and test whether it is moving

A profit-pool map tells you where profits appear. Bargaining-power evidence helps explain why they remain there or migrate. The relevant question is not simply “Is demand high?” It is: Who can retain the benefit of that demand after customers, suppliers, and competitors negotiate?

Cloud computing offers a compact example because a single customer workload depends on several parties. Read the value-chain description and the selected buyer- and supplier-power passages in Think Insights’ Industry Analysis: Cloud Computing. Use its descriptions to form hypotheses; do not treat qualitative claims about leverage as measured margins for every cloud provider.

A large cloud customer might negotiate a substantial discount because it can commit significant spend or credibly shift some workloads. The provider’s revenue can grow with that customer’s usage while its margin on the account narrows. At the other end of the chain, a scarce input supplier may negotiate higher prices or control allocation. Whether that supplier earns more profit depends on its own costs and capacity—not on scarcity alone.

For each cell of a map, look for observable consequences of power:

These are clues, not automatic proofs. For example, technical switching difficulty may help a managed platform defend price, while its dependence on a cloud provider may weaken its position on infrastructure costs. Bargaining power can point in both directions for the same business.

Growth adds a time dimension. In the illustrative table, AI applications grew fastest in the prior year. If their revenue subsequently rises 50% from $\$25$ billion and their EBIT margin improves from 8% to 14%, their EBIT would reach $\$5.25$ billion. If revenue rises by the same amount but competition and delivery costs reduce the margin to 3%, EBIT would be only $\$1.125$ billion—less than the starting $\$2.0$ billion. Revenue growth alone cannot distinguish those futures.

A short excerpt from Sequoia Capital’s AI’s Trillion-Dollar Opportunity is useful here as an investor thesis to interrogate, not evidence that AI application profits have already concentrated. Notice the prediction about application-layer value and the speaker’s separate insistence on durable revenue and a path to margins.

The discipline is to keep current profit, growth evidence, and forecast profit separate. A promising AI-native category may have strong demand yet limited current EBIT; an established software layer may have slower growth but a large, resilient pool. Neither observation settles where the pool will be in three years.

4. Produce a map you could defend in a client discussion

A first-pass profit-pool analysis need not pretend that every private company’s margins are visible. It does need to show how its conclusions follow from evidence.

Begin by fixing the boundary and period: for example, a particular set of enterprise workloads, the regions included, and one fiscal year. Decide whether you are measuring suppliers’ and intermediaries’ profit across the whole ecosystem or only profit earned from a defined group of end customers. That decision determines which transactions belong in the map.

Next, build a revenue-and-profit bridge for each layer. Use disclosed segment revenue and EBIT where they genuinely match a cell. Where they do not, state the allocation: perhaps a portion of a hybrid company’s reported segment belongs to managed platforms and the remainder to infrastructure. Allocate associated direct and shared costs as well; assigning high-margin revenue to a cell while leaving its R&D or sales costs elsewhere would inflate its pool. Preserve an unallocated or uncertain amount rather than forcing a false precision.

Then triangulate the gaps. Public filings can establish reported totals; transaction or workload measures can help estimate volume; comparable businesses and bottom-up unit economics can provide margin ranges. Check that each layer’s estimates reconcile with known company totals and that a fee or pass-through amount is not counted twice. If one poorly disclosed participant could change the ranking, show a range or sensitivity instead of a single confident bar.

Finally, explain the result in two sentences that separate observation from interpretation:

Observed or estimated: Cloud infrastructure currently has the largest aggregate EBIT in the defined sample, while managed platforms have a higher EBIT margin.
Interpretation to test: Managed platforms may defend that margin through embedded customer workflows, but customer discounting or higher infrastructure costs could reduce their share.

A strong map should also mark its direction of travel. Track changes in revenue share, realized margin, capacity investment, customer discounts, and supplier prices over successive periods. When those signals conflict, say so. A layer can gain revenue share while losing profit share; a temporarily scarce input can command high prices without sustaining high returns once capacity expands.

For consulting strategy, the map identifies where investigation is worthwhile, not where entry is guaranteed to succeed. A high-profit pool may require a control point an entrant lacks. A thinner pool may offer a compelling opportunity if better operations, distribution, or cost structure allow a particular firm to outperform the average. That distinction between industry attractiveness and a company’s ability to win is essential.

Conclusion

To locate a profit pool, define the industry boundary and layer-and-segment cells, estimate aggregate profit with a consistent measure, and reconcile allocations to reported evidence. Compare absolute profit with margins and capital requirements; then use growth and bargaining-power signals to assess whether the pool is expanding, eroding, or moving. Treat confident predictions about future value capture as hypotheses until prices, costs, and retained profit support them.

The next lesson turns to sector-specific operating metrics in earnings materials and filings—the disclosures that help test these estimates and monitor changes in the pools over time.