Infrastructure makes new capabilities available. Realizing their value requires judgment about where to apply them, the ability to integrate them into existing operations, and evidence that the resulting change improves performance. This report examines that distinction across enterprise IT, telecom and cloud, with Deming’s quality-management work as an earlier illustration.

The Conversion Premium describes the value attached to this proven application capability. It can reside in processes, customer relationships, institutional knowledge and delivery teams. These assets are harder to reproduce than access to a standard technology, although they remain exposed to competition, execution failures and changes in customer demand.

The historical evidence supports a conditional investment thesis. Implementation and operating capabilities can remain valuable after a buildout loses momentum, particularly when customers have durable operating needs. It does not establish that capacity always becomes worthless, that services always prosper, or that revenue growth necessarily translates into attractive investor returns.

Deming was invited to teach in Japan in 1950. His address to business leaders at Hakone linked quality to management responsibility, production methods and the needs of customers. His argument placed the responsibility for better output inside the organization: leaders had to connect technical methods with production practices and customer requirements. [1]

The economic relevance is the distinction between a teachable method and an effective operating system. Statistical techniques can be distributed widely. Applying them requires measurement, changes to processes and sustained managerial attention inside a particular organization. The analytical framework travels more easily than the working capability.

For an acquirer, that distinction changes the diligence question. Equipment and software establish what a business can do in principle. Its routines, people and record of delivery establish what it can reliably do in practice. The premium depends on whether that capability persists through a change of ownership and can be reproduced without dependence on a single individual.

The enterprise IT case is best illustrated by the growth of implementation businesses and the changing mix within Accenture. Infosys reported revenue of $413.8 million in FY2001 and $753.8 million in FY2003, an increase of approximately 82%. Its fiscal years end in March. These figures describe company-wide growth rather than an ERP-only revenue stream. [2]

Accenture’s FY2003 results show a more specific shift. Net revenue increased 2% to $11.82 billion, while outsourcing grew 37% and consulting declined 10%. Outsourcing generated approximately $3.57 billion, or 30% of net revenue. The useful finding is a shift toward operating and executing business processes, rather than uniform growth across services. [3]

The chart compares reported US-dollar growth over the same fiscal-year interval. It isolates the changing demand for two broad service categories within the same business. Accenture’s categories include activities beyond enterprise software implementation.

This evidence is consistent with enterprises continuing to pay for execution as technology spending priorities changed. It does not demonstrate that budgets moved directly from hardware purchases into outsourcing. Establishing that causal relationship would require customer-level spending data. The defensible conclusion is that useful operating capability retained demand even when other service categories contracted.Headline

Telecom illustrates how the usefulness of a network can outlast the financial assumptions behind its construction. Nortel’s as-reported revenue fell from $30.275 billion in 2000 to $11.418 billion in 2006. On a 2000 base of 100, the endpoint is approximately 38. These are reported company figures, not an organic-growth series, and the 2006 figure includes acquired revenue. [4]

The scale of Nortel’s losses requires equal care. Its Q2 2001 net loss of $19.4 billion included a $12.3 billion write-down of intangible assets. That composition does not justify attributing the entire loss to vendor financing. Acquisition valuations, operating performance and financing exposure should be analyzed separately. [5]

Equinix’s revenue rose from approximately $13.0 million in 2000 to $286.9 million in 2006 (22x). Its interconnection model brought networks and customers together in carrier-neutral facilities, adding value through access, reliability and customer density. Those features combined physical assets with operating expertise and network effects. Equinix therefore represents a hybrid infrastructure and services business rather than a pure asset-light implementation layer. [6]

Its growth also came with serious financial vulnerability. Equinix received a Nasdaq minimum-bid-price notice on August 15, 2002 and faced delisting. It completed its combination with i-STT and Pihana by the December 31 deadline set by the Nasdaq panel. Cost of revenues exceeded revenue through 2003, while acquisitions contributed to expansion. The 2000 starting point was a very small start-up base. Revenue growth should consequently be read alongside restructuring, losses and acquisition effects. [6]

The contrast supports a narrower claim than “the capacity owner lost and the conversion firm won.” Equipment supply and interconnection had different economics, and both involved capital and risk. What mattered was the ability to organize existing network capacity into a service customers continued to need. It is evidence of differentiation within the infrastructure stack, rather than proof of a universal rule about ownership.

MarchFirst provides a documented warning against treating services as protection from a technology bust. Court records establish that it entered Chapter 11 in April 2001 and converted to Chapter 7 on May 1. They do not establish that concentration in speculative dot-com clients caused its failure. It is therefore used here as evidence that a consulting business can fail, not as proof of a particular customer-mix explanation. [7]

Customer durability remains a relevant diligence test, alongside leverage, cash conversion, delivery economics and management execution. A durable customer base can support demand, but it cannot compensate indefinitely for an unsustainable operating model.

Cloud changes how enterprises obtain computing capacity and creates work around application architecture, data migration, security, integration and ongoing operation. This is a useful illustration of the conversion problem: standardized access to compute still leaves organization-specific decisions about how workloads should run.

The period from 2009 to 2013 should not be presented as another clean case of falling aggregate server revenue alongside growing services. A comparison beginning at the 2009 recession trough would also capture the subsequent recovery in server demand. Cloud is better understood here as a change in the organization of computing and the deployment work it requires. Broad Accenture and Infosys revenues cannot isolate the economics of cloud migration.

Cloud also exposes a limit of the capacity-versus-conversion distinction. Infrastructure operators can retain advantages through scale, location, reliability, distribution and integrated services. Implementation providers can face price competition and labor-intensive delivery. A position in the conversion layer is valuable only when its economics and differentiation can be demonstrated.

The historical cases give investors a way to examine AI businesses without assuming that access to compute or a model establishes an advantage. The relevant questions concern the result delivered, the institution paying for it and the operating capability that makes delivery repeatable. The analogy is a diligence framework, rather than a forecast that AI will reproduce any previous cycle.

The investable distinction is between access to a capability and a business that can turn it into a repeatable, economically useful result. The Conversion Premium is strongest where that result is proven, customers continue to need it, and the supplier can retain part of the value it creates.

Sources and evidence notes

Numbered references support the historical claims. The cloud discussion is deliberately qualitative. Revenue comparisons describe reported business performance and do not measure investment returns or establish a causal transfer of spending between firms.

1  Deming Institute
Speech by Dr Deming to Japanese Business Leaders in 1950. Primary transcript of the Hakone address. View source

2  Infosys FY2003 Form 20 F
Consolidated statements of income for fiscal years ended March 31. FY2001 $413.8M and FY2003 $753.8M; growth calculated as 753.8 / 413.8 − 1. View source

3  Accenture FY2003 Form 10 K
Fiscal year ended August 31. Management discussion of net revenues and consulting and outsourcing growth. Net revenues exclude reimbursements. View source

4  Nortel FY2000 and FY2006 annual filings
As-reported revenue endpoints. Index calculation: 11,418 / 30,275 × 100 = 37.7, rounded to 38. View source

4  Nortel FY2006 Form 10 K
FY2006 reported revenue and acquisition context. View source

5  Nortel Q2 2001 results Form 8 K
Reported quarterly net loss and intangible-asset write-down; no attribution of the full loss to vendor financing. View source

6  Equinix FY2003 Form 10 K
Early revenue base, business model, financial history and 2002 combination and Nasdaq events. View source

6  Equinix December 2002 Form 8 K
Combination with i-STT and Pihana and Nasdaq listing context. View source

6  Equinix FY2006 Form 10 K
FY2006 revenue endpoint. Growth includes acquisitions and starts from a small base. View source

7  US Bankruptcy Court Northern District of Illinois
In re marchFirst Inc; Maxwell v Novell Inc, court opinion. Bankruptcy chronology supports the failure example, not a causal client-concentration claim. View source