There is a common myth that maintaining an effective EA introduces a lot of cost to the Enterprise. An EA implemented in the right way reduces the level of effort and cost in an enterprise overall.
An effective EA needs to be supported by the right tools, and the right changes in associated processes (each of these processes in turn becomes more efficient and accurate) and be comprehensive in the set of stakeholders it supports.
A "Light weight" EA - where a superficial and partial set of information is assembled in the traditional way (i.e. without the right tooling, and without changing the touch point processes) may suit "light weight thinkers" - but as they fail to deliver substantial value and they add additional cost - they don't make any sense.
Establishing the changes in an enterprise that will allow an effective and complete EA to develop (and establishing the tooling, and changing associated processes) does have a cost but this cost is comparatively small if led from the right level within the enterprise, undertaken over a period of time (i.e. most cultural change is hard to effect quickly).
By way of one imperfect analogy - books and libraries
Assume your current strategy for managing books (i.e. knowledge - putting aside for now the fact that documents - especially in narrative form seldom make the semantics explicit, or relate knowledge in one book explicitly and atomically to knowledge in another book) is to leave the books scattered all over every office (and this is the form i.e. documents, and is the strategy essentially advocated at present; for most the design/governance artefacts associated with an enterprise).
Most people would find it quite difficult to find the correct information on most subjects, and if they wanted to record new information about subjects so that others could find it they would not know where to put it.
Say a proposed strategy is to establish a library function e.g. a classification system (Cf Zachman's framework), some places (physical or virtually) and some simple processes (that you would like people to follow). Now the actual cost of finding knowledge on any subject (i.e. what is said by a range of authors, with different perspectives) is reduced, and knowledge on subjects can know accrete in a natural way. Yes there is a set up cost.
A problem with this analogy is that most of the documents created in an enterprise are created by their authors for a very small and select audience (which in itself is odd), whereas most authors of books would ideally like their readership to be as broad as possible.
Another imperfect analogy - maps and the environment (built and natural)
If your current strategy for managing information about the environment (from rivers and mountains, to buildings and reticulation systems) is to leave documents scattered all over a wide range of offices/enterprise (and this was the form that the various stakeholders used until very recently). And a proprosed strategy is to establish a mapping function e.g. a classification system (a co-ordinate system, or a map projection), a collation mechanism and processes. The actual cost of finding out about any areas is reduced, and knowledge (on all areas) can accrete in a natural way (i.e. as the plans on buildings, cables, new information on soil types, etc. are deposited). Yes there is a set up cost.
No new information needs to be created - and in fact the effort associated with creating meaningful new information is reduced (e.g. it is easier to design changes if I can see clearly what already exists - buildings, cables, geological structures etc.) at the start of the exercise - rather than as I start construction (which is the ICT industry's approach).
Many artefacts are created by a function or unit for a very specific purpose and for a very small and select audience. The artefacts are usually focused on a narrow task (usually related to some delivery and/or construction activity), and specific timeframe (e.g. usually the immediate future). They are not oriented at: the ongoing maintenance, the effect on others, their discovery in future, the eventual replacement of what they is being delivering etc.). Governance (town planning functions) is slowly overcoming the impediments that the various engineering and cartographic specialists put in the way of this knowledge base by organising and sharing that knowledge.