Is Event-Driven Architecture and Complex Event Processing an Affordable Option for Most Businesses?

Any business today is event driven, more or less; otherwise it would not respond to real demands.
On the other hand complex event processing where events correlation is build in to simplify and improve decision making is desirable and on its way in the telecommunications field where a degree of automation exists.
But I am not sure what affordable in this context: cheap, implementable, mature...?
SOA and event processing may work together though.

Most Businesses, have no other choice. The real world and the enterprises business are Event Driven in nature. Not all events are simple, so most Businesses have to take care of CEPs as well. SOA Best Practices revealed limited success of implementing SOA, without implementing Event Driven Architecture (EDA) concurrently.
The practical question is a timing question. If the organization is not mature enough, it should postpone its EDA and CEP implementation, as well as postponing its SOA initiative. One thing is sure: even if you implement EDA and CEP right now in most cases it will take a long time to complete it.

It depends.

The size of the business and investment levels in IT will certainly be a driving factor in the feasibility of a corporation employing EDA/CEP. However, I think a more germane factor would be the level of (necessary) complexity and interconnectedness the organization has with others. Businesses that rely on weather forecasts, foreign market results, government statistics, or call center volumes/average speed to answer (ASA) all would benefit from being able to peer deeply into their operations at a statistical level.

EDA/CEP becomes an enabling tactic to assist. If I'm WalMart and the weather events indicate a hurricane heading to a particular city, I can start diverting supply chain resources in that direction. If I'm running a call center and ASAs are starting to increase, I can change the call routing patterns to other call centers.

I think EDA/CEP will soon become an essential part of all organization's Enterprise Architecture. One cannot performance tune anything - cars or enterprises - without the necessary diagnostic data. EDA/CEP provides that diagnostic data to the organization.

It's ultra affordable - all you need is some good developers with some EDA architecture chops.
If the emphasis is on affordability, there are great many open source and (license cost) free software products that can be used as EDA/CEP system building blocks.
For example, here are just some in the Java ecosystem:
* ActiveMQ as a message broker
* JGroups as a reliable multicast implementation
* Terracotta/EHCache as a distributed cache
* QuickFIX/J as a FIX engine
* Esper as a CEP engine

None of these products require any licensing costs (unless you want to pay for additional support) and can be used to build reliable EDA systems with CEP capabilities.

In addition to that, depending on the business problem, hosted solutions are also very affordable and can be used for building such systems. For example, Amazon SNS can be used as a backbone for a hosted EDA system:
http://aws.amazon.com/sns/

Disclaimer: I'm giving a talk titled "Open Source EDA" where I demonstrate how to build an EDA system with CEP capabilities using an open source Java stack: http://strangeloop2010.com/talks/14489

Looking at it as a cost is an incomplete way of looking at event and complex event processing (CEP). Event-oriented Architecture (EOA), which CEP is a subset, is the missing third architecture in most enterprise solutions, the first two being Service-oriented Architecture (SOA) and Data-Oriented Architecture (DOA). The statistics show that ver 98% of enterprise class solutions use some form of DOA in their data supply chain activities and between 30-40% use some variant of SOA. However, less than 5-10% are currently tapping into the benefits of what EOA can deliver.

EOA (events and CEP) offer answers and insights into knowledge, something SOA and DOA cannot. Think of the Wisdom Model: Data, Information, Knowledge, and Wisdom. At the lowest layer we have data. The universe if filled with it. For example, consider a university where you have data like students, books, building, colors, plants, etc. This is the realm of DOA. If you provide relevance to data, we call it information. Every Monday morning a university Provost is looking at a list of failing students and is trying to figure out just what to do. This is the realm SOA. When you study information, as in the case of the Provost, we call that knowledge. The provost is asking questions like, “Why are they failing?� This is the realm of EOA (CEP). Lastly, in depth reflection on knowledge provides us with wisdom. What is really important for the Provost is internalizing what it will take to turn failing students into successful students. This is the realm of EOA.

Once you map revenue potential for each of these layers in the wisdom model (Data, Information, Knowledge, and Wisdom) against each orientation (DOA, SOA, EOA), you soon come to the realization that EOA offers the highest growth rate for new revenue and margin. More on this later if you like.

So, a more business-oriented perspective would be to asked the question, “At what margin can we grow new revenue through EOA (event and complex-event) services? This is a conversation that most CEOs and CFO would gladly engage you in.

I guess another way to look at this is - can businesses (building out their own IT infrastructures) afford NOT to be event-driven?

As already commented - businesses are already event driven. It's just most IT systems are not ...