The case for 40 Gigabit and 100 Gigabit Ethernet is most clearly spelled out in a recent white paper posted to the Web site of the Ethernet Alliance.
In the paper (PDF), authors from Cisco, Infinera and Intel discuss the reasons behind the standards push for both speeds of Ethernet - not just from the perspective of service providers (their increasing bandwidth needs are well known), but also from a data center perspective.
They don't really get into the politics that nearly hamstrung the Higher Speed Study Group, the disagreements over whether to go with one speed or the other. Instead, they take the end result - the decision to pursue both speeds as standards, within the same task force - and explain the reasons why there is desire to go after both.
The paper points to several applications that would necessitate 40Gbps or 100Gbps across short distances. For instance, the rise of multicore processors is cited as accelerating the need: "Multi-core processors are particularly valuable for multithreaded computing applications used in high-performance compute (HPC) clusters where many servers are networked together to execute large computing tasks through parallel processing." Such processors also mean more consolidation on fewer servers, servers that will require greater bandwidth needs.
The paper also sees networked storage as a pressing application, and a higher-speed Ethernet could accelerate the deployment of converged storage and data networks, something I touched on last time.
The paper also notes that 40 Gigabit Ethernet would have physical-layer interfaces to cover distances "inside the data center up to 100m for inclusion in a full range of server form factors including blade, rack, and pedestal." Meanwhile, 100 Gigabit Ethernet would have physical interfaces for data-center networking, but it would also potentially stretch to up to 40 kilometers over single-mode fiber.
Lastly, the white paper says that the standard could be complete by 2010.