Okay, I'm sold on the cloud vision but there are far more technical hurdles to jump over than the tech industry is willing to admit. That said, the IBM System z (the great-great grand child of the 360 if I'm correct) may be far more technically suitable for cloud computing in the near-term.
I believe that this is true because the mainframe architecture already supports:
1. A horizontally scalable architecture. IBM mainframes support a scalable clustering technology called Parallel Sysplex. Parallel Sysplex can even be extended over moderate distances using Geographically Dispersed Parallel Sysplex (GDPS). A technology like GDPS could be the foundation for cloud-based statistical multiplexing for scale up/scale down compute requirements.
2. An open source platform stack. Mainframe growth is being driven by Linux partitions for net new applications. Combine mainframe Linux with Apache, MySQL, and PHP (the LAMP stack) and you've got a low cost open platform for development.
3. Unmatched security. I could go on and on here but the two things I'll highlight are data encryption and PKI. System z supports an encryption co-processor, key management, and several system level facilities to encrypt data sets. Combined with virtualization and access rules, partition "A" cannot read the data of partition "B." This enables secure resource sharing. Mainframe access control solutions like RACF, TopSecret, etc. can also act as a PKI infrastructure. To me, this is important for setting up trust relationships between in-house and cloud-based mainframe capacity.
4. Virtual networking services. A big obstacle to cloud computing is tranforming physical networks to virtual architectures and services. IBM faced and addressed this problem when it figured out how to make mainframe resources available over TCP/IP and higher layers of the OSI stack.
To be clear, the long-term future of cloud computing will likely be dominated by the Intel x86 architecture but mainframe clouds will also become a cloud player by providing a few valuable services. Users will be able to weigh the financial and operational costs of extending their mainframe environments with cloud services rather than purchasing additional boxes. This will become especially important as more IT professionals with mainframe skills retire. Additionally, a LAMP stack is a LAMP stack. If I am comfortable developing applications on LAMP and running them in the cloud, who cares what type of hardware it runs on? Finally, it is hard to argue with the environmental economics of the mainframe. This will be especially attractive in areas with high utility costs.
I remember when a lot of industry pundits proclaimed that the mainframe was dead. Ironically, as the mainframe meets cloud computing, many of these so-called "experts" will move on to the after-life far before the 360 architecture does.