Startling statement: Cisco is in the business of selling routers. So it’s not surprising that its announcement today, which was billed as news that will change the Internet forever, involves said piece of hardware. Analysts were expecting an update to the CRS family, and the new CRS-3 was not a complete surprise. It is not always the case that corporate announcements meeting analyst expectations should automatically be discounted. 
Factoring in the players taking part in the event(AT&T), I will rate this announcement as news that will will help transform a voice-centric carrier to a data-centric one.
Here are some observations:
Having read through my share of enterprise telecom specs, I find carrier-class router capacities have the power to humble. A single line card in a CRS-3 funnels in 140-Gbps, which works out to 1.2 Tera-bits/sec passing through an 8-slot shelf. There are different card configurations, and ultimately service providers can build a multi-chassis CRS-3 to achieve the stated total of 322 Tera-bits/sec. Now, that’s a router!
Backing up the increased speeds, is (drum roll) a new chip with the impressively sounding name QuantumFlow. It is a highly-parallel RISC-based packet processor or really processors, 40 of them. The real-time OS for each PPE (as each processor element is called) is written in C (uh oh), and can manage 4 threads, working out to a pool of 160 simultaneous ( or concurrent?) threads that can be executed by any of the 40 PPEs.
Cisco openly states that is a “programmable” environment for rapid feature development—they call it “feature velocity.” No doubt a carrier requirement.
There’s very low-level accounting of each flow to track bytes and packet counts, as well as duration, for video and other data services. Yes, the chip has the capability to peform lawful intercepts.
There’s special hardware assistance in the PPE’s for multicast packet generation, and I read some wording that processors can retain state between passes, presumably as a way to save on memory and other resources. I’m out of my depth here, so I assume that this is what’s needed to satisfy carrier customers’ IP-TV broadcasting goals.
The CRS-3 hardware allows for “massive’ VLANs that can extend across data centers.
During the presentation, the slide titled Multi-directional Networking clearly shows service providers as content hosters/generators.
Finally, in 4Q09, AT&T reported that 70% of its revenue came from wireless and wireline broadband. And the company is managing “unpredented mobile broadband growth,” up 5000% in three years.
Conclusion: While this may not be a revolutionary annoucement for average customers, it is very big news for carriers. And especially for AT&T, which needs a way to replace its declining voice revenue share while also giving them an opportunity to prove to the FCC that it is willing to make significant capital investments.