I'm a customer of Gnip's - they provider several enrichments and data normalizations on top of Twitter's stream. Could Twitter have hired X engineers and duplicated what Gnip does? Perhaps - but Gnip also has an existing customer base, existing revenue flow, infrastructure setup to handle customers creating their own matching rules etc.

There's more to Gnip than just re-selling the Twitter firehose.

I'll try to simplify it as much as possible.

Capital is raised by companies to accelerate growth. Growth acceleration can manifest itself in many ways. Specifically, in this case, as you rightfully point out, Twitter could easily build this service themselves. They also, at the time of Gnip's inception weren't totally sure there was a business model here.* So they let another organization take that risk for them, instead of using their own capital. They end up paying more capital in the long run but the risk is reduced.

*I recognize that most likely the certainty of Gnip's business model is (1) much higher due to the "duh, of course people want this" nature of their business and (2) that the people investing here are all "in this together". It, however, doesn't negate the fact that there is indeed risk.

Or buying their way out of the data contract Gnip has?

Twitter provides their raw data to multiple outside companies. These companies compete with each other to add value to the raw data, while Twitter focuses on solving its core problems. Everyone wins. (Except Twitter's users, who make these companies millions of dollars without a penny in return.)