We study a scenario where a buyer (e.g., Uber) buys cloud capacity from a seller (e.g., Amazon Web services) to run its business. One of the key factors that affects the quality of cloud services is congestion, and it has drawn considerable attention in recent years. Congestion leads to a potential loss of end users (e.g., riders and drivers of Uber) thereby adversely affecting the demand for cloud services. Discount has been a useful mean to stimulate demand and reward customer loyalty. However, in the presence of congestion, the effect of discount on demand is ambiguous. On the one hand, a higher discount leads to higher demand; on the other hand, higher demand can lead to higher congestion, thereby lowering the demand. Given that end users are both price and congestion sensitive, the choice of optimal discount under congestion is therefore not straightforward. Using a game-theoretic model, we study the dynamics between congestion and discount, and explore how congestion moderates both buyer's and seller's optimal decisions. Our results show that the buyer is not necessarily worse-off even when the end users are more intolerant to congestion. In fact, we find that when end-users are more congestion sensitive, the demand of cloud services can actually sometimes increase and the discount offered by the seller can decrease. These findings have important managerial implications on the seller's pricing and capacity decisions. We also observe that a lower cost of technology can sometimes hurt the buyer and the buyer can pass on lower benefit to end users. Moreover, given that the cloud services are prone to disruptions, buyer sources from multiple cloud vendors that further complicates the matter. We draw useful insights about the buyer's procurement decisions under congestion in a multi-cloud setup.