Because they got this through an out of court settlement, so didn't have the risk of taking it to court (which includes a risk of winning nothing, and of counter-claims), they avoid the delay and legal costs of trial, the embarrassing public disclosures of how both sides screwed up. NG never expected to win the full $1bn. Also, by taking any settlement, NG have created a situation where on causal observation it seems Wipro have informally admitted guilt...and in that respect, read these next couple of paragraphs
In court, NG would have got nothing for anything that was largely their own fault, and Wipro's lawyers would have been pointing to the myriad "client failures" that contribute in any big IT screwup. Even things that were largely Wipro's fault, if Wipro could show that NG contributed in some small way to the problem their liability is greatly reduced. But take a look at the link below, and read the list of bullet pointed errors under the heading "Audit Findings". These can be summarised as NG asked for too much, too soon, too cheaply, and without providing the necessary resource and data foundations to enable success. Those rest with NG's management, and although I get the impression that Wipro are invariably associated with cheap, low quality work, I'd say splitting the costs $75m to Wipro and $925m to NG sounds about fair. Indeed, a better headline for the article would have been Wipro Wipe Brow Thankfully As National Grid Chew On $925m Cost Of Fail"