In a region like silicon valley the opportunity and availability to walk away for a startup is huge this could act as a retention tool also. Better to let a talent disappear for 20% of their time than lose them altogether.
This is like someone betting $1000 on a game, and then calling it a $1000 loss up until the game is actually played (at which point it becomes indeed a true $1000 loss, or possibly a gain). It doesn't make any sense.
Wave is probably an even bigger disappointment: it would have done great as an enterprise product.
This might work for "moonshots" that are closer to their core business, but in other areas there's not going to be any inherent value in building expertise if the "moonshot" itself ends up failing, because at that point they will cease needing any expertise in that area at all.
Likewise, "flexing the idea muscle" would be a benefit if it was regular Google employees that were working on these projects but it's not. It's Alphabet's X lab which is it's own thing, with people working full time on these moonshots.
Internally, they're referred to as 'bets, a pun on the company name Alphabet.
Though I think even $1B for a $100M/year business is a pretty decent deal at this point...
And perhaps "bet" is the wrong word being used here. Usually with a bet in gambling terms, you either win big or you lose it all.
When Alphabet tries a "moonshot", they come away learning a lot about whatever problem it is that they were trying to tackle. They build expertise in-house, and they flex the idea muscle within the company. That isn't zero-value activity. In fact it's very valuable.
$1B a year is a small price to pay for what could eventually spurn a sustainable $10B per year or $40B per year business in the future (self-driving cars, for example).
Now this is just me being optimistic. It could also very well be the case that the X division is horribly mismanaged and the moonshots fail for more nefarious reasons, but I'm willing to give Google the benefit of the doubt.