Another problem with it is that food isn't as liquid as cash. I can't buy a car with their food. I can't buy a TV with their food. I can't even buy groceries with their food. Presumably, yes, I am going to have to buy some food anyway. However, perhaps I would prefer to buy something cheaper, and use the money for something else. Or, perhaps I would prefer to buy something more expensive, and not have to worry that I've already paid for Google's food on top of it. Or perhaps I just want something different. Saying, "We're going to convert $X of your salary to an equivalent value of our food" robs you of choice. I'd want a deep, deep discount in exchange. At least 50%. And no, I wouldn't accept an inflated number like $20k just so they could discount it to $10k. And yes, I'd get this perk in writing, including availability hours.
You're making it more complicated than it needs to be. When you're valuing a perk, the cost to the company should be ignored entirely. Any perk should be valued at what it's worth to you based on how often you'll use it, not what it costs the company.
You should definitely ask questions to see if there are reasons you wouldn't use it, but attempting to put that in the contract isn't going to work. There's no way a big company is going to complicate their facilities planning by making separate agreements with employees. Better to just assume it won't last forever.
Personally, not having to either cook or shop for dinner if I don't feel like it is worth quite a bit to me. To someone who goes home and eats with their family, it's not worth as much. I'm not going to put a dollar amount on it because I think that's just adding false precision to the calculation.
By the way, studies show that people often don't value the right things when they do a detailed calculation. For example, a short commute is more valuable (in terms of happiness) than an extra bedroom on the house, but many people will take the longer commute if they overthink it. [1]
