I disagree, this is an article about Meta’s finance department being full of risk averse accountants and that being used as validation, when they shouldn't be
The IRS does enforce this credit
Moving beyond that, its super easy to rationalize how these are used for experiments
People post framework after framework every day here that eeks more performance out of the same inference hardware
That will satisfy the experiments over operations need easily
The IRS isn’t the authority when they disagree, the courts are. And in court you can tell the IRS to pound sand and tell reporters to take it up with Congress
-- Meta's accounts are not 'risk averse' - if they were, they wouldn't be claiming the credit.
-- ""People post framework after framework every day here that eeks more performance out of the same inference hardware
That will satisfy the experiments over operations need easily""
No, not even remotely.
The easiest way to measure would be 'training vs inference' - and almost all training is R&D and almost all inference is Ops.
If we had to do something crudely.
The implication that 'the product itself is a market experiment' is absurd.
That's not the kind of R&D that gets tax credits.
This is a very simple story: Corporate entities stretching the tax code in ways it was not intended, until the authorities tell them not to. The authorities are manipulated by politics.
Yes, the move causes 'concern' with the 'finance department' because it's probably illegal.
Surely they are themselves as individuals risk averse, they usually all are, I mean to imply this action is not one of 'risk aversion' obviously, it's the opposite.
Surely Zuck is the one pushing his CFO to do this.
Its not clear, the IRS will presumably decide if, for example, Zuck gets an R&D tax deduction from his salary because he contributes to code, which no other large tech company has tried to claim.