I wouldn't say a overall loss leader, but definitely at lower profit margins than Intel. Like modern technology, CPUs require a huge amount of engineering to design and test the product.
My understanding is that for Intel, on a whole, these non-recurring engineering costs are about the same as the direct cost to make a CPU. Any company has to recover all of their costs, or else they will go out of business eventually. They also need to make a fair profit. How these costs and profit are divided up among market lines can be varied.
For nearly all tech companies, including AMD, Intel, Nvidia, Microsoft, etc., these costs are put more into the high end. For example, if you're Amazon and a server is bringing in $50,000 of revenue over its life, you can afford to pay $1000 of costs and profit to Intel. Whereas, if you're a home user, you don't want to spend more than $800 on a computer, you pay a much smaller share, maybe $100 on a $300 CPU.
What if you divided it up evenly? Amazon would love to pay $550 instead, but simultaneously, as a home user, you wouldn't buy a $750 CPU. So Intel would only be able to get $550, not $1100. They'd have to make a crappier product, or go out of business.
However, you're getting the same CPU core, which is why when you divide speed by cost, the higher-end CPUs seem less optimal.
Companies also want to recover their costs as soon as possible. You want the cash back quickly to reduce uncertainty and be ready to invest in something else, the next generation.
This all gives companies some knobs to tweak. Do you recover the costs quickly, charging more, or slowly? How much of the costs do you charge the low end vs high end? By accepting a lower profit, spending less in R&D, taking a longer return on investment, and adjusting the cost distribution, you
In spite of some of the experts and pros in this thread, I know people that work in film and animation for large studios and they are already planning on getting rack mounted version for their workflows. This is not a product for your home office. Its for companies that get handsome tax deductions spread over decent depreciation schedules that need a reliable workhorse.
Can you build a faster benchmark machine for a fraction of the cost, probably. But would you really put it in an environment with scores of expensive employees that charges in tenth of an hour intervals?
In 10 years i want one. By then it will still be ridiculously overkill for anything i can throw at it... But it would be fun to have. The case itself will still be an engineering marvel, even if tech has moved on.
