Tales From Nvidia: AI Is Now More Expensive Than Human Labor (Issue #200!)
Imagine a business that provided free humans to write code. Call that Business A.
Of course, every end customer would hire the free humans to write the code. They would review it is as productivity enhancing. Call these guys Business B.
Business A, if it had all of its losses funded (including by its suppliers), would grow like a weed, because the Business Bs would keep hiring the free humans. But, once Business A started charging a rate where they could make a profit on their humans, the Business Bs would stop hiring them and Business A would be out of business.
This is where AI is with respect to writing code. It may be OK at it (and even though it is better than it was, it is still not perfect and has issues and it’s mostly non-breakthrough production level code), but there are massive losses up and down the system to do it.
The provider (OpenAI, etc.) loses scads of money, their suppliers (the DC guys) lose tons of money. The only ones making money are the semiconductor guys, who also really do not make as much money as people think if you back out the cash they have to send out to fund all of their customers to keep the game going!
What if accounting rules required Nvidia ($NVDA) to count investments in their own customers as a cost of goods (expenses) and to amortize them into their P&Ls, how much money would Nvidia really be making?
The cash flow statement shows a very different story than their earnings do.
It is crazy what is being done — it is a way to move the discount off the P&L.
Anyway, the article below has it right. The free human analogy I used is pretty much exactly what AI is:
Position: None
