World’s smallest violin. Let’s break it down:

• Hardware - all paid to providers and more prominently Nvidia;
• Software - all the statistical relationship and logic was developed by handsomely paid staff;
• Input data - there’s no such thing as copyright, intellectual property or any sort of mechanism that prevents harvesting copious amounts of data that was created, refined and delivered as part of human experience or a business product. It’s free for all to take, why pay for data?
• Output of LLM - Based on the preceding paragraph, it’s free for all to take, why pay for data?

So, competitors can’t avoid the hardware costs but can save on developer costs? Nobody paid for input data anyway. Sounds like a VC’s wet dream.