@yhxxl0810@YounghoYh_PARK 인간이 비전위주로 운전하는건 카메라 방식을 지지하는 증거 맞습니다. 그런데 달성 목표가 인간 이상의 운전일 가능성이 크고, 거기에 도달하기까지 시간과 자원의 총량을 고려해야합니다. 테슬라는 영원히 도달하지 못할 가능성도 큽니다. 혹은 도달해도 엄청난 연산력을 동원해야할수도 잇네요
@jiahanjimliu Why is that such a terrible deal? If licensing helps more models grow, isn’t that good? Kimi seems good enough to justify even a 30% share. Jim, your perspective seems to be getting narrower.
@babyfolio Do we need to explain in more detail why DRAM is more widely preferred than SRAM, and why the chips are deliberately partitioned and clustered?
@jiahanjimliu If a Rubin rack delivers 6× the token throughput while costing 2× as much as a GB300 rack, it would produce roughly 3× more tokens per dollar invested.
@zephyr_z9 If a Rubin rack delivers 6× the token throughput while costing 2× as much as a GB300 rack, it would produce roughly 3× more tokens per dollar invested.
@CryptoBankz_65 It’s almost essential. Without upfront payments, raising capital would be far too dilutive. On top of that, to ensure stable revenue, at least 30% of the capacity needs to be locked into long-term contracts.
@RealJimChanos@MelvinInvests You’re a lazy old man. Go out into the market and check what it actually costs to rent GPUs today. You can do that, can’t you? Whining that there’s no evidence in a two-month-old financial statement just means you’re either too lazy to look or too afraid to admit you’re wrong.
@RealJimChanos@MelvinInvests Why are you ignoring the unit economics the company presented in Q2? If you believe the AI compute market is turning bearish, provide evidence. Simply doubling procurement costs and halving rental rates doesn’t make the argument stronger.