Well, how much is a robot? I’m assuming they are approaching a respectable six figure mark at this point. Let’s assume a unit pencils out to 100k today. The poor guy having to pick up temp work moving those boxes makes what, $15 an hour? Roughly speaking, you have 6,666 hours before he costs more than that one auto. Divide that by 40, then 8, supposing he works full-time for the holiday season for the whole eight weeks. You’re talking 30 holiday seasons? At that rate, the big movers will still be using humans for the foreseeable future. It’s too cheap. I think robots are a fad/way to keep attention on inflated stocks

---

EV enthusiasts pop these numbers out for every make and model that comes along, but the reality is that 10k unit sales per year is not a pathway to profitability. They need to sell ten times as many per year to offset the R&D of battery infrastructure, but those days don't appear to be on the horizon. EVs are fabulous economy cars. Trucks are not economy vehicles. It's just a bad fit.

---

Tesla creeping toward that 400 PE, with declining unit volumes. Complete and utter fraud

---

Numbers for the current Jags: unit production ~$200k, units pay for themselves after 18-20 months of operation

---

How much does a Waymo cost to produce per unit? What’s the annual cost to store and operate each unit and how many rides does it take to break even? I’m listening? Tesla bears can’t do math 

---

I think it’s an overbuilt bubble but the data is misleading. Older GPUs, H100/A100, will rent out cheaper over time as newer units, H200/Blackwell, are put into service. You need to compare msrp against rental rates at release for each unit and use that ratio to compare. There are more factors at play but that at least compares apples to apples.

---

Here is a few live examples of what they're going for. https://www.hyperbolic.ai/ h100 at 1.49 / hr. https://vast.ai/ - 1.65 Why do you keep bringing up hyperscalers? The market price is being driven by every other company who doesn't have lines of business to obfuscate the usage numbers. This gives us a better idea on demand *outside* the circular AI industry - it's important to ask - ARE the purchased cards being used to such an extent that they command higher sold out rates just 1 level down from Nvidia? But let's do some math just on the h100 because you doubt both the full report and the article. Per gpu H100s are around 25 to 30k on a single 80gb unit in 2024/2025 https://docs.jarvislabs.ai/blog/h100-price It CAN go up to 40k on some configs but let's use the middle band for average so 30k. https://www.clarifai.com/blog/nvidia-h100 Let's assume you want to have 70% utilization over only 3 years (NO one gets 70% but once again being generous) 3 years ≈ 3 × 365 × 24 ≈ 26,280 hours. At 70% utilization you rent ~26,280 × 0.7 ≈ 18,400 billable hours per GPU To just get your $30,000 capex back Required revenue per hour ≈ $30,000 ÷ 18,400 ≈ $1.63/hr. This *doesn't* include actual operating costs like power and cooling etc. Cooling, networking, rack space, maintenance, staff, overhead: easily pushes true hourly cost meaningfully above the capex‑only of 1.6/hr. Realistic datacenter overhead end up with breakeven more like $2.5–3.5/hr per GPU over a 3–5 year life at high utilization BUT that also works on alot of factors because datacenters can vary in cost. The only reason I track these is I want to see IF this apparent record level demand is sticky and goes beyond just Nvidia selling out - is it TRUE demand (i.e. deep and durable end user demand) or scalp demand. All signs point to the latter.

---