Musk Predicts Tens of Billions of Humanoid Robots and ASAP Terawatt in XAI AI Data Center

11 thoughts on “Musk Predicts Tens of Billions of Humanoid Robots and ASAP Terawatt in XAI AI Data Center”

1. 10s of billions of humanoid robots? For what? I keep trying to think of use cases in my and my wife’s modest home in NYC.
   First, where would we put a humanoid robot in our ~900sf home, already full of furniture, appliances etc. and stuck with 1960s electrical outlets already multiplied with powerstrips for all the electronics that didn’t exist in the 1960s when the building was built in NYC. BTW, there’s been a huge controversy over battery fires breaking out from faulty e-bike batteries being charged overnight – often multiply. I presume that $10K robots will have safer UL rated batteries, but they’re still going to draw a lot of power from a 15-20 AMP fusebox. And very few people will pay even THAT much – $10,000 – and we are a long way from anything useful at even that price.
   We both work from home & have to make ourselves get out of our chairs to exercise, get out, maybe do a little food shopping, laundry and for me: bike riding, outdoor gyms etc. Which of those would be:
   A. easier to accomplish with a humanoid robot,
   B. worth replacing, given that our health would suffer from not being active enough and we’d have to add more exercise to our routines than we have already?
   Food shopping? That’s what Amazon Prime monthly deliveries are for. Maybe Amazon will replace their driver with robots, but just for loading & unloading to lobbies? Hard to imagine, but that’s not billions of robots anyway.
   Laundry? Not as simple from basket to putting clothes away folded, ironed (?) in the right place as it seems. Is that worth it?
   As for robots in factories, we already have them – and they have one arm apiece, operate in the dark, doing the same repetitive tasks, 100s of times a 24-hour day, never needing to go anywhere to recharge, fixed in place. If there are tasks these custom robots can’t perform already, it’s hard to see how cheap multipurpose humanoid robots will do them better. Foxconn tried replacing its deadened, depressed sometimes suicidal million plus employees with robots and was only partially successful. Apparently, getting robots to tighten “all those little screws in an iPhone” – as Howard Lutnik put it – is not so easy.
   Humanoid robots on Mars to prepare for human habitation? Yeah, I can see something like that, but the big stuff is more likely to be performed by automated tractors and cranes.
   None of this needs 10s of billions of robots. Our mining and refining industries and electrical grid operators can relax a bit, and just worry mostly about AI server farms, for now, at least until the Deepseeks of the world greatly reduce the need there too.

2. Just testing whether all my comments are being moderated off simply for expressing a little bit of healthy skepticism.

3. One terawatt mentioned in Saudi can only be solar, so for about $250~300b at current prices that can be arranged. Won’t work at night though, and requires about x10 installed capacity to make a more or less stable terawatt. Point is, and someone has to say it, why is Musk still not on the nuclear theme? He needs it everywhere: on Mars, in space, on Terra, everywhere. Everyone and their cat have already converted from nuclear haters to nuclear fans. And yet, nothing notable from Musk. As with rocket motors, he knows where to buy advanced nuclear reactor tech to start with, and save ~50 years of R&D. SpaceX has not achieved its current capabilities just by being vocal proponents of such things.

   • Nuclear takes a long time to deploy. It is also more expensive than solar + batteries. In a different time line, the reverse might have been true, but we are in this one..

     • At ~42 months per typical 1GW+ unit, it is not such a long time, and with 60~80 years of active life (with current trend of extensions), there is nothing better even based on total “time to deploy” metric. With well-developed fuel cycle (as Russia has), the life cycle cost of energy of nuclear is much better than “solar + batteries”. With half-assed fuel cycle and rampant bureaucracy, you would be right: bottlenecked and expensive enrichment, ever problematic and expensive fuel storage, and endless construction (bonjour, EDF!).

       The real difference is the minimal tech level at which nuclear becomes the best option. Some have it, most don’t, hence all the papers on how “solar + batteries” magically provide better energy price than a line up of gigawatt units at 90%+ utilisation over 60~80 years life cycle. It is only better if there is no real choice.

   • “why is Musk still not on the nuclear theme?”

     Well, originally two reasons:

     1: He didn’t want the Greens after his throat.
     2: The regulatory burden was just absurd.

     #1 is kind of moot now, but #2 is still in play at the moment.

     • The regulatory burden will hopefully continue to be absurd.
       -nuclear waste and stolen/lost fuel is an environmental nightmare of a threat that has not proven to be manageable let alone allowing nuclear to proliferate. Now image how lost/stole fuel and fuel in less attentive countries is disposed of.
       -new option in fusion and geothermal on near horizon. Solar and batteries are a better option in the meantime.
       -cost overruns are hard to control
       -remain vulnerable to attack or accidents of significant environmental damage.
       -cost of military protection
       -a dumb vulnerable steam engine of the 20th century. Thorium MSR may be a better option but talking nuclear in general doesn’t discriminate.

       • nuclear waste is unburned nuclear fuel. It is 95% Uranium 238 by weight. Faster neutrons generated in reactors can use uranium 238 as fuel.

         About 65,000 tons of uranium 238 has to be stored every year. These are usually in casks stored onsite.

         The other energy sources
         coal – particulates and pollution from burning 8 billion tons of coal goes by the billions of tons into the atmosphere. A lot of this material ends up in peoples lungs. This and the air pollution from oil and natural gas accelerates the deaths of millions of people every year. There is also water pollution, mining deaths and deaths from transportation of billions of tons of material. 40% of rail traffic is moving coal. Hundreds of deaths moving a large amount of material.

         Nuclear waste is an imagined environmental threat but coal, oil and gas air and water pollution are actual deaths and health impacts. Bad air days cause a surge in hospital visits from those with asthma.

         Solar power uses more land. Solar on roofs causes more deaths as roofing is the 5th most dangerous profession. Solar and wind use ten times the steel and cement for the same energy generation.

         Nuclear energy has lower deaths per terawatt hour.

         Hydro dams alters the environment directly. Lakes are created where there was no lake before. Hydro dams have a history of deaths. China dam break killed 250,000 people.

4. As a electrician, my Turing test is for a humanoid bot to climb a ladder, and change a light bulb, by voice command, and me pointing to the bulb I want changed.
   Once that is achieved, that’s the first step in them taking my job.

5. I’d like to see Optimus demos using various tools – including driving cars and various equipment that’s operated from a drivers seat. It would be kinda fun to see a classic taxi with an Optimus bot behind the wheel wearing a drivers cap, or some classic car as a chauffeur. I assume that FSD ports to Optimus since both run on the same hardware but there’s probably a lot of work to adapt it.

   There is a whole vast sector of work vehicles and human operated machines that need to be either reinvented as robotic or operated by humanoids.

6. Tap dancing robots were inevitable.
   I look forward to the humanoid robot riding a unicycle while spinning plates on 10′ poles.

Comments are closed.