phil beisel

@Out_of_Spec I am sorry for your loss.

Ellie Passed Away youtu.be/2np5Fi_CNO0?si…

@Andercot Oh new Apple watch bands that look like satellites. Cool.

The best way to think about orbital data centers is this: "How many additional human-equivalents of intelligence does this add to the planet" A human brain is 20 Watts. An ODC is maybe 200kW. If computers are 1000x less efficient than humans, 100TW is five billion brains

@Harsch_Reality @CernBasher @APompliano I might suggest these could be different orbits, outside of Earth/sunbelt, since they would be long running async. processes (don't need much communication with Earthlings). Its just an economic question in the end, no?

@CernBasher @pbeisel @APompliano had a guy on his show about 9 months ago proposing bitcoin mining in orbit.

SpaceX Musk compensation incentive goal: 100 TW "operates data centers in space that provide at least 100 terawatts of compute capacity." Remember the famous Terafab chart showing 1 TW labeled "Expected compute demand from Tesla + SpaceX"? The Musk milestone is 100 TW. Have they gone insane? Actually no. We could hit 100 TW of demand in less than 10 years. Today global AI data center power capacity is about 30 GW. Let's check this against growth in the CPU growth era, 1980-2000's. The growth rate then was 58% per year (growing at 1.58x per year). By that measure, 30 GW will grow to 1 TW in 7.7 years (2034). And 100 TW at 17.7 years (2043). BUT, AI compute demand is growing much faster than the CPU-era growth. About ~3.4× per year. That’s more than twice as fast as the historical 58% rate. So the demand for 1 TW equates to only 3 years and 100 TW only 7 years. Even in the most optimistic case (e.g., 1.5× FLOPS/W efficiency gains every year from Blackwell/Rubin-class chips), we still hit 1 TW in ~4.3 years and 100 TW in ~10 years. And I think inference compute demand is going to accelerate from here!

@robcox23 @Speculator_io Very well.

@pbeisel @Speculator_io Think Google will do well tho 😉

Google was one of the most obvious winners of AI. It is the only company that owns the entire AI stack. 1. Chips: TPUs 2. Infrastructure: Google Cloud 3. Foundation Models: Deepmind 4. Applications: Gemini Every layer matters. The more layers you own, the bigger your advantage. And all while their core business prints cash. Not to mention: + Physical AI: Waymo + Ventures: Anthropic + Endless Proprietary Data

We are leaving the solar system and cruising around the Kuiper Belt now. Soon we will drive past Voyager I.

@CernBasher @rocktstl Agree, its about shifting industrialization off Earth. Compute + industrialization.

@rocktstl @pbeisel I'm not so sure we need to worry about getting the energy back to earth. In my opinion, what's more likely is that we will bring the manufacturing/industrial processes to space - particularly once we're getting raw materials from the moon & asteroids.

@CFlittard1904 @CernBasher @aaronburnett @teslayoda One day at a time.

@pbeisel @CernBasher @aaronburnett I just finished @teslayoda's book, Chip By Chip. It illustrates in great detail how difficult the Terrafab will be, by illustrating how much TSMC's culture is close to impossible to copy.

Do we remember when Elon said Google wins the west, China wins the world, SpaceX wins the rest? AI is consistently showing that vertical integration wins. Cursor on top of SpaceX is the Ai stack bet for the 2030s. The cheapest and most scalable access to the chips and energy. The only thing that stops this is dramatic slow down in demand for intelligence. Over the history of humanity intelligence demand seems like it only goes in one direction. Save maybe the dark ages.

@CernBasher @aaronburnett EXACTLY what I was thinking. And in the case of SpaceX another layer gets added, the fab.

@aaronburnett I love it when people completely ignore SpaceX/xAI when they put tables together (like the one in the post). The dark ages of analysis! Look beyond your toes and into the heavens...

@disruptivedean Right, non-AI DC are excluded.

@pbeisel The AI power is alongside 100GW of existing cloud or on-prem compute power Not all DCs are just used for AI

@markgurman Do you mean increased capex?

Power On: Apple has dropped the first hints about how John Ternus will run Apple, starting with holding onto more cash and paring back how much is given back to shareholders in the future. bloomberg.com/news/newslette…

@Maggie1728Q I've had 3.

@pbeisel This is one deserves an extra espresso ! ☕️😊

Today, May 3rd FSD (supervised) hit 10 billion miles. It only accelerates from here, massively. tesla.com/fsd/safety

Talking "Teraspace" with @RandyWKirk1. My interpretation of the SpaceX master plan. youtube.com/watch?v=6GI4vu…

@surfranchvibes No

Tesla community - do we think this is true

@JCChristopher I wouldn't even want a robot to deal with some people.

@pbeisel Tesla should run this as a Superbowl ad for Robotaxi 🤣🤣🤣 jk

If you’re an Uber driver and your gut is telling you to cancel and drive away, don’t 2nd guess your gut. Just cancel and go. And apologies to all the ladies out there who I know have to deal with some crazy Uber drivers. But I would recommend not saying the “K word” to a male Uber driver before you even get into his car. Also, don’t put your purse on the Uber driver’s arm and start dropping the contents of your purse on him. And yes, Robotaxi fixes all of this. There you are Tesla bulls, are you happy 😂

@David_Ogawa @RobertJThomas1 I think they did that on ST5.

@RobertJThomas1 @pbeisel It's not just the whale. You also need the water surrounding the whale.

T-11 days and counting On May 12, SpaceX will launch the first-ever Starship V3 from the new Pad 2 at Starbase. First time a version 3 (block 3) Starship is being launched. The key goals are testing the upgraded vehicle, attempting a Super Heavy booster tower catch, and a controlled Starship reentry with Indian Ocean splashdown. Starship V3 is taller with much larger propellant tanks than V2, giving it dramatically higher payload capacity. V3 is powered by advanced Raptor 3 engines: lighter, simpler, and significantly more powerful than the Raptor 2s on Version 2. The V3 booster features a built-in hot-staging ring and three larger grid fins for better staging and control. Bottom line: V3 aims for 100+ tons to orbit reusable, a massive leap forward from Version 2.

@David_Ogawa @WizZifnab Sounds reasonable, I've heard otherwise. But I would bet with you on that.

@WizZifnab @pbeisel The booster catch process is designed to handle diverting the booster to a crash site. But the V3 booster design/fab needs one test flight hovering above the ocean surface as proof. Every new block design will likely be expended at least once.