I don't think any Oxide racks come with GPU's at present, and the power density of modern GPU-centric AI compute is on a rather unprecedented level. Oxide racks are very well cooled but are no match for the racks in an AI datacenter that's literally burning a full gigawatt of power.
The overall sentiment of the discussion is *overwhelmingly skeptical and critical*. While a small minority of users defend Elon Musk’s track record of defying critics (citing reusable rockets and EVs), the vast majority view the "AI datacenters in space" proposal as scientifically unsound and economically nonsensical. Most commenters interpret the merger as a form of "financial engineering" designed to bail out underperforming assets (Twitter/X and xAI) using SpaceX's valuation ahead of a potential IPO.
The most robust debate focused on the physics of operating high-density compute in space.
* *The Cooling Problem:
* Numerous engineers pointed out that space is a vacuum and therefore an excellent insulator. While solar panels generate power, getting rid of waste heat from GPUs requires radiative cooling, which is inefficient compared to terrestrial convection (air/water). Users estimated that the radiator surface area required would be massive and structurally prohibitive.
* *Radiation & Durability:* Commenters noted that cosmic rays cause bit-flips and degrade electronics. Terrestrial hardware (like standard GPUs) would not survive long without heavy, expensive shielding, or would require "space-grade" hardening that lags generations behind in performance.
* *Maintenance:* On Earth, failed components are swapped; in space, a failed GPU or drive effectively bricks the unit or turns it into space junk.
* *Latency:* While Low Earth Orbit (LEO) offers better latency than Geostationary orbit, users questioned the utility of high-latency inference compared to fiber-connected terrestrial centers, particularly for complex AI tasks.
A significant portion of the thread analyzed the merger as a financial maneuver rather than a technological necessity.
* *The "Bailout" Theory:* Users widely believe this deal is designed to offload the heavy debt and losses from the Twitter (X) acquisition and the high burn rate of xAI onto SpaceX, which is viewed as Musk's most solvent and valuable company.
* *SolarCity Parallel:* Many compared this to Tesla’s acquisition of SolarCity—a move previously criticized as a bailout of a failing Musk-owned company by a successful one.
* *IPO Preparation:* Speculation suggests this is a play to juice the valuation of a SpaceX IPO by attaching "AI hype" to it, allowing early investors in the struggling X/xAI entities to cash out or convert their equity into SpaceX stock.
* *Conflict of Interest:* Commenters questioned the governance of private companies where one individual (Musk) controls the board and directs mergers that may benefit him personally at the expense of specific shareholder groups (e.g., SpaceX employees or investors).
### *Category 3: Economic Viability*
Users attacked the business logic of launching datacenters into orbit.
* *Cost of Launch vs. Land:* Even with the cost reductions promised by Starship, users argued that land and grid connections on Earth are orders of magnitude cheaper than rocketry.
* *Solar Efficiency:* While solar is more efficient in space (no night/clouds in specific orbits), users argued it is still cheaper to simply build more solar panels on Earth and use batteries than to launch infrastructure into orbit.
* *The "Million Ton" Claim:* Users crunched the numbers on Musk’s claim of launching "a million tons" of satellites, noting it would require an unrealistic flight cadence (e.g., launching massive rockets every few hours continuously).
### *Category 4: Musk's Reputation & Rhetoric*
The thread discussed Elon Musk’s history of promises versus delivery.
* *Skepticism:* Users cited a long list of missed timelines and unfulfilled promises (Hyperloop, Full Self-Driving by 2017, Robotaxis, Mars landings) as reasons to doubt the "space datacenter" timeline of 2–3 years.
* *Mockery of Language:* There was specific ridicule regarding the press release language, particularly the phrase "scaling to make a sentient sun," which many found to be "drug-induced nonsense" or "cultish."
* *The Defense:* A minority of commenters argued that betting against Musk has historically been a bad idea, citing the success of Falcon 9 and Starlink as proof that he can solve "impossible" engineering problems.
* *Jurisdiction Shopping:* Some speculated that moving AI to space might be an attempt to bypass terrestrial regulations regarding copyright, safety, or content generation (specifically referencing Grok’s lack of guardrails regarding CSAM/deepfakes).
* *National Security:* Concerns were raised that SpaceX is a critical US defense contractor, and merging it with a "chaotic" social media company and an AI firm introduces unnecessary risk and leverage over the US government.
* *Orbital Debris (Kessler Syndrome):* Users worried that launching millions of tons of disposable datacenter satellites would clutter low earth orbit, increasing collision risks and potentially locking humanity out of space travel.
### *Category 6: The "Why" (Strategic Speculation)*
* *Energy Arbitrage:* A few users attempted to steelman the argument, suggesting that if Earth's energy grid becomes the primary bottleneck for AI, space offers the only unconstrained solar power source, despite the cooling difficulties.
* *Vertical Integration:* Some noted this creates a conglomerate similar to Samsung or aggressive Japanese keiretsu, where the goal is total vertical integration of energy, transport, communication, and intelligence.
Stable storage would be limestone. To bring it down to pre-industrial levels it would mean that each person on earth would get a cube of 5 meters a side.
IDK, build houses out of limestone like we have been doing for ages.
Notoriously secretive, siloed Apple, where even internally, teams are said to be entirely in the dark about each other’s work? I think Apple, culturally, can’t do a public post mortem no matter how much they might want to. I would love to be proven wrong on this, because I would very much like to understand what happened.
The same Apple that reset a large number of iCloud passwords last year with no warning or notice, and no public acknowledgement or explanation? It was determined after to only have affected legacy Apple IDs that predated iCloud, but there was never any confirmation from Apple.
They absolutely SHOULD; but they absolutely WON'T because they don't even think they did anything wrong (as opposed to CloudFlare who hangs their hat on the mistake).
Companies commonly claim security/anti-fraud, then refuse to explain their actions, claiming (again, without evidence) that justifying themselves would help fraudsters in some way.
But really this has nothing to do with anti-fraud, and everything to do with duopolies out of control and weak consumer protections doing nothing to push back.
That's why Google, Apple, and Microsoft are notorious for this.
UK power grid has the Eastenders effect. Where the ending credits of the Eastenders soap signals a large increase in power draw from the grid as people will put on the tea kettle at the end of the show. The grid operators have to dispatch enough power to cover for this.
While the amount of energy used to boil water at 2kW is not significantly different from 3kW (2kW has a tiny amount of more atmospheric losses I think), there is a difference for the impact on the grid. Same energy but more power generating and transmission line capacity needed.
Here's a video showing an engineer at the national grid bringing hydro-electric plants online at the closing credits of a popular soap opera in anticipation of the millions of kettle that are about to be switched on!
And what he left out of this book (and included in the memoir or in some interview) was that there was a scientific study of women in the area at the time which discovered that a very high percentage of women had birthing complications serious enough for hospitalization that went untreated as they had to go back to their chores next day and there was no hospital anywhere close.
A few years ago I was running my UDM. Wifi on 2,4GHz and 5GHz was working fine. Then after an upgrade some of my devices start losing connectivity after a while, happened multiple times a week or day. Only restart helped.
I was going mad until I discovered that only the 2,4GHz ones are losing connectivity. Then I started researching and found out that it was not just me, it was just bad code in firmware that was not fixed in any of the versions that came out before I had enough, sold the device and swore off this garbage.
Nice, that's the optimized version - sounds actually a little different than the original one it's derived from. (actually better which I didn't expect)
Original: https://dittytoy.net/ditty/24373308b4
I like how music recognition flags it as the original Jarre piece.
I first did stuff like this when I was a teen using a 6502 machine and a synth card - using white noise to make tshhh snares etc. All coded in 6502. The bible was Hal Chamberlin's Musical Application of Microprocessors.
Then of course we had games abusing the SID etc to make fantastic tunes and then came very procedural music in size coded PC and Amiga demo coding that underneath the hood were doing tiny synth work and sequencing very much like dittytoy etc.
Shadertoy even has procedural audio but it doesn't get used enough.
reply