its the old joke about bandwidth versus latency. A suburban can hold an awful lot of 1TB uSD cards.

You really should do the math. 11 days, which is a still a long time.

The purpose of this type of storage is for use cases where the touch rate is very long. For data that is updated/accessed once a year in very big chunks, these work.

There is a JDEC spec for both consumer and data center SSDs. The consumer spec requires 1 year power off retention at 0% health remaining. The data center spec is only 90 days.

As the SSD wears, the retention goes down. Years ago, a tech magazine (remember those) did a torture test on an old MLC Samsung 3D SSDs. It was healthy with no errors at well over 10X the rated endurance. Unfortunately, the data retention was measured in days. This is why it is so important to watch SMART for wear and replace drives that are worn out. They might still seem error free, but they are not really storage at some point.

Other factors are temperature (cooler is better). Some have mentioned radiation, which probably is an issue if you are putting these in aircraft. Bottom line, if you really want the data to survive, replicate and keep it moving with good data integrity checks.

Fine, but the math gets a lot more complicated, especially when the voltage and current waveforms get out of sync (ie, power factor). But the conservation of energy still works with ideal components (which obviously do not actually exist). Also from a practical point of view, most 100 uF capacitors are electrolytic (or they are quite large) which does not play well with AC, but again the "ideal" part of the question does preclude this anyway.

I am not sure how a transformer would fit in with a DC capacitance / storage scenario. Transformers imply alternating current, and while they can be 1:1, I have never thought of a transformer as a DC component.

And yes, the AI explanation is not based on the basic formulas. That is actually the problems. Humans can sanity check an answer looking for conservation of energy. AI does not take this step. For these types of problems, a little recursion with some simple application of basic laws of physics should be able to push the "score" of a bad answer down to zero. But this is hard as a generalized solution.

sorry, my watt hours should be watt seconds. And one watt-second does equal one joule. So please explain the math here. moving from 100 V @ 10uF to 9.1V @ 110uF is exactly what you would expect. Just where does the 0.1X come from.

your question is not a paradox, you are just doing the math wrong. The initial state is 100 x .000010 = 0.001 watt hours. The final state is 9.1 x .000110 = 0.001001 watt hours. Remember the end point capacitance is 110 uF. So 9.1 volts make perfect sense.

Now what AI does with this is a completely different question.

I think the "developing products" in the summary is a big of a miss statement. I was at Flash Memory Summit (now Future Memory Summit) and a lot of the noise is other storage than NAND. Nothing yet compelling, but everyone is trying. A lot of this is that NAND is not staying still. A couple of vendors are likely to introduce PLC (5 level cells) as shipping solutions as early as next year, and for a lot of workloads, they are quite capable.

You need to do the math.

1. QLC SSDs are roughtly 4X the cost of HDD for a given amount of space.
2. While power is lower for SSDs, the cost savings does not even come close to the up front cost.
2a. The power for storage is literally in the noise compared to the GPU/AI compute power requirements.
3. The "other option" would be to use more older (ie smaller) HDDs, and even off-lease mining HDDs, not QLC SSDs.
4. NAND capacity is limited like HDD capacity in terms if exabytes. HDDs are still larger, although it is getting closer. Neither has the ability to ramp up quickly.

The people building these system do all the math. It is not knee jerk. They know every detail of reliability, power, space, availability, duty cycles, and multi vendor suppliers and take all these into account.

Several fallacies. As others have noted, how you define income is not trivial. If it is all the cash you receive, this means no one can ever loan money.

The other issues is that the simplicity is in no way related to a single rate. The rate can be progressive, and the method still simple. Don't try to push for one using the other argument. Classic straw-man.

Musk probably cancelled their hosting contract.

I just moderated you insightful, but then realized it was Tom Cotton, so redundant would have been more appropriate.

Sorry my moderation will go away because of this comment ;)

It is amazing, but your numbers are off.

50% is low. Perhaps really low.

The coders are some volunteers, and some companies that volunteer their employees. Still a community.

Some of the edge cases can get a bit scary. I tested a piece of iSCSI/ROCE code for a developer because he did not have the hardware to test it himself. It was a simple case, and the code worked, but the idea of a developer without a test-bed is at least "interesting".

So, still amazing, but like most things, more complicated.

The least disruptive is to just ditch vi and keep the current code. VS Code remove over ssh might be a good option. Unfortunately, the choice of editors is often more of a religious discussion that a technical one.

My math on this is 400 MW of solar, but with the typical over the day spread. The batteries can either charge or discharge up to 400 MW, but can hold 1600 MWh.

In that the panels and batteries are both at the same site, losses should be reasonable (and might already be built into the number). This balance should let the "combination" run pretty much one deep cycle per day most days:

In KWh:

1,600,000 x 300 x 10 = 4,800,000,000 or 4.8 billion KWh, which works out to $0.21/KWh.

Add in some incentives, plus the possibility to expand on the same site, and it makes a lot of sense, especially in that the power it sells is "peaker" power, not baseline like nuclear.