Comment by Betelbuddy
2 months ago
It seems we will run out of hardware by March?
"Hard drives already sold out for this year" - https://www.theregister.com/2026/02/20/ai_blamed_again_as_ha...
Time for an AI tax on the hyperscalers.
2 months ago
It seems we will run out of hardware by March?
"Hard drives already sold out for this year" - https://www.theregister.com/2026/02/20/ai_blamed_again_as_ha...
Time for an AI tax on the hyperscalers.
> It seems we will run out of hardware by March?
What happens when an unstoppable force (building everything in Electron because hardware is cheap) meets an immovable object (oh no hardware is expensive now)?
We go back to the demoscene days, being creative with what we have instead of shipping Electron junk.
Inshallah
1 reply →
Maybe we need to let go of our auto-scaled 100 pod service mesh for a todo list app, and just deploy it bare metal on 2 servers.
You joke, but I remember seeing a talk by Wunderlist CTO who has pretty much that. Also polyglot company and microservices in random languages. Can't find the talk now, but https://www.infoq.com/news/2014/11/gotober-wunderlist-micros... mentions 60 services at least.
1 reply →
I guess we have to get creative again.
I actually think you're right here.
Resource constraints have often helped me come up with stuff that I'm actually proud of.
1 reply →
consumer RAM is not what's creating shortage. Data centers doesn't run electron to train the model or for inference
Sure, consumer ram isn't causing a shortage, but it's affected by the shortage.
They effectively do. They’re trained by brute forcing 100TB of training data through them, rather than any logical learning technique.
A human doesn’t need 100TB of books to learn the alphabet.
10 replies →
Every RAM producer is stopping their consumer grade RAM production to provide ECC-RAM and VRAM now. Micron discontinued and closed down Crucial brand as a whole.
So, getting systems with higher RAM capacity is getting harder (from laptops to smartphones). So, for a couple of years, we need to stop using Electron so much and use what we have efficiently.
Data centers, esp. AI hyperscalers do not care about efficiency for now, because they can suffocate consumer-grade part of the hardware marketplace and get anything and everything they want. When their bubble pops, or the whole capacity ends, they need to learn to be efficient, too.
For reference, a well-optimized cluster runs at ~90% efficiency even though they have thousands of users. AI hyperscalers are not there. Maybe 60% efficient, at most. They waste a lot of resources to keep their momentum.
6 replies →
The problem is manufactures are shifting their production capacities towards the more profitable, high-performance "AI" components.
Stop using Electron to save massive amounts of RAM.
its not easy choice if you want mature crossplatform UI framework.
3 replies →
Everything gets more expensive?
2026 will be the year of Rust...
Due to lack of memory leaks which will stop increasing RAM prices?
9 replies →
[dead]
Why tax something that the market will figure out? This is normal and things will sort themself out.
Markets only "figure things out" in a petri dish economy where:
1) There are no barriers to entry for competitors (e.g. protectionist tariffs, equal access to capital for everyone)
2) There are perfect substitutes available, so transitioning to a competitor is seamless and free (e.g. no requirement to store data in Country X, no vendor lock-in, no security compliance)
3) The industry is not a "natural monopoly" when only a handful of vendors can operate due to capital investment and national/global distribution required (see power utilities, telecoms, petrochemicals)
4) Profitability attracts competitors (won't happen because of #3), but heavy competition prevents abnormal profits from accumulating to a single player (happens because of #1, #2 and #3)
When markets don't figure things out, as is the case around the world, you get a tangled mess of market failures, government intervention and lobbying to neuter proposed interventions.
Markets are never perfect but over the course of history they are a pretty good mechanism to solve these type of problems. Not sure why we think taxing hyperscalers differently is the answer. Government usually does worse than the market when it comes to sorting it out.
My argument is not that market is perfect but that the alternatives are probably far worse, like a new tax on a specific group of companies.
7 replies →
We see that market is very irrational now and it can stay irrational for long enough to destroy everything we know in tech.
By the time market figures things out, you may no longer have services, and hardware that you use daily. When such amounts of stupid money are pumped into a single industry, even if all AI companies went out of business tomorrow, it's going to take years for things to go back to normal.
FWIW, I'm not advocating taxes, as I think that won't really do anything. I don't know what the solution is either.
Sounds like hyperbole. Yes the world is connected yes we are seeing shortages, yes the market is imperfect and it lags but this is how things get fixed. Prices are sorted out, manufacturers make bets on long term capacity. Some will be losers, some will be winners.
10 replies →
> ...and it can stay irrational for long enough to destroy everything we know in tech.
Nah. For decades software engineers have been more expensive than the cost of buying the extra hardware needed for vastly inefficient software. There are orders of magnitude of inefficiency there. So there's a ton of slack in the world's software that can be taken up by software engineers while hardware is scarce, pushing back the date where there will really be a problem probably by decades more.
Of course software engineers will see a problem though, because they'll have to learn to to write efficient software again.
ie. "Great, but now make it work with less RAM" will be a thing again, instead of "It needs more RAM so order some as it's cheaper than your time to fix the code".
What we're seeing is the natural conclusion of VC distortion in a market. There is so much money being pumped into AI speculatively now that it's hurting normal and sustainable businesses in other parts of the economy.
The solution might have to be mandatory rationing of some kind to avoid a situation where only a handful of AI giants are able to buy essential components. We can't just throw the rest of the economy under a bus to support the AI bubble for a few more months.
I'm working with a business right now that would like to buy some new servers for sensible, boring business reasons. It is having trouble because the prices from their normal suppliers are now extremely high - if the components are even available at all. This business has nothing to do with AI or Big Tech and yet it's at risk of being unable to continue normal operations in much the same way that a business would be affected if the phone networks were all switched off or the water supply to its office was cut. We regulate those industries because their continued reasonable operation is essential to make sure everyone else can continue to operate reasonably as well.
4 replies →
> it can stay irrational for long enough to destroy everything we know in tech
What does this even mean? I know people on the internet sometimes exaggerate, but I cannot even begin to find a more charitable meaning with this, what exactly will be "destroyed" in "tech" because of prices going up for a year or two?
10 replies →
> We see that market is very irrational now and it can stay irrational
That meme refers to speculation on stock market prices. Nobody is buying up RAM with the expectation of making speculative gains on it.
People keep parroting this but I don't see GPU prices sorting themselves out.
“Parroting” is not very constructive language but I will respond.
How would you propose solving it? My opinion is that government cannot solve the problem better than the market. That’s not to say the market is ideal or perfect but one of the better tools available. GPU prices might stay high for a number of years. I don’t think that is inherently bad. Constraints breed innovation and help guide market participants into the right direction.
I think folks often get hung up on the market thinking it’s a perfect tool or it will realign issues instantly. That’s not true. Demand is high, price catches up and eventually either that thesis behind the demand is correct and eventually supply increases or that thesis is proven wrong and demand collapses below original baseline.
Obviously that’s a simplified version of it above but I don’t know what folks like yourself are poking. How is a tax on hyperscalers effective? I suspect most folks repeat this idea because they are in the anti-AI camp. Should we tax EV manufacturers because they may be buying up battery supply? I don’t know if I want the government making those kind of decisions.
4 replies →
They didn’t the last two memory crunches. Litigative action figured it out first.
Price fixing is not a free market, and that’s (probably) not what’s happening this time. It’s simple supply and demand.
Because this perfect version of capitalism you think exists, doesn't.
We live in a world with markets dominated by cartels of tech companies who don't play by the rules. Every other industry that impacts society in a negative way typically pays some sort of specialized tax to offset that, I don't know why these tech oligarchs shouldn't have too. It's wild how people just want to let them do whatever they want.
Everyone says we need to deregulate tech, and certain industries to get ahead of China.. Isn't it funny how their largely government controlled economy (to a degree) is annihilating the west on all fronts economically. We need far more regulation.
China will defeat the West solely because it regulates its billionaires, not the other way around like we have it in the West. And I hope so, the world is rooting for you China.
Way to put words in people mouths. Markets are imperfect but I do believe on average they are one of the better tools to solve supply and demand issues.
I don’t know who will come out winners but I do agree that China did well taking the playbook from Singapore and navigating their country through incredible amounts of growth. They are still facing depressing housing prices and deflation in other parts of the economy.
There are absolutely areas where markets breakdown, thinking problems where impacts are on longer horizons but for simple supply and demand like what we are seeing today, things will sort out in a couple years.
Reads a bit like the Paperclip Maximizer appearing way ahead of schedule? Implemented not as AI, but as emergent behavior in the ways of the financial class (that happens to be about AI, singularity and all that).
> Time for an AI tax on the hyperscalers.
Like the purchase price + increased cost? The thing is that these parties are sitting on billions and billions of investor money, they don't care that hardware is 400x as expensive. Which companies like nvidia have capitalized on a few years ago, they were already able to price their hardware at a 400% markup compared to pre-crypto times, and shift their focus from consumer graphics chips to datacenter compute chips, causing their revenue to go up 6x (if my interpretation of [0] is correct)
[0] https://www.macrotrends.net/stocks/charts/NVDA/nvidia/revenu...