If you look at valuation of a company like Micron (MU), it's pretty clear that the market expects a collapse in their profits. They are growing at absolutely insane rates, whether you look at YoY (well over +300%) or QoQ (over +40%) in both revenue and profits. Yet their PE is below 10. SK Hynix is similar.
A typical company that trades at a sub-10 PE is a flatline-to-dying business whose last juice is being squeezed out before the inevitable sale of remaining valuable assets and dissolution.
Of course the current market price is not a perfect predictor of the future of a business but despite all the memes and anecdotes, the market is far more often basically right than dramatically wrong (it's just that when it does get things dramatically wrong, that's what gets all the publicity, for obvious reasons).
So, assuming the market is roughly right in its valuation of Micron or SK Hynix, either demand will collapse or supply will surge or, some combo of the two. Either way it seems extremely likely that within a couple of years at most the prices of DRAM and NAND will come waaaay down.
The market as a whole is irrational [1] and, in general, doesn't know anything that we here don't know. In fact, the market typically reacts to things in a much baser and less analytical manner.
Micron and SK Hynix are at sub-10 PE because they are run by people who are poor salesmen and don't know how to drive or ride a hype cycle. If they focused less on engineering and more on PR, media outreach, "socials," performative patent applications and gimmicks, they'd be trading at double their current valuations, or more.
It would also be worse for society. Puffery is rewarded. Thomas Carlyle's example of the seven-foot lath-and-plaster hat paraded through 19th century London. The memestock. Crypto.
[1]-Does anybody genuinely believe that Google's worth 3.43T and Anthropic is worth 1.5T? This goes beyond irrationality and into lunacy...
Google expects to make 200 billion in profit in 2026, so their PE ratio is somewhere between 15-20. You can argue if that profit is sustainable over the long term, but a PE ratio of 15-20 does not seem like lunacy to me.
It's trivial to make bets that the market is dramatically wrong by using long-dated options. These are multi-year-long bets at high leverage (i.e. you bet relatively little and win relatively a lot if you're right). If you are extremely confident that you are right and the market is wrong, you could use these and become rich. I would not recommend it -- the market is generally way better at valuing businesses than us casuals.
Edit: FWIW, Google at 3.4T seems almost a bargain -- not that I would go all in on it instead of just keeping my broad index funds. I'm less sure about Anthropic at 1.5T
It's the ratio of Google to Anthropic that is off. If Anthropic is genuinely "worth" 1.5T, then Google should be much greater. They have everything Anthropic has, and a veritable mountain of assets that Anthropic does not have. But the vibes-based market doesn't price it in.
I'd much rather invest in market fundamentals as an individual. Many billions are poured into (and made on) vibe-based market predictions. It's much safer to go after smaller returns on well priced stocks than play predict the market irrationality against behemoths. After all, it worked for Buffet.
My point, which admittedly could have been clearer, was that the ratio between the two companies makes no sense. Think about what we know Anthropic has, in relation to what we know Google has.
Both have data, both have AI. Google is more established and likely has way more data broadly, but the perceived number 1 in AI is now Antrophic. I never payed google 200$ a month (and always used adblockers), but now I do pay so much for Antrophic and for me it is very worth it. And governments and companies seem to think likewise.
So basically also when it is about the ratio, it does not seem so irrational to me. In the big picture with all data in the open it might, but from my limited point of view it is definitely not obvious wrong.
The market has participants with varied and contradictory motives.
Main street investors rely on stocks to produce returns.
Large investors are willing to invest at a loss, if the investment provides control / access to a company which may augment their other investments. Sure, your $100M investment in Anthropic may lose 50% of its value, but if you turn a blind eye to it, then you may lose 100% of your $500M consulting business. The $100M investment may also include information that maybe is or isn't "insider", but maybe it's tailored to your needs and allows you to cut so much in costs in your other businesses.
(Numbers are probably grossly out of scale)
Some are paying to own the company, not to get a share of the profits.
Your guess at those valuations is as good as anyone’s. Turns out if you put lots of such anyones in a room it makes a market. If you doubt the market you doubt the wisdom of the crowd, which is very smart if you know more than everyone else (which means you know more than the market). Not impossible, but to quote a classic, it’s much easier to be first.
What you're saying was the rock-solid conventional wisdom of 1998 or 2005.
It absolutely doesn't apply in 2026. A confluence of things has risen to thoroughly destroy it, right down to its foundations...
...from BTC (valuations rooted in nothing), to the memestock (valuations rooted in nothing), to Doge and Fartcoin with multi-billion dollar market caps (valuations rooted in jokes), to a 10-person+zero-revenue herbal medicine company hitting $33B on the Nasdaq (RGC, valuation rooted in software glitches), to a former beverage company rebranding around 'quantum' for a $2.8B valuation on $200K a quarter (QUBT, valuation rooted in buzzwords), to dying e-commerce firms jumping 3,000% overnight just for announcing they'll buy crypto tokens (valuations rooted solely in clever PR), to pre-revenue AI and nuclear startups fetching $30B+ (valuations rooted in pure vibes)...
And that's just the tip of the iceberg. Ain't it funny how DJT commanded a $5B+ market cap on $3.6 million in annual revenue and over $400 million in losses? (Valuation rooted in tribal vibes.)
'The wisdom of the crowd' in 2026 is a punchline, and not a funny one.
The root cause for all of the above seems to be a mix of a market that is able to operate on the greater fool theory, and an immense amount of liquidity available to those playing the market. The end result seems very likely to be a huge wealth transfer to the existing elite, as the 'fools' left holding the bag will be institutional investors (where the actual control of the individuals whose money is involved is de facto nonexistent) and the public (via the seemingly inevitable bailouts).
Go short that IPO and buy some Google! I agree that the market is irrational and millions of flies can't be wrong, etc. I just also agree that my opinion about it doesn't matter a single bit - the market is made of people like us and they all believe something.
Hot take: Google is probably worth that much but not OAI or ANT.
Both of those are cool but what do they do ultimately? Change some stuff in a Google doc or pull a message from Slack? And where is the stickiness, because as a developer I can route my requests to a different provider with a toggle.
In a way, it is good that companies like Google own sizable stakes in ANT. While ANT is an important company, I don’t think their product is because Google has insane sticky distribution and is very hard to knock off.
ANT products are only months behind open source and half a year behind China. I can easily ditch ANT for OAI with a toggle. Even cloud isn’t this simple for primary workloads.
The underlying idea is that there is sufficient competition that people have an incentive to prevent … laziness ?
That outcome is highly dependent on the market participants performing in a specific manner.
Even then, pricing accurately during the introduction of new is impossible.
Ideally the players wouldn’t have so much cash to burn in a bonfire. This would make it easier for people with alternate positions to be a credible counter balance to excess.
Not saying anything new, just that the market isn’t good at dealing with new general purpose tech, and impaired further when there is a massive capital disparity between the bullish and bearish.
> The market as a whole is irrational [1] and, in general, doesn't know anything that we here don't know.
Yes, and? The price of stock is determined all information available about it, which is why (a) there are rules on insider trading for folks that may have non-public information, and (b) why people try to get an information edge, e.g., hedge funds using satellites:
It should also be noted that The Market™ is made up of different actors with different motivations: from pensions funds which want stability to generate predictable cash flow for thirty years, to day traders and algorithms that only need returns over the course of a couple of days (or hours). Do a search for the term "investor heterogeneity".
Some folks are looking at good companies (or at least good investments, which may not be the same thing), and others are looking for companies they think other people will think are good investments, a meta-game:
> A Keynesian beauty contest is a metaphorical beauty contest in which judges are rewarded for selecting the most popular choices among all judges, rather than those they may personally find the most attractive. This idea is often applied in financial markets, whereby investors could profit more by buying whichever stocks they think other investors will buy, rather than the stocks that have fundamentally the best value, because when other people buy a stock, they bid up the price, allowing an earlier investor to cash out with a profit, regardless of whether the price increases are supported by its fundamentals and theoretical arguments.
What is the (your) purpose for investing: cash flow? growth? a game/competition/fun? Are you investing for yourself (retirement) or for others (maximum profit for funds? cash flow for retirees?)?
Depending to the answers to these questions, the actions in The Market™ can be quite rational for a particular actor depending on how they answer them (and irrational from your perspective given how you answer them). Your goals and others' goals will determine actions and what is "rational".
These are cyclical companies. So at the peak of the cycle in terms of earnings their valuation looks the cheapest (similar to oil companies, just not as extreme).
For semi conductors a relatively reliable metric is Price to tangible book or P/TB ratio. Here it is at 5.81. Lower than the 7.89 for the full of 2026, but historically Micron has traded around 2 P/TB and at the bottom of the cycle 1.5.
You are correct, I oversimplified my explanation. But the conclusion still stands: we're near the peak of the cycle and MU's profits will collapse going forward (until the next cycle, whenever that is). It's very hard to see how current prices can stay anywhere near current levels while profits are collapsing.
Micron can be overvalued and crash and RAM can still be completely unavailable. For example, suppose a bunch of companies have to cancel future commitments for 2028 and beyond. Terrible for Micron’s stock and doesn’t move present supply constraints at all
Is that so? Everything that I have read indicates that CXMT capacity will take a year or so to come online and it's not much capacity (at first) neither.
Because HBM is an integrated product. Nvidia etc would actively have to partner with the Chinese HBM suppliers. So it is a much safer bet for CXMT to sell non-integrated DRAM chips. It's a commodity market after all.
> A typical company that trades at a sub-10 PE is a flatline-to-dying business whose last juice is being squeezed out before the inevitable sale of remaining valuable assets and dissolution.
I don't know if I'd agree with this characterization... plenty of established businesses have gone on for decades at a P/E of 10. It may mean they have no significant growth potential left, but that's not the same as dying
It's unclear what "PE" means and whether it really indicates that investors expect a collapse in profits. Even if they do, you are making predictions based on investor predictions, which is a two-step prediction. It would be better to first find out why the investors are predicting a decline in profits (if indeed they do) and only then use that as a basis for further predictions.
All just speculation and almost no hard facts. Demand could still drop dramatically if they run out of investors / can't build all the datacenters they are planning. I dont care if manufacturing is "booked" for 100years if no one will be actually buying. Manufacturers know this is not going to continue like this. Thats why no one is building out manufacturing capacity.
> feels like plants outside of Taiwan or South Korea are basically political stunts rather than significant production
Investment seems to be emphasizing "advancements such as process technology upgrades, higher-layer stacking, hybrid bonding, and high-value products such as HBM" over capacity expansion [1].
But capacity is being built out, albeit modestly. (Mostly, it seems, by Micron.)
Some are legitimate efforts but everyone (other than SSD manufacturers) forgets how complex the supply chain is for modern chips. The companies making the ultra-pure chemicals and substrates have already contracted out their future production. Even something local like ultrapure water requires experts who themselves are in demand and are already employed wherever they reside now. Few will be willing to move to a greenfield factory in a country without an established chip industry.
Can't place it in time but there was an interview with one of the big memory manufacturer firms said that they were waiting out to see if it was a temporary trend.
That said, it might've already been a year ago already when CXMT was less of an actor and they probably hadn't had time to reap enough profits from the craze yet to justify spending.
> I dont care if manufacturing is "booked" for 100years if no one will be actually buying.
They can't just not buy what they've booked for this year and next (and 2028+ if they have made agreements further in advance) if their immediate DC plans fall through, they have a contract to buy and there will be financial penalties if they pull out (unless they go out of business or have a nice loophole hidden in said contract).
Even if they could pull out (via loophole or being willing to pay the termination fee), they won't because that would give competitors access to the inventory (which is part the point of block booking future stock) and when they do need the stuff in the near future it will very likely cost more than the current arrangement will be setting them back. It would be much better to take the inventory and sit on it ready for other DC plans, they won't all fall through.
Isn't this is one of the reasons why hyperscales use SPV (special-purpose vehicles) to build AI DC? If they'll decide to slow down they will just let SPVs to go bankrupt and if there is no business there are no penalties. The parent company will stay protected.
When you have a term for the shell company you're going to use to avoid liability you can bet your butt someone using is going to point that out to a judge when they sue you.
It’s like when the Ellisons were buying Paramount and Paramount wouldn’t accept an offer from the _revocable_ trust.
When they’re giant companies you have to let them be punished for not doing due diligence or be punished for taking an above market rate price at the risk of the deal breaking (the data center shell companies don’t trade as well as the original company.
Pulling out or not, if they don't need them, they can just sell the contract on to a retail supplier. Even if the booked devices are not suitable for retail, if they aren't actually produced yet, there's the possibility of renegotiating, even paying a penalty to change the contract.
Exactly. Thats how serverpartdeals and others get their stock basically. Massive overbuying from hyperscalers and HotSpares that get sent back to the manufacturer and recertified.
>Thats why no one is building out manufacturing capacity
But they are.
And there is no putting this genie back into the bottle.
The more companies and users use AI, the more data is generated, the more storage is required.
Byte-per-byte, my data generation at least 1000x'd. Code, projects, user storage for each of it, images, videos, songs, voiceovers, games, sessions, worktrees, throw-off sideprojects, work data, inference done on data, trained model weights... insanity.
And with coming years, more and more people will do that. The upcoming capacity isn't even close to meeting basic extrapolated demands, but we still have the "economic disbelief" that it will scale like that.
On another note, Apple somehow still offers the cheapest cloud pricing out of all providers.
>Users will stop using AI to generate images, videos or text
This is actually the worst customer segment for AI as they won't pay for anything but it's using a lot of resources. Most model providers found out pretty early on the target market is businesses. Practically the same way as social media is "free" to end users and funded by companies.
> Byte-per-byte, my data generation at least 1000x'd
Call me crazy, but I don't think AI will create more data than putting video recording cameras into everybody's pockets. Your code projects can't take up more gigabytes than my dog videos.
A lot of the generated data is slop and often not worth storing long-term. People will realize this and introduce more aggressive deletion mechanisms - how much and how quickly depends on the pricing of storage.
This has happened in the past: Email providers (e.g. Microsoft) offered "infinite" storage, i.e. growing over time, but eventually crawled back and applied a hard limit.
You can't just claim scaling will go on like this forever. Extrapolating the number of Huang's affairs makes it look like he will have dozens of them by next year.
I find my 1TB MacBook disk to be already at its limits with a single project. Optimized AI models as well as image assets with texts machine translated to 30 languages are taking a lot of disk space.
It isn't just me, I understand GitHub usage is way up year over year.
That said, with growing production capacity, it looks like 2028 could be a good year to buy new GPU hardware, memory, and storage.
But we are at what now - lets say year 4 - of a new, incredibly innovative technological wave that is radically changing how we approach, use and work with technology.
It is incredibly early, we are at the beginning stages of a massive buildout, and to treat it as just a "hype" thing that will pass or a "bubble" is nothing but a massive misapplication of pattern recognition, as in "oh dotocom was a buble, crypto was a bubble, this surely is a bubble too".
If you asked people in 2000's how much storage they need, they'd say 40GB is surely enough.
If you said to someone "Amazon will make nearly 200B a year renting servers", people would think you were mad.
We have had the chance to be alive and witness how one insane technological revolution (internet) progressed and evolved, changed how humans interact with the world forever and built dozens of different industries. At year 4, the scales of today were unimaginable to anyone but the wildest hackers.
A motion to dismiss a class action lawsuit for RAM price fixing just got thrown out by the judge and the whole AI circular economy is under a lot more scrutiny after the recent NVidia "scandal".
I swear by magnetic hard drives for long-term storage. SSDs are highly volatile if you leave them unpowered for extended periods of time. Magnetism will remain stable for decades if not centuries. Worst case at least some of the data can be read back, whilst with SSD you lose every single bit of information.
They’re much cheaper to produce than SSDs (and their consumer and enterprise prices reflect that), and they’re plenty good for many uses. I don’t like making predictions about the future of technology, but it seems unlikely that they’ll stop being made any time in the next 20 years. They’ll keep getting bigger and cheaper-per-GB.
I wouldn’t want to run the latest AAA video game off one, but they’re fine for most other uses. They’re significantly cheaper per GB, more reliable, and not much slower for most uses. The demand will stick around as long as SSDs don’t become cheaper than them.
Data recovery for HDDs is also generally a decent more likely to be successful.
Obviously it should be your last resort and is not a backup strategy; but the mechanicalness of a HDD makes it far more practical.
Failed heads? Replace the heads.
Failed PCB? Swap in a donor one.
While I’d never recommend it for any important data, if you’re in a pinch and don’t care (and can’t afford professional recovery), it is actually relatively DIY—able even without a clean room. I’ve done it before.
Voxel based glass mediums for non-volatile / archival storage will take the place of spinning rust and tape.
Because glass unlike all other mediums (including optical discs) is exceptionally durable and resilient. You don't need to control environmentals for humidity / oxidation. There's no "spinning" involved so there's no wear and no heat generation to worry about. Also virtually silent operations / no noise.
The only problem being the read write io is garbage. Using a single head yeilds worse speeds then USB 2.0
That said, it's just an issue of parallelization and caching. Same kind of stuff we've been doing with HDD's and other archival storage for decades. But it will mean serious capacities (anything above ~16TB) will probably only be available to data centers.
Microsoft has demonstrated nice improvements in this recording method, including higher read/write speeds, but they are still far from having a commercial product, mainly because of a chicken-and-egg problem.
This needs certain kinds of pulsed lasers and optical components that are currently too expensive, mainly because they are not mass-produced, like e.g. the optical components required to make a Blu-Ray drive. But these components are not mass-produced because for now they are not incorporated in any product with a large potential market, which cannot happen as long as they are so expensive.
Tape is slow. Around 1TB/hour, so writing a 30TB tape takes an entire day. The faster solution is to write to a couple smaller (8TB is a good compromise) HDDs.
yeah, I gave up on upgrading/purchasing any gaming hardware at all tbh... can't justify the price; enjoying some 6y+ old games tho; too bad I didn't get myself any handheld as I thought it was too expensive (and now it costs pretty much 2x, oh well)
me 1.5 years ago: who needs a TB of storage ? everything is backed up in cloud - even if i want an ssd is so much cheaper than the price apple wants ! lemme just get the 256 gb MBP
I strongly advise against using the cloud as your only backup strategy. If you have not already done so please read about the 3-2-1-rule of backups. Do not let a out-of-the-blue-apple-locked-my-account-event stand between you and your digital archives.
Yes absolutely, While not perfect, I have all of my important personal stuff ( photos, documents..etc ) backed up once on the cloud and once on my home PC.
My grievance with not getting more storage on MBP is more related to dev stuff, where as AI has exponentially increased the number of stuff I'm working on in parallel. I have to clear up storage once a week now
Wasn't it given as a major factor in the exit polls for the US midterms? It definitely also ties into affordability, which definitely was one, so I'd assume there is quite some mandate to start curtailing their rolll-out. But maybe it's not enough as you say.
There are lots of places where data centers could be built. If 100 towns reject data center construction and one allows it, the data center gets built. Doesn't even have to be in the US.
Well given the total sum of proposed datacenter project electrical consumption exceeds that of the current generation capacity of the US, the is a reason why legislators started freaking out when they saw the numbers.
> Well given the total sum of proposed datacenter project electrical consumption exceeds that of the current generation capacity of the US, the is a reason why legislators started freaking out when they saw the numbers.
We need to get ready to start powering down homes to power AI datacenters. If we don't do it, China might win!
A lot of the push back excuses against data centers has now been unmasked as total lies - like the water story - and other negatives can maybe be addressed - and the benefits are downplayed because F billionaires and F ai, right? But as rightfully pointed out, datacenters have many places they could go, and as emotionally made decisions deprive communities of the benefits, people in other places will wake up and say "uhm we want the benefits THAT place over there got from a datacenter" .... Of course the more people push back for emotional reasons the higher the probability that many communities will have permanently missed the boat, because their pushback forced it to go overseas (much cheaper!) or even space.
Given these datacenter projects appear to do the same ploy of paying off politicians to grant them those property tax breaks and other nonsense. The benefits to local communities is really hard to argue.
A modern large scale data-center may have 20 full time employees at most, this is peanuts in economic benefits to local communities compared to the impact on local resources (Water and electric). I expect the FT employee count to go down slowly in the next decade as there are various companies working overtime to even automate the servicing of server racks when equipment goes down.
> A lot of the push back excuses against data centers has now been unmasked as total lies - like the water story - and other negatives can maybe be addressed
Total lies? "The water story" isn't a total lie. And the other negatives aren't total lies either. It's ironic to dismiss opposition as total lies while not actually saying what the benefits are.
not really, if you look at the % of people citing AI + data centers as crucial or a major factor, it is far lower at 56% than for other factors (89% cost of living, 90% government corruption, 72% immigration enforcement).
How often were such places built remotely and then the towns around them built as worker housing? There's a powerful difference between "This is how this place has always been since I moved here" vs "This new thing is changing how my long term home is".
yeah. the opposition in the US will simply drive them abroad, where people dgaf. the cost of staffing an overseas DC with a few American overseers is nothing compared to the savings. it's been done a million times during the deindustrialization of US and Europe.
If RAM will also follow this trend then it's done with insane token prices and large compute data centers - AI, LLMs will become commodity. Most companies will either rely on small scale national providers or have some compute in-house sufficient to run coding and other agents.
Even if prices fall in the second half of next year, I'll believe it when I see retail prices actually drop. The recent spikes have made me hold off on any upgrades for now.
SSD and HDD prices going up always seemed strange to me, because LLMs don't need huge amounts of either. So the AI induced increase in demand can't be that large. What LLMs do need is large amounts of DRAM.
So it would be more interesting to predict when DRAM price could go down again. Several new fabs are planned to come online in 2027/2028/2029, but it seems unclear whether or not the substantially increased supply will be absorbed by further growth in DRAM demand from AI companies.
IMO mi think its probably for surveillance capitalism purposes. the ai companies wanna keep data on how u use ai, and other companies now wanna run all ur data thru ai, so its probably an exponential growth of data surrounding the ai boom
High RAM prices are IMO more annoying, but in general,
why do governments not protect the people anymore?
Even in die-hard capitalistic countries such as the
USA, the idea would be that free market competition
leads to lower prices. Clearly this is not the case
anymore. We have de-facto monopolies undermining
basic assumptions of capitalism.
This isn't about "protecting consumers" but capitalism and supply and demand. The A.I. boom will or can lead to enormous profits if it pans out the way its developers foresee it. That consumers are "collateral damage" is unfortunate but the U.S. could profit handsomely from it.
And capitalism also means that supply will eventually catch up with demand lowering prices.
In this case, the high prices are not due to monopolies but due to high demand. The market is doing what it should - people are reducing their consumption of scarce goods (buying less RAM). Eventually, prices will normalize.
Anti-trust laws aren't international in scope, and the US has allowed the NAND/DRAM manufacturing to move off shore, to the benefit of Korea and China. How would the governments protect their people? Price controls? Tariffs? Dumping penalties? None of those will increase supply to lower price.
There is now a natural oligarchy of manufacturers in Korea and China. Without extensive subsidies and/or trade barriers, the US and EU cannot compete.
China has its own needs and is capturing the lower end and consumer market, and will eventually capture the higher end, purely because of their own demand.
They're following the same path as they did with cars and other technology.
Except that this time they're not #4 or 5 in the market, they're #2, and arguably #1.
No contradiction. Free market just means markets are steered by the forces that naturally act on them, which is supply and demand, instead of primarily being steered by government decree or courts or huge complex webs of (burdensome) laws of regulations. Most free market guys don't argue for ZERO regulation, just minimal and clear regulation.
How exactly would protecting work if OpenAI just says hey we bought 80% of the global capacity until 2035? "No, you don't"? Where is competition suddenly supposed to come from in cyclical markets like ram and storage, where every plant takes years of planning and building, and you better not end up with too much capacity if demand dips again?
Keep in mind that the whole AI topic is being framed as a race between China/US, so telling OAI that sorry, only 50% of the future production may be reserved, shafts your own companies and your future world domination attempts.
Was gonna say: I'm so glad it's the honest outlook, imagine we'd get a dishonest outlook, no human would want to read it. So be honest... straight from the system prompt into TFA!
Yeah, it will take some time for Alibaba sellers to prepare the interposer boards that will make enterprise NAND chips scavenged from the AI bubble rubble useable on normal computer people have at home. 2027Q3 sounds about right with the Shenzen shops getting hand of populating the boards.
Wait what? That's just when we'll all be living on Mars and cancer will be cured, and we'll have high universal income and also autonomous self driving cars for the masses.
Wild speculation at timing aside, I am looking forward to prices going down eventually (and inevitably). What I’m hoping for is that memory makers realize that there’s a market to be had selling lots of memory to consumers for local AI usage, as well as depressing SSD prices further by expanding into large capacities for consumer and prosumer market segments. We’ve generally been stuck at 64GB RAM/4TB SSDs for half a decade now, waiting for those price/capacity ratios to finally stop being manipulated by crypto and AI bubbles and return to normal growth patterns and compute to evolve alongside them.
Just think: 128GB Mac Minis driving Hermes with open weight frontier models and 2TB Mac Studios serving trillion-parameter models in a single chassis. Home NAS systems built on SSDs but offering storage capacities similar mid-tier spinning rust drives (8TB to 16TB drive sizes built on high endurance QLC).
I’m excited for the inevitable price drop relative to capacity, and all the cool stuff we’ll build with it.
It might go up slower which will mean it is getting cheper since there will be THE BEST inflation in a year because of the BEST president THE WORLD HAS EVER SEEN.
I mean, I agree given the way this was rigged that they’re going to see how far they can keep the prices. I suspect we’re going to see a recession soon, so let’s see how long they can keep this racket going if they end up with excess inventory or low sales.
If I can quote the great thinkers known as backstreet boys - quit playing games with my heart. The only way silicone to become cheap is the AI bubble to pop and hard or massive new capacity to come online. In the first case we won't have money to buy and the second - we all will be graybeards.
I call bullshit – the "article" also stinks of AI slop.
As long as the US/Israel stuff is going on in the middle east, there's going to be no downwards movements in any prices, SSDs or otherwise. End of story.
Why ?
For a start, before you consider anything else, stuff needs to be shipped around. For that you need (a) Oil (b) The ability to use shortest-route transport routes.
The middle east nonsense messes all that up completely. And even if the war stopped tomorrow, it will still take some time for things to return to some semblance of normality, such is the extent of the disruption.
The other day my firm was looking to ship about 180kg of expensive materials packed into 9 large crates from Korea to the US -- which, if it were SSDs, would be a lot of SSDs.
You're absolutely right that it's much more expensive than ever, what with the war and the holidays and all. But we were still quoted just $2600 for Air Freight and $1300 for Sea Freight. Which, if it were SSDs, would work out to pennies per unit. (For expensive goods, tariffs are a much more punishing cost.)
It's demand driven. The Iran War might raise prices more by cutting off Asia from helium used in chip production, rather than by driving up oil prices...
imho i think alot of prices for things like shipping are a gouge and have nothing todo with iran war. look at what happened during covid when every company with supply of anykind tried to leverage their position to gouge everything and everyone they could.
the iran war itself is a gouge on the american taxpayer by the defense industry, if war ends , defense indutry gravy train goes back to normal. sad how the US spent so much money on bombs, and yet the war continues... do they not know where to drop the bombs to end the war?!, seems like they drop billions in munitions in the middle of the desert
Shipping cost affects cheap (compare to mass/volume) products but in RAM price shipping is a negligible fraction. Where Iran war affects the more AFIK is the cost of credit which slows investments into new capacity by the manufactures. All thing being equal low interest rates allow to expand capacity faster.
Most SSD/RAM and chips generally ship by air. At the value density of such things, nobody wants to maintain stock sitting on ships. Interest rates alone would cover the cost of air cargo. Just think of how much ram can be on a single pallet, let alone in a 40-foot shipping container.
If you look at valuation of a company like Micron (MU), it's pretty clear that the market expects a collapse in their profits. They are growing at absolutely insane rates, whether you look at YoY (well over +300%) or QoQ (over +40%) in both revenue and profits. Yet their PE is below 10. SK Hynix is similar.
A typical company that trades at a sub-10 PE is a flatline-to-dying business whose last juice is being squeezed out before the inevitable sale of remaining valuable assets and dissolution.
Of course the current market price is not a perfect predictor of the future of a business but despite all the memes and anecdotes, the market is far more often basically right than dramatically wrong (it's just that when it does get things dramatically wrong, that's what gets all the publicity, for obvious reasons).
So, assuming the market is roughly right in its valuation of Micron or SK Hynix, either demand will collapse or supply will surge or, some combo of the two. Either way it seems extremely likely that within a couple of years at most the prices of DRAM and NAND will come waaaay down.
The market as a whole is irrational [1] and, in general, doesn't know anything that we here don't know. In fact, the market typically reacts to things in a much baser and less analytical manner.
Micron and SK Hynix are at sub-10 PE because they are run by people who are poor salesmen and don't know how to drive or ride a hype cycle. If they focused less on engineering and more on PR, media outreach, "socials," performative patent applications and gimmicks, they'd be trading at double their current valuations, or more.
It would also be worse for society. Puffery is rewarded. Thomas Carlyle's example of the seven-foot lath-and-plaster hat paraded through 19th century London. The memestock. Crypto.
[1]-Does anybody genuinely believe that Google's worth 3.43T and Anthropic is worth 1.5T? This goes beyond irrationality and into lunacy...
Google expects to make 200 billion in profit in 2026, so their PE ratio is somewhere between 15-20. You can argue if that profit is sustainable over the long term, but a PE ratio of 15-20 does not seem like lunacy to me.
I think they are saying anthropic's valuation is unreasonable, not Google's
It's trivial to make bets that the market is dramatically wrong by using long-dated options. These are multi-year-long bets at high leverage (i.e. you bet relatively little and win relatively a lot if you're right). If you are extremely confident that you are right and the market is wrong, you could use these and become rich. I would not recommend it -- the market is generally way better at valuing businesses than us casuals.
Edit: FWIW, Google at 3.4T seems almost a bargain -- not that I would go all in on it instead of just keeping my broad index funds. I'm less sure about Anthropic at 1.5T
It's the ratio of Google to Anthropic that is off. If Anthropic is genuinely "worth" 1.5T, then Google should be much greater. They have everything Anthropic has, and a veritable mountain of assets that Anthropic does not have. But the vibes-based market doesn't price it in.
I've been trying to tell many people this. Their valuations can make sense, but not at the same time.
I see, I misunderstood what you meant. If I had to bet on one, I would agree with you and bet on Google over Anthropic. But I'd rather not bet at all
The OP is right, but the shorts always lose because the market will always be wrong.
If I had a million bucks I'd spend it developing a trading algorithm that completely ignored company fundamentals and analyzed only Vibes.
I'd much rather invest in market fundamentals as an individual. Many billions are poured into (and made on) vibe-based market predictions. It's much safer to go after smaller returns on well priced stocks than play predict the market irrationality against behemoths. After all, it worked for Buffet.
Very crowded business, that
"Does anybody genuinely believe that Google's worth 3.43T and Anthropic is worth 1.5T? "
They control much of the information flow and therefore decisions that are made world wide. Is it not obvious to you, that this power is worth a lot?
My point, which admittedly could have been clearer, was that the ratio between the two companies makes no sense. Think about what we know Anthropic has, in relation to what we know Google has.
Both have data, both have AI. Google is more established and likely has way more data broadly, but the perceived number 1 in AI is now Antrophic. I never payed google 200$ a month (and always used adblockers), but now I do pay so much for Antrophic and for me it is very worth it. And governments and companies seem to think likewise.
So basically also when it is about the ratio, it does not seem so irrational to me. In the big picture with all data in the open it might, but from my limited point of view it is definitely not obvious wrong.
The market has participants with varied and contradictory motives.
Main street investors rely on stocks to produce returns.
Large investors are willing to invest at a loss, if the investment provides control / access to a company which may augment their other investments. Sure, your $100M investment in Anthropic may lose 50% of its value, but if you turn a blind eye to it, then you may lose 100% of your $500M consulting business. The $100M investment may also include information that maybe is or isn't "insider", but maybe it's tailored to your needs and allows you to cut so much in costs in your other businesses.
(Numbers are probably grossly out of scale)
Some are paying to own the company, not to get a share of the profits.
Your guess at those valuations is as good as anyone’s. Turns out if you put lots of such anyones in a room it makes a market. If you doubt the market you doubt the wisdom of the crowd, which is very smart if you know more than everyone else (which means you know more than the market). Not impossible, but to quote a classic, it’s much easier to be first.
What you're saying was the rock-solid conventional wisdom of 1998 or 2005.
It absolutely doesn't apply in 2026. A confluence of things has risen to thoroughly destroy it, right down to its foundations...
...from BTC (valuations rooted in nothing), to the memestock (valuations rooted in nothing), to Doge and Fartcoin with multi-billion dollar market caps (valuations rooted in jokes), to a 10-person+zero-revenue herbal medicine company hitting $33B on the Nasdaq (RGC, valuation rooted in software glitches), to a former beverage company rebranding around 'quantum' for a $2.8B valuation on $200K a quarter (QUBT, valuation rooted in buzzwords), to dying e-commerce firms jumping 3,000% overnight just for announcing they'll buy crypto tokens (valuations rooted solely in clever PR), to pre-revenue AI and nuclear startups fetching $30B+ (valuations rooted in pure vibes)...
And that's just the tip of the iceberg. Ain't it funny how DJT commanded a $5B+ market cap on $3.6 million in annual revenue and over $400 million in losses? (Valuation rooted in tribal vibes.)
'The wisdom of the crowd' in 2026 is a punchline, and not a funny one.
The root cause for all of the above seems to be a mix of a market that is able to operate on the greater fool theory, and an immense amount of liquidity available to those playing the market. The end result seems very likely to be a huge wealth transfer to the existing elite, as the 'fools' left holding the bag will be institutional investors (where the actual control of the individuals whose money is involved is de facto nonexistent) and the public (via the seemingly inevitable bailouts).
Go short that IPO and buy some Google! I agree that the market is irrational and millions of flies can't be wrong, etc. I just also agree that my opinion about it doesn't matter a single bit - the market is made of people like us and they all believe something.
Hot take: Google is probably worth that much but not OAI or ANT.
Both of those are cool but what do they do ultimately? Change some stuff in a Google doc or pull a message from Slack? And where is the stickiness, because as a developer I can route my requests to a different provider with a toggle.
In a way, it is good that companies like Google own sizable stakes in ANT. While ANT is an important company, I don’t think their product is because Google has insane sticky distribution and is very hard to knock off.
ANT products are only months behind open source and half a year behind China. I can easily ditch ANT for OAI with a toggle. Even cloud isn’t this simple for primary workloads.
The underlying idea is that there is sufficient competition that people have an incentive to prevent … laziness ?
That outcome is highly dependent on the market participants performing in a specific manner.
Even then, pricing accurately during the introduction of new is impossible.
Ideally the players wouldn’t have so much cash to burn in a bonfire. This would make it easier for people with alternate positions to be a credible counter balance to excess.
Not saying anything new, just that the market isn’t good at dealing with new general purpose tech, and impaired further when there is a massive capital disparity between the bullish and bearish.
> The market as a whole is irrational [1] and, in general, doesn't know anything that we here don't know.
Yes, and? The price of stock is determined all information available about it, which is why (a) there are rules on insider trading for folks that may have non-public information, and (b) why people try to get an information edge, e.g., hedge funds using satellites:
* https://newsroom.haas.berkeley.edu/how-hedge-funds-use-satel...
It should also be noted that The Market™ is made up of different actors with different motivations: from pensions funds which want stability to generate predictable cash flow for thirty years, to day traders and algorithms that only need returns over the course of a couple of days (or hours). Do a search for the term "investor heterogeneity".
Some folks are looking at good companies (or at least good investments, which may not be the same thing), and others are looking for companies they think other people will think are good investments, a meta-game:
> A Keynesian beauty contest is a metaphorical beauty contest in which judges are rewarded for selecting the most popular choices among all judges, rather than those they may personally find the most attractive. This idea is often applied in financial markets, whereby investors could profit more by buying whichever stocks they think other investors will buy, rather than the stocks that have fundamentally the best value, because when other people buy a stock, they bid up the price, allowing an earlier investor to cash out with a profit, regardless of whether the price increases are supported by its fundamentals and theoretical arguments.
* https://en.wikipedia.org/wiki/Keynesian_beauty_contest
What is the (your) purpose for investing: cash flow? growth? a game/competition/fun? Are you investing for yourself (retirement) or for others (maximum profit for funds? cash flow for retirees?)?
Depending to the answers to these questions, the actions in The Market™ can be quite rational for a particular actor depending on how they answer them (and irrational from your perspective given how you answer them). Your goals and others' goals will determine actions and what is "rational".
These are cyclical companies. So at the peak of the cycle in terms of earnings their valuation looks the cheapest (similar to oil companies, just not as extreme).
For semi conductors a relatively reliable metric is Price to tangible book or P/TB ratio. Here it is at 5.81. Lower than the 7.89 for the full of 2026, but historically Micron has traded around 2 P/TB and at the bottom of the cycle 1.5.
You are correct, I oversimplified my explanation. But the conclusion still stands: we're near the peak of the cycle and MU's profits will collapse going forward (until the next cycle, whenever that is). It's very hard to see how current prices can stay anywhere near current levels while profits are collapsing.
I agree with you, I just wanted to add some info.
NAND “spot price” already dropped 20% in the last month, so it could already be happening.
Micron can be overvalued and crash and RAM can still be completely unavailable. For example, suppose a bunch of companies have to cancel future commitments for 2028 and beyond. Terrible for Micron’s stock and doesn’t move present supply constraints at all
I upvoted your comment because I want to believe
CXMT is going to make enough DRAM soon enough to more than make up for whatever was lost converting DRAM fabs to HBM.
Less familiar with the NAND market.
Is that so? Everything that I have read indicates that CXMT capacity will take a year or so to come online and it's not much capacity (at first) neither.
And why would CXMT miss the hype train others are riding?
Can you trust Chinese memory? Will you try it out if it's cheaper?
It's an easy time to grab market share.
Can you trust Chinese everything else? Lots of people do
Because HBM is an integrated product. Nvidia etc would actively have to partner with the Chinese HBM suppliers. So it is a much safer bet for CXMT to sell non-integrated DRAM chips. It's a commodity market after all.
> A typical company that trades at a sub-10 PE is a flatline-to-dying business whose last juice is being squeezed out before the inevitable sale of remaining valuable assets and dissolution.
I don't know if I'd agree with this characterization... plenty of established businesses have gone on for decades at a P/E of 10. It may mean they have no significant growth potential left, but that's not the same as dying
.
It's unclear what "PE" means and whether it really indicates that investors expect a collapse in profits. Even if they do, you are making predictions based on investor predictions, which is a two-step prediction. It would be better to first find out why the investors are predicting a decline in profits (if indeed they do) and only then use that as a basis for further predictions.
price-to-earnings (P/E) ratio
[delayed]
All just speculation and almost no hard facts. Demand could still drop dramatically if they run out of investors / can't build all the datacenters they are planning. I dont care if manufacturing is "booked" for 100years if no one will be actually buying. Manufacturers know this is not going to continue like this. Thats why no one is building out manufacturing capacity.
> no one is building out manufacturing capacity
This is unfair, since I’m asking you to support a negative, but do you have a source for this?
EDIT: The claim is bunk [1][2].
[1] https://en.sedaily.com/business/2026/11/03/sk-hynix-picks-ne...
[2] https://www.blocksandfiles.com/flash/2026/08/12/sk-hynix-exp...
Good to check but feels like plants outside of Taiwan or South Korea are basically political stunts rather than significant production.
> feels like plants outside of Taiwan or South Korea are basically political stunts rather than significant production
Investment seems to be emphasizing "advancements such as process technology upgrades, higher-layer stacking, hybrid bonding, and high-value products such as HBM" over capacity expansion [1].
But capacity is being built out, albeit modestly. (Mostly, it seems, by Micron.)
[1] https://www.trendforce.com/presscenter/news/20251113-12780.h...
Some are legitimate efforts but everyone (other than SSD manufacturers) forgets how complex the supply chain is for modern chips. The companies making the ultra-pure chemicals and substrates have already contracted out their future production. Even something local like ultrapure water requires experts who themselves are in demand and are already employed wherever they reside now. Few will be willing to move to a greenfield factory in a country without an established chip industry.
Can't place it in time but there was an interview with one of the big memory manufacturer firms said that they were waiting out to see if it was a temporary trend.
That said, it might've already been a year ago already when CXMT was less of an actor and they probably hadn't had time to reap enough profits from the craze yet to justify spending.
> I dont care if manufacturing is "booked" for 100years if no one will be actually buying.
They can't just not buy what they've booked for this year and next (and 2028+ if they have made agreements further in advance) if their immediate DC plans fall through, they have a contract to buy and there will be financial penalties if they pull out (unless they go out of business or have a nice loophole hidden in said contract).
Even if they could pull out (via loophole or being willing to pay the termination fee), they won't because that would give competitors access to the inventory (which is part the point of block booking future stock) and when they do need the stuff in the near future it will very likely cost more than the current arrangement will be setting them back. It would be much better to take the inventory and sit on it ready for other DC plans, they won't all fall through.
Isn't this is one of the reasons why hyperscales use SPV (special-purpose vehicles) to build AI DC? If they'll decide to slow down they will just let SPVs to go bankrupt and if there is no business there are no penalties. The parent company will stay protected.
When you have a term for the shell company you're going to use to avoid liability you can bet your butt someone using is going to point that out to a judge when they sue you.
Why?
It’s like when the Ellisons were buying Paramount and Paramount wouldn’t accept an offer from the _revocable_ trust.
When they’re giant companies you have to let them be punished for not doing due diligence or be punished for taking an above market rate price at the risk of the deal breaking (the data center shell companies don’t trade as well as the original company.
Pulling out or not, if they don't need them, they can just sell the contract on to a retail supplier. Even if the booked devices are not suitable for retail, if they aren't actually produced yet, there's the possibility of renegotiating, even paying a penalty to change the contract.
Exactly. Thats how serverpartdeals and others get their stock basically. Massive overbuying from hyperscalers and HotSpares that get sent back to the manufacturer and recertified.
They might go out of business.
> there will be financial penalties
do we know how large are those?
Who cares about contracts when the whole company will likely just fold?
>Thats why no one is building out manufacturing capacity
But they are. And there is no putting this genie back into the bottle. The more companies and users use AI, the more data is generated, the more storage is required.
Byte-per-byte, my data generation at least 1000x'd. Code, projects, user storage for each of it, images, videos, songs, voiceovers, games, sessions, worktrees, throw-off sideprojects, work data, inference done on data, trained model weights... insanity.
And with coming years, more and more people will do that. The upcoming capacity isn't even close to meeting basic extrapolated demands, but we still have the "economic disbelief" that it will scale like that.
On another note, Apple somehow still offers the cheapest cloud pricing out of all providers.
Ai is compressing the internet to a usable prediction machine. Who wants to store stuff that can be regenerated any time
HDD manufacturers arent building out anything.
> basic extrapolated demands
Basic extrapolation used to say human population will approach infinity this November...
What are the possibilities here tho?
>Companies will stop AI adoption and technology won't penetrate the market
Not likely. Only a small subset has adopted it for now and doesn't seem to be slowing down.
>Users will stop using AI to generate images, videos or text
Also not likely. AI still has small market penetration and you can already see GPT slop everywhere.
>We will find a 10-fold improvement to SOTA compression
Not likely, but even if we did, you'd need to scale it globally in a few years.
>Less people will buy phones, consoles, computers
This could be possible due to high prices, but not in a manner that impacts the overall market demand.
So if these are all not likely, what is left as likely?
>Prices grow because demand skyrockets >New players enter the market to catch up with demand
And even if a few neoclouds fail and some data center buildout stops, it really is just a temporary setback to an evergrowing machine.
Ones that IMO are not unlikely:
- AI companies significantly overestimated the size of the market
- AI companies, on average, significantly overestimated the fraction of the market they will capture
>Users will stop using AI to generate images, videos or text
This is actually the worst customer segment for AI as they won't pay for anything but it's using a lot of resources. Most model providers found out pretty early on the target market is businesses. Practically the same way as social media is "free" to end users and funded by companies.
or just not so many people will use this tech for whatever reason, eg. it's too expensive
Is it? I know a lot of people that aren't in tech and use free versions of it all.
It sounds like you are agreeing with OP. If it was not too expensive presumably these people you know would pay.
How does income from free versions pay inflated prices for SSDs?
relevant xkcd: https://xkcd.com/605/
> Byte-per-byte, my data generation at least 1000x'd
Call me crazy, but I don't think AI will create more data than putting video recording cameras into everybody's pockets. Your code projects can't take up more gigabytes than my dog videos.
A lot of the generated data is slop and often not worth storing long-term. People will realize this and introduce more aggressive deletion mechanisms - how much and how quickly depends on the pricing of storage.
This has happened in the past: Email providers (e.g. Microsoft) offered "infinite" storage, i.e. growing over time, but eventually crawled back and applied a hard limit.
You can't just claim scaling will go on like this forever. Extrapolating the number of Huang's affairs makes it look like he will have dozens of them by next year.
I have no plans of deleting my projects.
I find my 1TB MacBook disk to be already at its limits with a single project. Optimized AI models as well as image assets with texts machine translated to 30 languages are taking a lot of disk space.
It isn't just me, I understand GitHub usage is way up year over year.
That said, with growing production capacity, it looks like 2028 could be a good year to buy new GPU hardware, memory, and storage.
I'm not claiming it will go on forever.
But we are at what now - lets say year 4 - of a new, incredibly innovative technological wave that is radically changing how we approach, use and work with technology.
It is incredibly early, we are at the beginning stages of a massive buildout, and to treat it as just a "hype" thing that will pass or a "bubble" is nothing but a massive misapplication of pattern recognition, as in "oh dotocom was a buble, crypto was a bubble, this surely is a bubble too".
If you asked people in 2000's how much storage they need, they'd say 40GB is surely enough.
If you said to someone "Amazon will make nearly 200B a year renting servers", people would think you were mad.
We have had the chance to be alive and witness how one insane technological revolution (internet) progressed and evolved, changed how humans interact with the world forever and built dozens of different industries. At year 4, the scales of today were unimaginable to anyone but the wildest hackers.
> my data generation at least 1000x'd.
And you extrapolate that to everyone?
Can’t even spell this handle without “hype”.
Found Gaving Belsons HN account. Data-Geddon is right around the corner for decades now.
The backlog from all the non AI money is growing though.
I'm waiting it out.
since when are tech or stock people interested in real world events? They rather live in the fantasy trading world.
A motion to dismiss a class action lawsuit for RAM price fixing just got thrown out by the judge and the whole AI circular economy is under a lot more scrutiny after the recent NVidia "scandal".
I hope you are right. I would also like RAM and GPU Prices to go down aswell.
I’m guessing there will be a lot of pent up demand as the prices go lower.
Yes but once supply stabilizes, demand will go down, as well prices.
> spinning drives
I'm slightly surprised that these are still being manufactured, but it is obvious to me that their pricing is different from SSDs.
I swear by magnetic hard drives for long-term storage. SSDs are highly volatile if you leave them unpowered for extended periods of time. Magnetism will remain stable for decades if not centuries. Worst case at least some of the data can be read back, whilst with SSD you lose every single bit of information.
- 3x cheaper per byte
- can be rewritten an infinite number of times
- don't lose data just by sitting unpowered
- tells you when it's about to fail
- data recovery is possible
They’re much cheaper to produce than SSDs (and their consumer and enterprise prices reflect that), and they’re plenty good for many uses. I don’t like making predictions about the future of technology, but it seems unlikely that they’ll stop being made any time in the next 20 years. They’ll keep getting bigger and cheaper-per-GB.
I wouldn’t want to run the latest AAA video game off one, but they’re fine for most other uses. They’re significantly cheaper per GB, more reliable, and not much slower for most uses. The demand will stick around as long as SSDs don’t become cheaper than them.
Data recovery for HDDs is also generally a decent more likely to be successful.
Obviously it should be your last resort and is not a backup strategy; but the mechanicalness of a HDD makes it far more practical.
Failed heads? Replace the heads.
Failed PCB? Swap in a donor one.
While I’d never recommend it for any important data, if you’re in a pinch and don’t care (and can’t afford professional recovery), it is actually relatively DIY—able even without a clean room. I’ve done it before.
I could see a hyper-dense optical medium replacing magnetic hard drives as the preferred "spinning disk" in the next 20 years.
https://www.nature.com/articles/s41586-023-06980-y
Negative.
Voxel based glass mediums for non-volatile / archival storage will take the place of spinning rust and tape.
Because glass unlike all other mediums (including optical discs) is exceptionally durable and resilient. You don't need to control environmentals for humidity / oxidation. There's no "spinning" involved so there's no wear and no heat generation to worry about. Also virtually silent operations / no noise.
The only problem being the read write io is garbage. Using a single head yeilds worse speeds then USB 2.0
That said, it's just an issue of parallelization and caching. Same kind of stuff we've been doing with HDD's and other archival storage for decades. But it will mean serious capacities (anything above ~16TB) will probably only be available to data centers.
Microsoft has demonstrated nice improvements in this recording method, including higher read/write speeds, but they are still far from having a commercial product, mainly because of a chicken-and-egg problem.
This needs certain kinds of pulsed lasers and optical components that are currently too expensive, mainly because they are not mass-produced, like e.g. the optical components required to make a Blu-Ray drive. But these components are not mass-produced because for now they are not incorporated in any product with a large potential market, which cannot happen as long as they are so expensive.
You can already just go and buy a 32TB drive, although demand for them is pretty high.
LTO-10 is around $10/TB. Demand for that is also high. More data is being produced than ever before in human history.
Demand is up and $/TN has gone up in price significantly.
You have to have them to do large scale outs.
Wait until you learn that tapes are still being produced with Fujifilm extending production by quite a lot in the last few years.
I wish tape drives weren't so expensive. Otherwise they'd be amazing for backups
Tape is slow. Around 1TB/hour, so writing a 30TB tape takes an entire day. The faster solution is to write to a couple smaller (8TB is a good compromise) HDDs.
I agree. I'd love them for my home setup.
Much bigger capacity on the spinning drives for much less - well at least until this price spike.
I can’t imagine that companies backing up locally are using ssd over hdd.
I regret not upgrading my PS5 hard drive space when I had the chance.
yeah, I gave up on upgrading/purchasing any gaming hardware at all tbh... can't justify the price; enjoying some 6y+ old games tho; too bad I didn't get myself any handheld as I thought it was too expensive (and now it costs pretty much 2x, oh well)
With luck you will have that chance with the PS6.
me 1.5 years ago: who needs a TB of storage ? everything is backed up in cloud - even if i want an ssd is so much cheaper than the price apple wants ! lemme just get the 256 gb MBP
me rn : (T_T)
I strongly advise against using the cloud as your only backup strategy. If you have not already done so please read about the 3-2-1-rule of backups. Do not let a out-of-the-blue-apple-locked-my-account-event stand between you and your digital archives.
Yes absolutely, While not perfect, I have all of my important personal stuff ( photos, documents..etc ) backed up once on the cloud and once on my home PC.
My grievance with not getting more storage on MBP is more related to dev stuff, where as AI has exponentially increased the number of stuff I'm working on in parallel. I have to clear up storage once a week now
Wishful thinking. Opposition to data centres is nowhere near universal enough to dent demand, nowhere near enough supply is coming online.
Wasn't it given as a major factor in the exit polls for the US midterms? It definitely also ties into affordability, which definitely was one, so I'd assume there is quite some mandate to start curtailing their rolll-out. But maybe it's not enough as you say.
There are lots of places where data centers could be built. If 100 towns reject data center construction and one allows it, the data center gets built. Doesn't even have to be in the US.
Practically the issue is not towns allowing it but the electricity connections, which are being blocked by state legislators.
Even Texas has paused electricity hookups which has stalled several builds and expansions.
Well given the total sum of proposed datacenter project electrical consumption exceeds that of the current generation capacity of the US, the is a reason why legislators started freaking out when they saw the numbers.
> Well given the total sum of proposed datacenter project electrical consumption exceeds that of the current generation capacity of the US, the is a reason why legislators started freaking out when they saw the numbers.
We need to get ready to start powering down homes to power AI datacenters. If we don't do it, China might win!
A lot of the push back excuses against data centers has now been unmasked as total lies - like the water story - and other negatives can maybe be addressed - and the benefits are downplayed because F billionaires and F ai, right? But as rightfully pointed out, datacenters have many places they could go, and as emotionally made decisions deprive communities of the benefits, people in other places will wake up and say "uhm we want the benefits THAT place over there got from a datacenter" .... Of course the more people push back for emotional reasons the higher the probability that many communities will have permanently missed the boat, because their pushback forced it to go overseas (much cheaper!) or even space.
A pushback assuming the facts are correct:
https://backofmind.substack.com/p/they-paved-paradise-and-pu...
What benefit do you get from a data center in your backyard that you wouldn't get from a data center a hundred miles away?
An unberable noise and a higher power bill
Electricity is already pricy enough, I don't need to pay out the ass for AI that I try to stay away from.
Given these datacenter projects appear to do the same ploy of paying off politicians to grant them those property tax breaks and other nonsense. The benefits to local communities is really hard to argue.
A modern large scale data-center may have 20 full time employees at most, this is peanuts in economic benefits to local communities compared to the impact on local resources (Water and electric). I expect the FT employee count to go down slowly in the next decade as there are various companies working overtime to even automate the servicing of server racks when equipment goes down.
> A lot of the push back excuses against data centers has now been unmasked as total lies - like the water story - and other negatives can maybe be addressed
Total lies? "The water story" isn't a total lie. And the other negatives aren't total lies either. It's ironic to dismiss opposition as total lies while not actually saying what the benefits are.
not really, if you look at the % of people citing AI + data centers as crucial or a major factor, it is far lower at 56% than for other factors (89% cost of living, 90% government corruption, 72% immigration enforcement).
Opposition is much more localized/NIMBY-like.
https://www.nbcnews.com/politics/2026-elections/exit-polls
56% still sounds pretty high to me, regardless of how it sits relative to other issues.
Uranium enrichment plants, coal mines and chemical factories have been built all over the US, despite nobody wanting to live near one
How often were such places built remotely and then the towns around them built as worker housing? There's a powerful difference between "This is how this place has always been since I moved here" vs "This new thing is changing how my long term home is".
When people are poor they will accept anything.
yeah. the opposition in the US will simply drive them abroad, where people dgaf. the cost of staffing an overseas DC with a few American overseers is nothing compared to the savings. it's been done a million times during the deindustrialization of US and Europe.
If RAM will also follow this trend then it's done with insane token prices and large compute data centers - AI, LLMs will become commodity. Most companies will either rely on small scale national providers or have some compute in-house sufficient to run coding and other agents.
Instead of thinking about more storage, we should be looking at better compression algorithms. Less is more.
Imagine if we just do both!
Ha! That’s the way!
Do not buy RAM, compress https://news.ycombinator.com/item?id=50557661
Even if prices fall in the second half of next year, I'll believe it when I see retail prices actually drop. The recent spikes have made me hold off on any upgrades for now.
This is an ad.
Nah, it’s cool. They just want to give you a free, closed source app that crawls your hard drive.
You don’t have trust issues, do you?
SSD and HDD prices going up always seemed strange to me, because LLMs don't need huge amounts of either. So the AI induced increase in demand can't be that large. What LLMs do need is large amounts of DRAM.
So it would be more interesting to predict when DRAM price could go down again. Several new fabs are planned to come online in 2027/2028/2029, but it seems unclear whether or not the substantially increased supply will be absorbed by further growth in DRAM demand from AI companies.
The LLMs might not need large amount of storage space but the data to train them does. Libgen alone takes 100s of TB and that's just books.
I saw claims that hyperscalers were anticipating needing to store large amounts is AI generated content.
IMO mi think its probably for surveillance capitalism purposes. the ai companies wanna keep data on how u use ai, and other companies now wanna run all ur data thru ai, so its probably an exponential growth of data surrounding the ai boom
High RAM prices are IMO more annoying, but in general, why do governments not protect the people anymore? Even in die-hard capitalistic countries such as the USA, the idea would be that free market competition leads to lower prices. Clearly this is not the case anymore. We have de-facto monopolies undermining basic assumptions of capitalism.
This isn't about "protecting consumers" but capitalism and supply and demand. The A.I. boom will or can lead to enormous profits if it pans out the way its developers foresee it. That consumers are "collateral damage" is unfortunate but the U.S. could profit handsomely from it.
And capitalism also means that supply will eventually catch up with demand lowering prices.
In this case, the high prices are not due to monopolies but due to high demand. The market is doing what it should - people are reducing their consumption of scarce goods (buying less RAM). Eventually, prices will normalize.
Anti-trust laws aren't international in scope, and the US has allowed the NAND/DRAM manufacturing to move off shore, to the benefit of Korea and China. How would the governments protect their people? Price controls? Tariffs? Dumping penalties? None of those will increase supply to lower price.
There is now a natural oligarchy of manufacturers in Korea and China. Without extensive subsidies and/or trade barriers, the US and EU cannot compete.
China has its own needs and is capturing the lower end and consumer market, and will eventually capture the higher end, purely because of their own demand.
They're following the same path as they did with cars and other technology.
Except that this time they're not #4 or 5 in the market, they're #2, and arguably #1.
We have capitalism, what we don't have are free markets.
You can't have both. Free markets means markets are free, but capitalism means markets are controlled by whomever has the most capital.
No contradiction. Free market just means markets are steered by the forces that naturally act on them, which is supply and demand, instead of primarily being steered by government decree or courts or huge complex webs of (burdensome) laws of regulations. Most free market guys don't argue for ZERO regulation, just minimal and clear regulation.
How exactly would protecting work if OpenAI just says hey we bought 80% of the global capacity until 2035? "No, you don't"? Where is competition suddenly supposed to come from in cyclical markets like ram and storage, where every plant takes years of planning and building, and you better not end up with too much capacity if demand dips again?
Keep in mind that the whole AI topic is being framed as a race between China/US, so telling OAI that sorry, only 50% of the future production may be reserved, shafts your own companies and your future world domination attempts.
The 'honest outlook' and 'straight answer' in this article are a bit suspicious
Was gonna say: I'm so glad it's the honest outlook, imagine we'd get a dishonest outlook, no human would want to read it. So be honest... straight from the system prompt into TFA!
Did the article make you genuinely curious?
The analysis part is missing from this analysis article.
Nitpicking are we? /s
Yeah, it will take some time for Alibaba sellers to prepare the interposer boards that will make enterprise NAND chips scavenged from the AI bubble rubble useable on normal computer people have at home. 2027Q3 sounds about right with the Shenzen shops getting hand of populating the boards.
Wait what? That's just when we'll all be living on Mars and cancer will be cured, and we'll have high universal income and also autonomous self driving cars for the masses.
Oh right, never trust tech predictions
I'll believe it when I see it.
“Anyone who causes harm by forecasting should be treated as either a fool or a liar.” —NN Taleb
Wild speculation at timing aside, I am looking forward to prices going down eventually (and inevitably). What I’m hoping for is that memory makers realize that there’s a market to be had selling lots of memory to consumers for local AI usage, as well as depressing SSD prices further by expanding into large capacities for consumer and prosumer market segments. We’ve generally been stuck at 64GB RAM/4TB SSDs for half a decade now, waiting for those price/capacity ratios to finally stop being manipulated by crypto and AI bubbles and return to normal growth patterns and compute to evolve alongside them.
Just think: 128GB Mac Minis driving Hermes with open weight frontier models and 2TB Mac Studios serving trillion-parameter models in a single chassis. Home NAS systems built on SSDs but offering storage capacities similar mid-tier spinning rust drives (8TB to 16TB drive sizes built on high endurance QLC).
I’m excited for the inevitable price drop relative to capacity, and all the cool stuff we’ll build with it.
yeah right
I doubt it. Maybe costs will get lower, but end-user prices? Hardly ever
The price of computer components has decreased a lot since the 80s.
See you guys in 2066
It might go up slower which will mean it is getting cheper since there will be THE BEST inflation in a year because of the BEST president THE WORLD HAS EVER SEEN.
I mean, I agree given the way this was rigged that they’re going to see how far they can keep the prices. I suspect we’re going to see a recession soon, so let’s see how long they can keep this racket going if they end up with excess inventory or low sales.
If I can quote the great thinkers known as backstreet boys - quit playing games with my heart. The only way silicone to become cheap is the AI bubble to pop and hard or massive new capacity to come online. In the first case we won't have money to buy and the second - we all will be graybeards.
I call bullshit – the "article" also stinks of AI slop.
As long as the US/Israel stuff is going on in the middle east, there's going to be no downwards movements in any prices, SSDs or otherwise. End of story.
Why ?
For a start, before you consider anything else, stuff needs to be shipped around. For that you need (a) Oil (b) The ability to use shortest-route transport routes.
The middle east nonsense messes all that up completely. And even if the war stopped tomorrow, it will still take some time for things to return to some semblance of normality, such is the extent of the disruption.
No, that hardly matters.
The other day my firm was looking to ship about 180kg of expensive materials packed into 9 large crates from Korea to the US -- which, if it were SSDs, would be a lot of SSDs.
You're absolutely right that it's much more expensive than ever, what with the war and the holidays and all. But we were still quoted just $2600 for Air Freight and $1300 for Sea Freight. Which, if it were SSDs, would work out to pennies per unit. (For expensive goods, tariffs are a much more punishing cost.)
It's demand driven. The Iran War might raise prices more by cutting off Asia from helium used in chip production, rather than by driving up oil prices...
> might raise prices more by cutting off Asia from helium used in chip production,
Which is why I used the words "for a start".
To spell it out, I was saying oil was ONE THING, not the only thing which is affected by the stupid war.
Please don't try to tell me the war doesn't affect RAM or SSD production or shipment in the slightest. I think we all know that's not the case.
imho i think alot of prices for things like shipping are a gouge and have nothing todo with iran war. look at what happened during covid when every company with supply of anykind tried to leverage their position to gouge everything and everyone they could.
the iran war itself is a gouge on the american taxpayer by the defense industry, if war ends , defense indutry gravy train goes back to normal. sad how the US spent so much money on bombs, and yet the war continues... do they not know where to drop the bombs to end the war?!, seems like they drop billions in munitions in the middle of the desert
Shipping cost affects cheap (compare to mass/volume) products but in RAM price shipping is a negligible fraction. Where Iran war affects the more AFIK is the cost of credit which slows investments into new capacity by the manufactures. All thing being equal low interest rates allow to expand capacity faster.
Most SSD/RAM and chips generally ship by air. At the value density of such things, nobody wants to maintain stock sitting on ships. Interest rates alone would cover the cost of air cargo. Just think of how much ram can be on a single pallet, let alone in a 40-foot shipping container.