Sounds like misinformation, let's compare: "The clothing and textile industry uses roughly 170 times more."
The difference is even larger if you look at the state of the art: "As of November 2026, the state of the art is direct-to-chip liquid cooling combined with closed-loop dry coolers. This allows a data center to operate with essentially zero water consumed for cooling, even for high-density AI infrastructure."
You might not want to put an older/less efficient data center in a very dry area, but other than it seems quite reasonable.
That makes sense, anthropic is in a rapidly changing market. Revenue could crater if a better model is released by a competitor. AI demand could fall behind. Doesn't seem like that will happen, but its definitely not investment-grade debt worthy either.
Even if Anthropic invents a godlike frontier model that can do anything, it feels like they won't have a moat for very long before competitors like DeepSeek offer something almost as good for a fraction of the price.
If AGI is created, then it will be used by someone else to create a cheaper AGI, making the finances of the person who borrowed trillions to create it in question.
An extremely hypothetical experiment but what if this is some sort of MAD (Mutually Assured Destruction) and in our timeline, we have accepted it with AI.
Deepseek can do MAD on Anthropic and release open weights model (which means that they might not earn the most or basically burn millions/billions to then give us all essentially the same thing and we would all be on equal footing, oh they also released their inference engine right now.)
This might explain why the stock market got into an bubble and is fearing itself right now.
Mutually Assured Commoditization. (Credit where it's due I pasted my comment to Chatgpt and it gave me this term ironically and it seems to fit well to the comment that I am writing, I might write a blog about it someday.)
MAD / MAC assumes everyone has the same leverage on each other- that they are all as heavily invested in capital.
Where it falls apart is when you get companies that take open models, rent compute and sell tokens at a profit, without heavy investment into research or training. They spend way less, and if they can keep distilling the frontier models, they can produce reasonable products without the massive losses of subscription plans. In your example, DeepSeek et al aren't nearly so vulnerable. There's nothing mutually assured.
It is much more of a classic race to the bottom. Every new frontier model shifts the goal posts a bit, but eventually the house of cards will crumble - or at least, that's what the market seems to be thinking will happen.
Are you trying to say "the Anthropic API surely cannot be highly marked up! Therefore subscriptions surely are selling at a loss!"? If not, I'm not sure how your comment is particularly relevant to the discussion?
I mean, it's worth pointing out because either API pricing is crazy marked up and thus subscriptions are still profitable if they are used at full amount or subscriptions are subsidized and API pricing is more true cost. Sure, it could be second one but due to average usage, it's profitable overall.
Right. Either way, there's nothing at all to support that subscription plans are "massive losses" except the original commenter's imagination. A subscription/API price delta does not a massive loss make.
This moat issue comes up a lot, but look at what's been happening so far; open models are consistently 6-12 months behind. What if their moat is simply staying ahead?
And once their models start to accelerate their progress, which may already be happening, this would further solidify their lead.
Demand is currently not sustainable because they - ant and OAI - refuse to raise prices.
Infact they’re going to end up in price competition unless they field an army of ‘FDE’ who go to every large firm and reorganise the operations around their models and displace labour.
I was under the impression this was only on the consumer side? I have heard of contracts expiring and them wanting a lot more money now (forcing from generously subsidized enterprise seat plans to all API usage). Specifically Anthropic in this case.
GitHub copilot then comes in and offers multi provider support including open source models, and gets the contract due to Microsoft’s existing establishment and lower price point.
Anthropic can’t raise their prices without loosing market share of Claude code
Every time it gets posted most assume that the forest was supposed to be protected and they decided to cut it down anyway.
The details are much less dramatic: The company doing the work messed up on a part of the approval process that was supposed to be completed first. It’s technically possible that the environmental impact review would come back with some obscure exception that required them to revise their plans, but it’s unlikely that anything was going to stop those trees from being cut down.
The 250K trees headline appears dramatic if you don’t know that Finland is covered in trees and cutting down a lot of them is part of building anything.
In the US there's also a third category besides "permit approved" and "permit denied", "no permit needed": the thing you are doing is allowed and not covered by the permitting process. You just do it. This shows up in both building code (you typically don't need a permit for most repairs or cosmetic renovations) and environmental protections (you don't need an environmental study for logging in a non-protected forest).
The tricky part with "no permit needed" is that while you can just go and do whatever, if someone charged with enforcement suspects you're doing something that requires a permit, they can still demand you prove you don't (and stop work until you do), and if you guessed wrong and you did indeed need a permit, you're double-fucked.
That's why for many of these processes, there's also something like a certificate of no-permit needed, where you go through an abbreviated permitting process spending a lot of time and money while all of the clerks and officials stare at you like you're a moron before finally giving you a piece of paper saying you didn't need to do any of that in the first place.
And that's why a lot of people with a lot of experience and a good understanding of the law skip getting a "no permit needed" certificate.
It also doesn't necessarily protect you from anything by obtaining it.
If you, say, go and get a certificate of no-permit needed when you're replacing all of your kitchen cabinets, and then you decide to move the sink while you're at it, you're back to still needing a permit for the plumbing change (and even if you aren't, if your neighbor calls the inspector and says they think you might be, the inspector might say, "no, we already know about that, it's just cabinets", or they might wonder if you are and then you're back in the situation you'd be in without the certificate).
A friend worked in the US tree cutting permit process. There were two main goals behind the permits (as I understand it):
1) To double check the harvest was not going to affect things outside their property (neighbors, land slides, erosion, etc), or special govt. interests on their property (endangered species)
2) To maintain their significant tax break. The break was for owning forest land, and you needed a plan to replant successfully.
Keep in mind that these are mostly 'working forests' that are more akin to wheat fields than wild forests.
It sounds like the issue was insufficient analysis of #1, resulting in excessive water run-off to neighboring privately owned property.
That being said, it could also be #2. The same friend worked on 'conversion' permits, wherein the owner asked to turn forest land into a residential or business property, and special considerations had to be addressed.
• You wouldn't need this kind of permit for clear-cutting 3.3 sq-km of forest, if the forest stays as forest (i.e. you will plant with tree seedlings)
• You need this permit when the land use changes (from forest to build area)
• The subcontractor might not have known that they are clear-cutting without all permits, or at least they might have plausible deniability
• The decision makers of the town of Muhos want the data center, as it bring jobs and taxable income
• The town building inspector had given a green light for Google to proceed
• Finland is, still, a high-trust society. Punishment for this kind of wrong-doing is likely lenient (some relatively small fines), the shame of having been sentenced in court is supposed to be the main deterrent in itself
• Thus, it might make financial sense for Google to cut corners and proceed faster, even if it means they end up paying some fines
• Unless the bad publicity ends up costing more than the saved project time
Google (through its local subsidiary) asked the regulators if they could skip doing an environmental impact assessment. The regulators told them that they could not skip doing it. They'd already started logging before asking for the exemption and continued to log for months after being told that there would not be an exemption.
They'd also applied for a landscaping permit for the site, but were denied one by the municipality because they'd still not performed the environmental impact assessment.
It's google trying to ask for forgiveness rather than permission, the news cycles making money on the cockup, and HN whiteknighting for Google because Google pays their wages directly and indirectly.
Yeah, and it's being done by a local contractor who presumably knows the process. It'll definitely be interesting if some moustache-twirling CEO at a foreign company is shown to have yelled "BURN THE FOREST DOWN, I WANT IT NOW!" at the contractor, but that feels unlikely in this case.
That's the first thing that comes to mind here too. Very very very few contractors are going to do work without a permit. Maybe Finland needs to investigate their own government offices?
Funny thing is, if this "forest" is to be protected, why doesn't the govt buy/own the land? What an idiotic thing to say you plant trees in your land but you are not allowed to cut it. Don't approve of a data center in a land where you won't let them clear it. Bureaucracy must prevail!
Significant soil depth can be part of the solution, which no doubt requires additional strength in the structure underneath the roof..
For a worked example:
"The 5.4-acre green roof has soil depths that vary from 18 inches to 4.5 ft. This layered planting system is a structurally engineered component of the building that functions to balance seismic shifting, collect and filter stormwater, and irrigate the gardens. Geosynthetic fill was used to build up the rolling topography of the park while adding minimal weight to the structure."
Yes, Claude + skills is what we use too. Big picture, all of the old BI players are living on borrowed time. If you have to drag and drop; you'll be dropped.
Remote with that timezone difference seems tricky to find something. Maybe consider starting something on your own? Pro: You can be remote, and dont need to apply, also uncapped upside. Con: It's hard.
I've met a guy who went from software into being a sherpa. Taking people up the mountain for hikes. Kinda fascinating.
But this post is not representative of the state of the market. The market for tech talent is still very hard/scarce. Finding good people is very difficult. You can validate that by seeing the trend of software salaries. It's not coming down.
One day a deer friend, and brilliant engineer, called me. He asked if he could come see me at my office in San Francisco. Surprised at what this could be I simply said yes. He arrived, we chatted. He showed me his plane ticket. He quit his job that day, and that evening he was flying abroad to become a monk. Haven't talked with him in a decade. Hope he is doing well!
The short answer is that those were conceived during economic boom times before the median American was sold down the river with outsourcing, financialization, and good old fashion corruption. 50 years after this crime was commited against the people, the ideals espoused in those fictions has become hollow and sour.
Star Trek is fiction, a hopeful vision of a far future and not a description of the present. ToS was also dark and verging on a horror series most of the time. The rampant optimism of the 1990's and TnG are NOT found in ToS. The Federation of ToS isn't a utopia or implied to be one, it's just more egalitarian than the present at the time which was defined by generational, racial, and economic fighting. It took place during the context of a war that claimed tens of thousands of American lives, most drafted from the ranks of our youth.
It's very easy to imagine a past defined by fictional narratives and idealized memories of the Cold War. Don't give in to nostalgia, it's a lie.
Neither Star Trek nor the Apollo program were popular at the time. The original Star Trek was cancelled after three seasons with no further content produced until a movie revival a decade later; contemporary polls of the Apollo program consistently showed that most people thought it was a waste of money. If (if) SpaceX and Tesla deliver on Musk's dreams, I expect people in the 2050s will wax nostalgic about the 2020s and complain that we can't summon the same spirit of adventure for the Dyson spheres Claude wants to build.
He lied about the Cybertruck being faster than it is. It's fast anyway. There was no need to lie about it but Musk is pathological. He simply can't help himself:
My memory is the media was pretty negative back then too. You have to filter out the political person A says bad stuff about B shouts at C, unfair to D which is constantly in the news in all time periods and look at the cool things going on instead.
One of the good things about Musk is at least he has cool sounding projects like switching to carbon free transport, reusable rockets, cities on Mars, amazing abundance etc.
Star Trek was pretty negative about this period in history.
In that universe, we had the Eugenics War/World War III during this time period. The most depraved and bloody war in Earth’s history that wiped out 1/3 of Earth’s human population and rendered hundreds of thousands of species extinct.
I just looked up - apparently WW III starts in 2026.
>In 2026, at the start of the war, Colonel Phillip Green led a faction of ecoterrorists that was responsible for the loss of thirty-seven million lives.
The difference is even larger if you look at the state of the art: "As of November 2026, the state of the art is direct-to-chip liquid cooling combined with closed-loop dry coolers. This allows a data center to operate with essentially zero water consumed for cooling, even for high-density AI infrastructure."
You might not want to put an older/less efficient data center in a very dry area, but other than it seems quite reasonable.
reply