mgh95

You're talking about roughly ~5T (the prior estimates I'm aware of) in costs for replacing the grid with a HVDC transmission line. That's a massive cost and undertaking. It's much more likely we see an incremental transition to point-to-point transmission in major industrial corridors (Los Angeles/Irvine to IE solar generation capacity in the future, Lousiana Basin, North Texas to the gulf, etc.) where major benefits can be cost effectively be realized.

A "full" transition is unlikely in our lifetimes due to the fact that the majority of the benefits can be reaped without needing such an expense.

> Or the wind generators across the continent if you have to.

Transmission losses are typically very substantial in most grids that are AC based. For example, a cross-country power transmission with the USAs grid would result in ~36% losses (napkin math at about 20% loss per 1000km).

Reality isn't as simple as "ship the electricity" unfortunately; it makes a lot of sense to keep generation near consumption.

Edit: Since people like this comment, take a look at this: https://patternenergy.com/projects/southern-spirit-transmiss...

Reading this makes me angry.

C# has its own share of warts when it comes to AOT -- it doesn't yet support OpenBSD; it does not yet support all reflection oriented keywords; it requires significant changes to msbuild to work with in a standalone context.

But this article reads like someone with minimal background in software went on a LLM prompted delusion and lacks an understanding of the C# language architecture, associated build system, and philosophy. Additions of keywords like the proposed "overspace" attempting to shoe-horn both inappropriate approaches and ideas are a plague on the community.

> Cloud providers are currently “un-differentiated”, but there are three huge ones making profits and some small ones too. Hosting is an economy-of-scale business and so is inference.

Anybody who has worked in a compliance heavy segment (PCI-DSS, HIPAA, etc.) will tell you the big 3 clouds have very significant differences from the smaller players. The differentiation is not on compute itself, but on the product. It's partially why products like AWS Bedrock exist and are actively placing model providers both in competition with eachother and AWS itself which is exactly the market dynamic they should seek to avoid.

> The idea that OAI is un-differentiated is just weird. They have a massively popular consumer offering, a huge bankroll, and can continue to innovate on features. Their consumer offering has remained sticky even though Claude and Gemini have both had periods of being the best model to those in the know.

This is exactly where this line of reasoning goes off the rails. The consumer market is problematic (see the recent post about the segment its growing in; basically young women of limited spend in low income countries); a huge bankroll is also a huge liability, model providers are on a clock to get huge or die, and the innovation we are seeing is effectively attempting to "scale-up" models, not provide novel features.

> Their consumer offering has remained sticky even though Claude and Gemini have both had periods of being the best model to those in the know.

This isn't a good thing with current market mix.

> And generally speaking there are huge opportunities to do enterprise integrations and build out the retooling of $10T of economic activities, just with the models we have now; a Salesforce play would be a natural pivot for them.

Do you have any indication these are achieving buy in or profitable? Most significantly, we have seen a recent study by MIT that 95% of generative AI pilots fail. The honeymoon period is rapidly coming to a close. Tangible results are necessary.

> I don't disagree but you're moving the goalposts. I never said that they could achieve the profits of a typical tech business, just that they could be profitable. Also, the whole distilling problem doesn't happen if the model is proprietary.

In the absence of typical software margins, they will be eroded by providers of "good enough" margins (AWS, Azure, GCP, etc.) who gain more profit from the bundled services than OpenAI does from the primary services. This has happened multiple times in history, either resulting in smaller businesses below IPO price (such as Elastic, Hashicorp, etc.) or outright bankruptcy.

Second, the distilling happens on the outputs of the model. Model distillation refers to the usage of a models outputs to train a secondary smaller model. Do not mistake distillation for training (or retraining) to sparse models. You can absolutely distill proprietary models. In fact, that is how DeekSeek-R1-Distill-Qwen and the DeepSeek-R1-Distill-Llama are trained. This also happens with Chinese startups distilling OpenAI models to resell [2].

The worst part is OpenAI is already having to provide APIs to do this [1]. This is not ideal, as OpenAI wants to lock people into (as much as possible) a single platform.

I really don't like OpenAIs market position here. I don't think it's long term profitable.

[1] https://openai.com/index/api-model-distillation/

[2] https://www.theguardian.com/technology/2025/jan/29/openai-ch...

Amazon was gross margin profitable -- and significantly so -- the entire time.

It just turns out they were a server farm subsidizing a gift shop.

This only applies if you think one of two things; First, that it is guaranteed that this specific line of inquiry will lead to development of a form of superintelligence or otherwise broadly applicable development; or second, the form of machine learning technologies that unlocks or otherwise enables a market which would otherwise be inaccsesible that justifies this investment.

To date, no evidence of either even exists. See Zuckerbergs recent live demo of Facebooks Ray Bans technology, for example.

I think it's possible we are in datacenter GPU overcapacity already, and NVIDIA is burning its stock to avoid the music stopping.

It's already happening in China that datacenters are at GPU overcapacity. I wouldn't be surprised if it occurs here.

> OpenAI is profitable if they stop training their next generation models. Their unit economics are extremely favorable.

But OpenAI can't stop training their next generation models. OpenAI already spends over 50% of their revenue on inference cost [1] with some vendors spending over 100% of their revenue on inference.

The real cash cow for them is in the business segment. The problem here is models are rapidly cloned, and the companies adjacent to model providers actively seek to provide consumers the ability to rapidly and seamlessly switch between model providers [2][3].

Model providers are in the situation you imagine cloud providers to be in; a non-differentiated, commodity product with high fixed costs, and poor margins.

[1] https://www.wheresyoured.at/why-everybody-is-losing-money-on...

[2] https://www.jetbrains.com/help/ai-assistant/use-custom-model...

[3] https://code.visualstudio.com/docs/copilot/customization/lan...

"Marginally ROI positive" works in a ZIRP environment. These are huge capital investments; they need to at least clear treasury return hurdles and importantly provide attractive returns.

I am fundamentally skeptical of "scaling inference". Margins are not defensible in the market segment OpenAI is in.

Currently, selling LLM inference is a red queen race: the moment you release a model, others begin distilling and attempting to sell your model cheaper, avoiding the expensive capitalized costs associated with R&D. This can occur because the LLM market is fundamentally -- at best -- minimally differentiated; consumers are willing to switch between vendors ("big labs", as you call them, but they aren't really research labs) to whomever offers the best model at the lowest price. This is emphasized by the distributors of many LLMs, developer tools, offering ways to switch the LLM at runtime (see https://www.jetbrains.com/help/ai-assistant/use-custom-model... or https://code.visualstudio.com/docs/copilot/customization/lan... for an example of this). The distributors of LLMs actively working against LLM providers margin provides an exceptionally strong headwind.

This market dynamic begets a low margin race to the bottom, where no party appears able to secure the highly attractive (think the >70% service margin we see in typical tech) unit economics typical of tech.

Inference is a very tough business. It is my opinion (and likely the opinion of many others) that the margins will not sustain a typical "tech" business without continual investment to attempt to develop increasingly complex and expensive models, which itself is unprofitable.

I don't believe I am. Investors (value investors, not pump and dump investors) provide capital to companies on the expectation of profit, not revenue.

Revenue != profit, and you don't need to become net negative margin to be net unprofitable. Expensive researchers, expensive engineers, expensive capex, etc.

Inference has extremely different unit economics from a typical SaaS like Salesforce or adtech like google or facebook.

What revenues do these GPUs generate for OpenAI? OpenAI is not currently profitable, and it is unclear if its business model will ever becomes profitable -- let alone profitable enough to justify this investment. Currently, this only works because the markets are willing to lend and let NVIDIA issue stock to cover the costs to manufacture the GPUs.

That's where the belief that we are in a bubble comes from.

I have a feeling if you hire more Indians in India (which is already coming out to ~40$ per hour billing rate) you are going to be stuck with the problem that you now have the typical Indian outsourcing problem which is why companies want them locally: to keep an eye on them.

When the C-suite moves to India, I'll believe it.

> Theres the little matter of the US imposed Plaza Accords that have been skipped in the timeline there , which are widely acknowledged to have precipitated Japan's lost decades.

Actually, no. The CPC generally argues this is the case, but it is largely incorrect. Japan responded to the Plaza accords by attempting to massively increase stimulus and preserve the structure (significant industrial, limited service/consumption economy) they started.

In contrast, (West) Germany kept the stimulus relatively low and rotated into higher productivity enterprises and R&D and had a substantially lessened impact. See this (https://www.atlantis-press.com/article/125961733.pdf)

The whole "US destroyed Japan with the Plaza accords" is basically CPC propaganda.

China will have its own issues, but I would guess they look more like the 1992 sterling crisis where China cannot hold its yuan peg as it continues to have trade expansion without corresponding foreign exchange growth.

Yeah the comment was aimed at people who think like this. I think you should spend more time in the physical world.

Ask any business if they care about revenue or customers. $ is the goal.

Ok. I think you drink the coolaid of China far too much. The problem with Chinas production has much more to do with the fact it has a substantial imbalance in consumption and production. This is what is at the root of why the involution crackdown has by in large failed to yield results (.4% contraction CPI; 2.9% PPI); countries are increasingly shutting out the state sponsored production endorsed by the CPC in favor of domestic or better balanced trade partners.

Without the large population not to perform "dirty jobs" but to participate in Chinese society to generate domestic demand, it is unlikely that China will continue to require ongoing stimulus just to keep the economic model functioning. See this (https://www.omfif.org/2025/03/china-has-just-raised-its-debt...) regarding the increase in debt held by China and its localities.

This need for "permastimulus" just to keep the economic model working is the problem: China needs to be rebalancing its economy away from production and into consumption. Unfortunately, a declining population also has declining consumption.

Oh sure. But the important thing -- especially for the OC, who is having a rough time in the job market -- is getting paid to do it.

I think the combination of birth rates in China and Russia banging on the door of the EU is going to help the US more than H1Bs for the US. With current demographics, the US is slated to be more populous than China at the turn of the next century.

In your opinion, what is false about it? Canada's growth segment in exports was minerals and tourism services. That's Canada's situation, unfortunately.

I don't know what to tell you but on the front of:

> There are plenty of other places to travel is the rationale. besides, if a strong greenback was the reason for decrease of leisure, wouldn’t it also be responsible for a decrease in business travel, too? Certainly businesses are also bound to macroeconomic shifts.

Canada has shockingly little choice when it comes to trade partners. They are literally physically attached to the US and trade is much simpler when working with the US, whether Canada likes it or not. CETA hasn't yet been ratified; Canada has bungled trade with China since harper; Mexico is at best a cheap labor destination that can replace India for Canada. The only real staying power Canada has is exporting raw materials, and even that effectively turns Canada into a resource extraction colony. That's not a happy ending, either.

Canadian businesses have fewer options than many would like to admit, so it makes sense they are keeping up their economic activity with the US.

Sorry - I mean the "only way" as a manager, not as a software/technical professional.

You can't for example, use manager of a McDonalds to qualify as a manager under a TN as far as I understand it. If you are a manager, you need to be technical in nature.

Ok; all I can say is what the statistics indicate. It seems like at least some Canadians accept assignments which result in business travel to the US. As with the OC, this may result in more desirable employees to Canadian employers who wish to continue to do business with the US, hence the original question.

Business travel from CAN to US has remained stable (see: https://globalnews.ca/video/11436758/business-travel-to-u-s-...). It's leisure that is declining.

Meanwhile, total spending in the US from int'l visitor tourism is up in the US (see: https://www.hotel-online.com/press_releases/release/internat...).

Honestly, I think macro factors -- namely, poor Canadian household finances due to increasing cost of living and declining real incomes in Canada coupled with a strengthening US dollar against the Loonie -- are what are killing tourism from CAN to US right now.

The CV info you posted basically says upper management. I'm seeing increasingly that companies centralize management in an office (either city satellites or a hq) and permit developers to either work remote within the geographic region with occasional onsites. I don't know of anybody that would permit a manager (much less someone that is a sr. manager) to be remote which is why I asked.

Not really. I even live in one of the cities that is "under siege". It's basically another day.

People read headlines, lock themselves in a cage of their own making, and assume the world is on fire. People would do better trusting their own eyes and ears.

Well, that's a personal decision. And personal decisions can have financial costs.

Yeah. That's actually the only way to get a TN -- to oversee technical work in a managerial capacity. Technical managers of software engineers are 100% able to enter under TN.