It's reasonably likely that a lot of people linked to the federal government want to ban DeepSeek. You can tell it's being presented away from "they gave us a free set of weights" and towards "they destroyed $1T of shareholder value." (By revealing that Microsoft et al. paid way too much to OpenAI et al. for technology that was actually easy to reinvent.)
> "they destroyed $1T of shareholder value." (By revealing that Microsoft et al. paid way too much to OpenAI et al. for technology that was actually easy to reinvent.)
The value was highly speculative, an illusion created by PR and sentiment momentum. "Hype value" not real value (unless you're able to realize it and dump those bags on someone else before fundamentals set in). Same thing happening with power companies downstream of the discovery that AI is not going to be a savior of sagging electricity demand. Overdriving the fundamentals is not value destruction, it is "I gambled and lost."
What they really destroyed was the idea that OpenAI would be able to charge $200/month for their ChatGPT Pro subscription which includes o1. That was always ridiculous IMO. The Free tier and $20/month Plus tier along with their API business (minus any future plan to charge a ridiculous amount for API access to o1) will be fine.
> The Free tier and $20/month Plus tier along with their API business (minus any future plan to charge a ridiculous amount for API access to o1) will be fine.
Actually no! If we take their paper at face value, the crucial innovation to get a strong model with efficiency is their much reduced KV cache and their MoE approach:
- where a standard model needs to store two large vectors for each token at inference time (and load/store those over and over from memory) deepseek v3/R1 only stores one smaller vector C that is a „compression“ from which the large k,v vectors can be decoded on the fly.
- They use a fairly standard Mixture of Expert (MoE) approach, which works well in training with their tricks, but whose inference time advantages are immediate and equal to all other MoE techniques, which is to say that from ~85% of the 600B+ params that are inside the MoE layers, the model at each token inference step will only pick a small fraction to use. This reduces FLOPs and memory io by a large factor in comparison to a so-called dense model where all weights are used for every token (cf Llama 3 405B)
The two podcasters who do the Acquired podcast spoke to Ballmer about some of Microsoft’s failed initiatives and acquisitions. He told them that at the end of the day “it’s only money”.
All of the BigTech companies have enough cash flow from profitable lines of business to make speculative bets.
It must be EZ mode to be a big tech executive, you somehow have all the power to make every decision while also having the ability to never take the fault for these decisions.
I would much rather have a company with a culture that isn’t afraid to take calculated risks and not be afraid of repercussions when they take risk as long as it doesn’t cause consumer harm.
"Not doing consumer harm" is carrying a lot of weight there.
Either way what you describe is perfectly achievable for the workers, but at some point management needs to own up to their failures and getting rewarded because the board is also made up of executives at other big tech companies is a perverse incentive to never actually improve.
I mean forcing copilot everywhere I don't want it (nowhere) while jacking up prices to justify it and using Windows 11 to serve ads is harmful to me. There's also you know... the anticompetitive company that thinks buying new sectors is healthy.
What does that have to do with what I'm saying? Today Copilot is being shoved in services I don't want which they are then in turn using to justify cost increases.
How is that not consumer harm?
Hopefully in the future the FTC will break up Microsoft, forcing them to split Azure, Office, and Windows. They clearly can't be trusted with all 3.
Since the time when companies en masse stopped paying cash dividends on owned shares, the value has become highly speculative. In the absence of dividend payments, the stock pricing mechanism is not essentially different from Solana or Ethereum "price" discovery.
I don't disagree that price discovery is harder, but I can with more certainty give an honest valuation of CLF or DOW vs OpenAI's "who knows what money will look like after we succeed, you should view your investment as a donation" nonsense. Speculation is inevitable when forward looking, but there is a difference between error bars and various projections vs unicorns.
> when companies en masse stopped paying stock dividends
Do you mean cash dividends [1]?
Also, the premise is false. Dividend yields have roughly tracked interest rates [2]. (The difference is a dirty component of the equity risk premium [3].)
I changed the typo, thanks. Chash dividends. This analysis does not negate common sense: when a company does not pay cash dividends, owning its stock is purely speculative, like owning Solana. When it does, you get cash dividends funded by the company's tangible revenue, proportional to your number of shares.
I saw a some Europeans hoping that the US would ban DeepSeek, because then there would be less traffic interfering with their own DeepSeek queries.
The US can ban all they want, but if the rest of the world starts preferring Chinese social media, Chinese AI, and Chinese websites in general, the US is going to lose one of its crown jewels.
The way the US behaves is a problem and makes a lot of people prefer alternatives just for the sake of avoiding the US, which is why it's important that the US get along with other nations, but--well, about that...
Agreed, you've highlighted one of the key problems with protectionism and nativism. Banning competition just weakens America's global influence, it doesn't make it stronger.
This statement doesn't seem to hold true. China has banned nearly all US tech companies and social products. It has not decreased the influence of China's influence (which has been through manufacturing/retail influence and tech influence).
I don't think your statement holds with current behavior.
> China has banned nearly all US tech companies and social products. It has not decreased the influence of China's influence
Being hostile does not bring you friends. Sure, various countries can have reasons to suck it up anyway (e.g. because of sanctions, or because China makes an offer too good to pass, although even that comes with strings attached). But in the long run you just create clients or satellites who will escape at the first occasion.
The American foreign policy around the middle of the 20th century relied very effectively on soft power, which is something you can leverage to get much more out of your investments than their pure monetary value. It is not required in order to gain influence, but it is a force multiplier.
Then how can you explain that China’s hostility towards Western tech companies being present inside their own country has not created what you’re describing?
Is hostility a bad idea only for America? Sure hope not.
> Is hostility a bad idea only for America? Sure hope not.
I think protectionism is long-term bad for every country, but it's especially and uniquely bad for the biggest economy in the world who has net benefitted the most from free trade and competition. There's no denying that China is influential – the argument is that they could've been (and still can be) so much more influential by embracing western tech instead of walling themselves off.
America is reliant on purchasing cheap goods from elsewhere and selling expensive technology. If it’s hostile toward the suppliers of cheap goods or the buyers of expensive technology, well, what purpose does it have on the global scale?
I am saying that they could have got much more, particularly considering the spectacular mistakes western countries kept making for the last ~2 decades.
But China has never been a global leader in tech or social media. They undoubtedly have influence in these areas, but they've never dominated them like the US has. Banning foreign competition in a field where you already dominate, like tech and AI, has different consequences than banning it where you're playing catch up.
What is your definition of "tech"? A very large amount of the electronics products in the world are made in China (specifically in/around Szhenzen and the wider Guangdong province). Both consumer goods and industrial goods. From the cheapest stuff to the most advanced and everything in between. They provide the manufacturing for brands fron all over the world, including goods "from the west".
The amount of economy that depends entirely on this low-cost, high-quality manufacturing is insanely large - both directly in electronics goods but also as part of many other industries because you need electronics to build anything else.
By "tech" I'm sort of vaguely handwaving at Silicon Valley et al. I agree that China has built up a massive manufacturing industry that the west depends on, but I don't think that "being a significant cog in the machine," so to speak, buys as much influence or bargaining power as being the maker or owner of the machine. It's better to have the Apples and Googles of the world than it is to have the SG Micros or BYD Electronics.
American consumer and industrial electronics companies are increasingly unable to deliver products without the Chinese supply chain. How does that not give significant bargaining power? Also factor in that the Chinese manufacturers also manufacturers for everyone else in the world, so they don't have to sell that capacity to USA. And that the share of production capacity that companies from America use is trending down anyways. Mostly due to Asia, Middle East and South America are still growing a lot. Then Africa is following, delayed by some decades.
Of course owning the end customer is generally better. But moving production is not something to take lightly.
I've recently cancelled my Github Copilot subscription and now use Mistral. When the US starts threatening allies with tariffs or invasion, using US services becomes a major business risk.
Not only a business risk. It also becomes a moral imperative to avoid if you can. Don't support bullies, is my motto. It can be hard to completely avoid, but it is important to try.
And the path to pleasing the EU would be straightforward: make a EU subsidiary, have it host or rent GPUs and servers in Europe, make sure personally-identifiable data is handled in accordance with GDPR and doesn't leave that subsidiary, make sure EU customers make their accounts with and are served by that subsidiary.
Meanwhile, to please the US they would probably have to move the entire company to the US. And even that may not be enough
Banning the site would be fine. The model itself will still be available from a variety of providers as well as locally. The US is more likely to ban the model itself on the basis of national security
Theoretically this should be good for OpenAI - in that they can reduce their costs by ~27x and pass that along to end users to get more adoption and more profit.
I wish more people had understood that spending a lot of money processing publicly available commodities with techniques available in the published literature is the business model of a steel mill.
It’s the business of commodities. The magic is in tiny incremental improvements and distribution. DeepSeek forces us to question if AI—possibly intelligence—is a commodity.
No, but it's good enough to replace some office jobs. Which forces us to ask, to what degree is intelligence--unique intelligence--required for useful production? (We can ask the same about physical strength.)
I find it interesting that so much discussion about “LLM’s can do some of our work” is centred around “are they intelligence” and not what I see as the precursor question of “are we doing a lot of bullshit work?”
My partner is in law, along with several friends and the amount of completely _useless_ work and ceremony they’re forced to do is insane. It’s a literal waste of their talent and time. We could probably net most of the claimed AI gains by taking a serious look at pointless workloads and come out ahead due to not needing the energy and capital expenditure.
Surely that would be amazing for NVDA? If the only 'hard' part of making AI is making/buying/smuggling the hardware then nvidia should expect to capture most of the value.
No. Before Deepseek R1, Nvidia was charging $100 for a $20 shovel in the gold rush. Now, every Fortune 100 can build an O1-level model with currently existing (and soon to be online) infra. Healthy demand for H100 and Blackwell will remain, but paying $100 for a $20 shovel is unlikely.
Nvidia will definitely stay profitable for now though, as long as Deepseek’s breakthroughs are not further improved upon. But if others find additional compression gains, Nvidia won’t recapture its old premium. Its stock hinged on 80% margins and 75% annual growth, Deepseek broke that premise.
There still isn't a serious alternative for chips for AI training. Until competition catches up or models become so efficient they can be trained on gaming cards Nvidia will still be able to command the same margins.
Growth might take a short-term dip, but may well be picked up by induced demand. Being able to train your own models "cheaply" will cause a lot more companies and departments want to train their own models on their own data, and cause them to retrain more frequently.
The time of being able to sell H100 clusters for inference might be coming to an end though.
Maybe they even suppressed algorithmic improvements in their company to preserve moat. Something akin to Kodak suppressing internal research on digital cameras because they were world leading company that produced photo film.
You're fooling yourself if you think OpenAI is going to pass up implementing the same strategies to get a ~27x cheaper model.
> Unlike a social network, network effects won't help them - their users don't care how many other users they have, only about the AI output quality.
Google Search doesn't have a network effect. Everyone on HN has been saying Google Search is complete garbage for a decade. It still has the same market share (roughly) as it did a decade ago.
Not directly. The 27x is about costs. What it means is some order of magnitude of more competition. That reduces natural market share, price leverage and thus future profits.
Valuations are based on future profits. Not future revenues.
You can theoretically lower your costs by 27x and end up with 2x more future profits - if you're actually 45x cheaper (which DeepSeek's method claims to be).
You mean charge a 27x lower price, but have 45x lower costs, so your profit margin has doubled?
Your relative margin may have doubled, but your absolute profit-per-item hasn't. Say you had a 10% margin before, at a $100 price and $90 cost, for a $10 profit-per-item. Reduce price 27x and cost 45x, so $3.7 price, $2 cost, and $1.7 profit-per-item. 6x less profit - not as bad as 27x, but not good if you're OpenAI.
> Your relative margin may have doubled, but your absolute profit-per-item hasn't.
ChatGPT doesn't have any profits right now.
We have no idea what investors are expecting future profits to be.
> Say you had a 10% margin before, at a $100 price and $90 cost, for a $10 profit-per-item. Reduce price 27x and cost 45x, so $3.7 price, $2 cost, and $1.7 profit-per-item. 6x less profit - not as bad as 27x, but not good if you're OpenAI.
Now do the same thing but assume you have 10x more subscribers because the prices are ~27x lower.
You end up with almost 2x more total profit.
Just take ChatGPT's ~$200 subscription. Hardly anyone is going to pay ~$200 a month. Reduce that by 27x - and you're at $7.5 per month. Maybe 10% of people on the planet will pay that.
> Now do the same thing but assume you have 10x more subscribers because the prices are ~27x lower.
You're in various spots of this thread pushing the idea that their 1B MAUs make them unassailable. How are they gonna get to 10B in a world with less than that total people?
> Just take ChatGPT's ~$200 subscription. Hardly anyone is going to pay ~$200 a month. Reduce that by 27x - and you're at $7.5 per month. Maybe 10% of people on the planet will pay that.
if ChatGPT starts selling ads on chat results that will probably improve revenue. I've seen social media ads recently for things I've only typed into ChatGPT so that leads me to believe they're already monetizing it to advertising platforms.
Google spends immense amounts of resources every year to ensure that their search is almost always the default option. Defaults are extremely powerful in consumer tech.
> Google Search doesn't have a network effect. Everyone on HN has been saying Google Search is complete garbage for a decade. It still has the same market share (roughly) as it did a decade ago.
It absolutely does. People use Google for search -> Websites optimise for Google -> People get “better” results when searching with Google.
The fact that it’s market share is sticky and not responding quickly to change in quality is sort of indicative of the network effect.
It probably counts pretty much anyone on a newer iPhone/Mac (https://support.apple.com/en-au/guide/iphone/iph00fd3c8c2/io...) and Windows/Bing. Plus all the smaller integrations out there. All of which can be migrated to a new LLM vendor... pretty quickly.
« The thing I noticed right away when Claude came out is how little lock-in ChatGPT had established. This was very different to my experience when I first ran a search on Google, sometime in the year 2000. After the first time I used Google, I literally never used another search engine again; it was just light years ahead of its competitors in terms of the quality of its results, and the clarity of its presentation.
This week I added a third chatbot to the mix: DeepSeek »
My guess is that OS vendors are the real winners in the long run. If Siri/Goolge can access my stuff and core of LLMs is this replicable then I don't see anyone downloading any apps for their typical AI usage. Specially that users have to go out of their way to allow a 3rd party to access all their data.
This is why OpenAI is so deep in the product development phase right now. They have to become the OS to be successful but I don't see that happening
MySpace and Friendster both claimed ~115M peak users.
> It's literally orders of magnitude.
Sure, and the speed at which ChatGPT went from zero to a billion is precisely why they need a moat... because otherwise the next one can do it to them.
Your argument is like a railroad company in 1905 scoffing at the idea that airliners will be a thing.
Peek users is not the amount of users they had when Facebook started.
Facebook probably would've never became a thing if MySpace already had ~115M users when it started.
MySpace had ~1M.
That's why DeepSeek (or anyone else) is going to have an incredibly difficult time convincing ~1B to switch from ChatGPT to their tool instead.
Can it happen? For sure.
Will it happen? Less likely.
If anyone unseats ChatGPT - it's much more likely to be a usual suspect like Google, Apple, or Microsoft - then some obscure company no one has ever heard of.
There is no network effect (amazon, instagram, etc.) not an enterprise vendor lock-in (Microsoft Office/AD, Apple Appstore, etc.) In fact, it's quite the opposite, the way these companies deliver ouput is damn near identical. Switching between them is pretty painless.
> You don't need a moat when you're in first place
There are different moats [1]. You’re describing incumbency, an intangible moat. It’s nice, but it’s fickle. Particularly with something with low switching costs.
OpenAI could argue, before, that it had a natural monopoly. More people use OpenAI so it gets more revenue and more data which lets it raise more capital to train these expensive models. That may not be true, which means it only has that first, shallow moat. It’s Nike. Not Google.
> There are different moats [1]. You’re describing incumbency, an intangible moat. It’s nice, but it’s fickle. Particularly with something with low switching costs.
Google has a low switching cost, and hardly anyone switches.
> Google has a low switching cost, and hardly anyone switches
Google has massive network effects on its ad business and a natural monopoly on its search index. Crawling the web is expensive. It’s why Kagi has to pay Google (versus being able to pay them once and then stop).
just for the chatbot, it's trivial to switch, create a new account and start asking questions from deepseek instead. There is nothing holding the users in chatgpt.
Businesses don’t even want to maintain servers locally. They definitely aren’t going to start managing servers beefy enough to run LLMs and try to run then with the reliability, availability, etc of cloud services.
This will make the cloud providers - especially AWS, GCP and to a lesser extent the also ran clouds more valuable. The other models hosted by AWS on Bedrock are already “good enough” for most business use cases.
And then consumers are definitely not going to be running LLMs locally on their computers to replicate ChatGPT (the product) anymore than they are going to get an FTP account, mount it locally with curlftpfs, and then using SVN or CVS on the mounted filesystem and then from Windows or Mac, accessed the FTP account through built-in software instead of using cloud storage like Dropbox. [1]
Whether someone comes up with a better product than ChatGPT and overcome the brand awareness is yet to be seen.
[1] Also the iPod had no wireless, less space than the Nomad and was lame.
There is a reason I kept emphasizing the ChatGPT product. The (paid) ChatGPT product is not just a text based LLM. It can interpret images, has a built in Python runtime to offload queries that LLMs aren’t good at like math, web search, image generation, and a couple of other integrations.
The local LLM on iPhones are literally 1% as powerful as the server based models like 4o.
That’s not even considering battery considerations
> The local LLM on iPhones are literally 1% as powerful as the server based models like 4o.
Currently, yes. That's why this is a compelling advance - it makes local LLMs much more feasible, especially if this is just the first of many breakthroughs.
A lot of the hype around OpenAI has been due to the fact that buying enough capacity to run these things wasn't all that feasible for competitors. Now, it is, potentially even at the local level.
Training costs are not the same as inference costs. DeepSeek (or anyone hosting DS largest model) will still need a lot of money and a bunch of GPU clusters to serve the customers.
Not sure I agree with your premise, but what exactly are they going to ban?
They can stop DeepSeek doing various things commercially I guess, but stopping Americans using their ideas is simply impossible and stopping use of their source or weights would be (likely successfully) challenged under the first amendment.
There is no law against simply destroying trillions of dollars of shareholder value.
American researchers had already made enough progress to prove that LLMs were not an incomprehensible trade secret based on years of secret knowledge - investors and tech executives were simply lead to believe otherwise. Well-connected people are probably very mad about this and they may try to lash out like the emotional human beings they are.
It doesn't matter from the US government perspective if all of the tech is replicated by US companies and US user continue to use US AI technology. But if US users start to use Chinese AI tech, then protectionism urges will appear that will likely figure out how to ban its use or subject it to large tariffs (e.g. TikTok, BYD, network equipment, solar panels, etc.)
It matters because their goal was hyping up how advanced and difficult their tech is, propping up their valuations.
DeepSeek proved the emperor had
no clothes and wiped out a lot of their valuation when investors saw reaching parity to Chtgpt is not really that difficult.
I think parent was asking would it even matter if there was a ban. To which the answer would be "no" because as you said the point has been made. And, as parent pointed out, it's repeatable anyway.
>It's reasonably likely that a lot of people linked to the federal government want to ban DeepSeek.
It took them years and years to move forward with the ban ok Tiktok and it still hasn't been banned yet. There is no way they are going to ban some MIT-licensed weights.
There's a lot of egg on people's faces now. DeepSeek shows there's nothing special about America or its economic system that breeds innovation. DeepSeek shows how these tech oligarchs greatly overplayed their hand and along with the president, bamboozled the taxpayer to enrich each other. I just hope the voters remember this in 2 years, 4 years and beyond.
If the allegations are true, the special thing about OpenAI is that it didn't have to be trained off DeepSeek. But either way, you maybe don't want to invest billions in something if someone else will be able to copy it for less.
Yeah they're setting this up to ban it. Crazy that they think this kind of approach will work in any way. Banning H100s didn't work, and actually pushed them to innovate. Now someone has found a more efficient way to train a model and they decide the best way forward is for the US not to benefit from access to it? This is clear evidence of collusion between OpenAI and the US Government to disadvantage competitors. Beyond that, it will never work. If they need to be reminded of just how little power they have to control the distribution of open source models, I think we would all be happy to enlighten them.
