ASML Changes AI Narrative As DeepSeek Scrutiny Grows
HTML-код
- Опубликовано: 5 фев 2025
- ASML surged the most since 2020 after booking orders worth twice as much as analysts expected, as the artificial intelligence boom fuels demand for its chipmaking machines. ASML shares rose as much as 12% and more than fully recovered from a drop earlier in the week on concern that a new AI model by Chinese startup DeepSeek can provide comparable performance to Western chatbots at a fraction of the price. ASML makes machines needed to produce some high-end chips that the Chinese company is barred from purchasing. Meanwhile Microsoft and OpenAI are investigating whether a group linked to Chinese AI startup DeepSeek obtained data from OpenAI's technology without authorization. The group allegedly used OpenAI's API to exfiltrate a large amount of data, potentially violating OpenAI's terms of service. Tom Mackenzie joins Lizzy Burden and Stephen Carroll on Bloomberg radio to discuss it all.
--------
Watch Bloomberg Radio LIVE on RUclips
Weekdays 7am-6pm ET
WATCH HERE: bit.ly/3vTiACF
Follow us on X: / bloombergradio
Subscribe to our Podcasts:
Bloomberg Daybreak: bit.ly/3DWYoAN
Bloomberg Surveillance: bit.ly/3OPtReI
Bloomberg Intelligence: bit.ly/3YrBfOi
Balance of Power: bit.ly/3OO8eLC
Bloomberg Businessweek: bit.ly/3IPl60i
Listen on Apple CarPlay and Android Auto with the Bloomberg Business app:
Apple CarPlay: apple.co/486mghI
Android Auto: bit.ly/49benZy
Visit our RUclips channels:
Bloomberg Podcasts: / bloombergpodcasts
Bloomberg Television: / @markets
Bloomberg Originals: / bloomberg
Quicktake: / @bloombergquicktake
If you can’t compete, then smear. US is very good at smearing.
Usa is very different usa today now..usa is now weak and useless now,,,,,?...
But today usa are no good at all,?
Usa mismanage usa, until all totally dont trust usa anymore now,?....❤❤❤❤❤❤
Why can't they accept that the Chinese are surpassing western nations in AI. I have worked with many Chinese. They are superb mathematicians and excellent in programming
whyte people > all people. but yeah can't have that
Google 'Yellow Peril' - that's exactly why.
Explain what is "OpenAI data"? Their system is closed source and they themselves trained on millions of copyrighted material without compensation.
Breaking news: World leaders are congratulating China , DEEPSEEK for helping the world for FREE!❤❤❤❤
ASML is trying to make money while they still can. They know their days are numbered.
That will be the final shock which is not very far. Fingers crossed.
Just free yourself from the great hegemon,you will be free to make as much money as you can freely
@@xianqingduJevon Paradox. If AI is cheaper, then more GPUs will be needed. Not less! FYI, NVDA owns CUDA & PTX. DeepSeek only works on NVDA GPUs!
@ Yes, agree with you completely. But you assume that only Nvidia can produce high end GPUs, just like OpenAI assumed that no Chinese company could produce comparable AI models before DeepSeek. The next shock will come when all Chinese AI models are trained using Chinese homemade GPUs and they export those GPUs at a much lower cost to other countries. Huawei is already producing chips as powerful as Nvidia’s H100 now.
@@tringuyen7519 When Jevon’s paradox is applied the ICE cars, it fails again in China because they will just use EVs.
OpenAI can change to CLoseAI. Do CloseAI obtain all Data legally?
ASML days are numbered wait till 2028 when market will be flooded with chinese cheap lithography machines
Yup, US consumers are clamoring for 14nm Chinese chips, & not TSMC’s latest, greatest 2nm chips. Do you even hear yourself speak?😂😂😂
这是个开源的模型，有全球海量用户和开发者参与的开源社区，美国政府官员根本就不懂这些技术的意义，就算是你禁止了也没有多大的意义，因为普通用户也可以下载到本地运行。AI模型竞争的关键你的算法是否更优秀真正实现AGI的终极目标，福利所有人类。而不是利用国家安全建立护城河，没有海量用户和开发者的参与和测试，想独自建立个闭源模型垄断整个AI行业根本就是不可能的事。
It is just a cheap copy at the end.
@@RolopIsHere Anything that has its own charactericts is not just a "copy". by your logic you're just a cheap copy of some billionare's child then? Or a cheap copy of some kind of genius? I know my statement is dumb but yours is dumber.
@@RolopIsHereu make it sound like a bad thing. A cheap copy is quite literally the goal.
@@RolopIsHerethat actually a good thing, they can do the same thing faster better and cheaper. It’s actually very fast I have both apps.
@ "better and cheaper", no, they destiled the weights of Chat GPT, they added some "improvements" but its like putting some make up to a pig, they wont be able to replicate any future models since this was a copy of the previous one by abusing chat gpt api. So they very likely won't be able to do significant improvements. And if proven, this might only deepen the sanctions against China.
ASML results are a lagging indicaror. DeepSeek news is a leading indicator. This guy is comparing the two to arrive at a nonsensical conclusion.
Jevon Paradox. If AI becomes cheaper, then everyone will use more GPUs, not less. & DeepSeek runs exclusively on NVDA GPUs bc NVDA owns both CUDA & PTX architectures!
ASML is the back-end of the back-end of the back-end. They won't be affected much.
The entire value chain of AI will be affected. The domino effect will play out in the coming two years. It is matter of when not if.
@@yasinhyDeepSeek uses PTX & CUDA which are owned by NVDA exclusively. It’s pushing NVDA dominance, not reducing it. You’re so, so naive!😂😂😂
, or he's deliberately trying to be "clueless."
reasoning, reinforcement learning, and distillation will improve AI model. DeepSeek is doing that.
All of which runs on NVDA GPUs. DeepSeek uses CUDA & PTX which are both owned by NVDA. DeepSeek is pushing NVDA dominance!
And if this is something to not be concerned about please explain to me why the US House of Representatives was moving to ban deep-seek
Can't compete? No worries Papa government will protect you. :)) You will eventually innovate nothing.
Can they sue DeepSeek without content providers suing OpenAI? There are many laptops now sold with NPUs, AI chips that are useless for local models because of the large size of the model. The chips are manufactured by Intel among others. DeepSeek can run on these laptops so that must be a win for those manufacturers.
it's absolutely hilarious. if they sue DeepSeek, it will ultimately boil down to how "Open"AI trained their models. hmm i wonder if "Open"AI trained their data on legitimate sources, just like that one "Open"AI whistleblower "discovered" the legitimacy of the data for sure.
Numbers don’t speak for themselves; the earnings are for the quarter ending December 2024. The deepseak sell-off was just yesterday.
I am not saying there is no story, but blindly being a bull is not an intelligent thing to do.
The actual numbers that would matter are those of the next quarters.
The data from ASML is actually from before Deepseek hit the AI market. So you need to wait for next earnings to see the real results of the affects of deepseek on them.
ASML is the monopoly! That’s why they can name any price. Just a couple years ago a lithography machine costs $150 million and today $375 million. They can make like bandit!
past order doesn’t equal to future orders. Hyperscalers will cut back on Capex since LLM is 12 times cheaper to produce.
1. If DeepSeek just distilled their model from openAI and other LLMs, it proves that innovation DOES cost billions, even if copying costs less
2. Atleast until AGI is reached, it seems more likely that hyperscalers will parlay software advances with their enormous supercomputer builds, instead of replacing them. There is still a lot of innovation to come and that requires compute, period.
We will see how Microsoft / xAI / openAI respond. It's certainly possible DeepSeek just fuelled the fire of an international arms race, which would only inflate Capex spend.
@ They published their results, anyone can reproduce their methods. Computing cost is 2.88 million hours of H800, that won’t change no matter how much inference they did using openAI.
Using openAI inference to training deepseek allow them to find the cheapest way to train. Once the cheaper alternative is found, everyone will use their way, not bother building models from the ground up, since there are tons of open sources LLMs.
@@thejeffinvadeDeepSeek uses CUDA & PTX which are both owned by NVDA. They’re pushing NVDA dominance! & Jevon Paradox means more GPUs will be needed, not less.
@@tringuyen7519 Inferencing a LLM uses much less computing than building a model. Even if use cases increase more than 12 times, you still need less nvidia gpu as before. And I doubt 1/12 cost will produce 12 times more AI demands.
Where is my money, excuses for failure
Because the US government threatened them.
How is it able to show you its thoughts?
My friends are taking coding classes at Moonpreneur before enrolling in AI courses. Can I pursue AI courses without prior coding experience? Please clarify my doubts
Ooh now they have moved to blame game. Anyway quen 2.5 and r1 are not as good as opening 01 buy much cheaper..and most importantly they are OPEN-SOURCE.
War and economics experts suggest that asml may has to be prepared to lose some market to deep seek, to prevent asml from completely lose everything to deep seek,,,???...❤❤❤❤❤❤
DeepSeek R1is said to be “Open Source” but it isn’t - since they haven’t provided details on the training data etc it doesn’t meet the definitions which means people can get into copyright and IP trouble if they build apps on top of it - this is why Huggingface is now trying to reverse engineer it. Furthermore when you push R1it says it’s ChatGPT!!!
I love how everyone is now an expert on open A.I. in 72 hours lol
The collapse of the collapse of the American markets
Yes!
i am your retribution, i am your retribution.
As an investment enthusiast, I often wonder how top-level investors are able to become millionaires through investing. I have a significant amount of capital to start with, but I'm unsure about the strategies and direct I should take to help me generate substantial profits like some people are this season.
I’m not in a position to offer financial advice, but given the significant amount of capital you're working with, it would be wise to consult a financial advisor who can guide you in developing a strategy tailored to your goals and risk tolerance.
Factos!!. After experiencing a major portfolio loss in 2020 amid the COVID pandemic while trying to manage my investments on my own, I reached out to an investment advisor. They helped me turn my little investment into a seven-figure portfolio by providing the guidance and strategy I needed. Having that expertise made a significant difference in my investment journey.
Thanks for sharing your experience! I’ve been managing my portfolio myself, but it’s not working out. Do you have any recommendations for a good investment advisor? I could really use some help.
Nikki Leigh Brooks is a highly respected figure in her field. I suggest delving deeper into her credentials, as she possesses extensive experience and serves as a valuable resource for individuals seeking guidance in navigating the financial market.
Thank you so much for the suggestion! I really needed it. I looked her up on Google and explored her website; she has an impressive background in investments. I've sent her an email, and I hope to hear back from her soon!
Seriously? Bying more for what? When talent is at the lowest, silly decisions are made.
Too fragile to bet on
All indian IT engineers should realize the usa strategy , its time to leave Usa and make india and china stronger than Usa in IT .. best time to leave usa forever
FYI, DeepSeek uses CUDA & PTX which are both owned by NVDA. So AI startups in India & China will buy more NVDA GPUs, not less. NVDA is a US company!
Typical Western envy
OpenAI uses only PUBLIC data.
This is what GPT says: "OpenAI follows strict policies to ensure that training data comes from publicly available sources, licensed data, and information that complies with legal and ethical guidelines."
It said it "has faced multiple lawsuits related to copyright infringement"
if they say so, it must be true
@@clairsnow6404 yup that "Open"AI whistleblower found out they were using legitimate data so he decided to off himself.
@@justdummy-pr9t yeah nothing more depressing than realizing that you can't contribute to making the world a better place because it's already amazing 😂
Ask DeepSeek about Taiwan or Tiananmen square. Go ahead - then tell me it’s not controlled by Beijing.
Happy 35th anniversary of crushing an American failed color revolution in Tiananmen Square, Beijing, China. 🎉🎉🎉🎉
熱烈慶祝中國大陸在北京天安門廣場成功粉碎美國失敗的顏色革命的35周年 🎉🎉🎉🎉
@@directxxxx71DeepSeek uses CUDA & PTX which are both owned by NVDA. NVDA is a US company with a Taiwanese CEO!😂😂😂
Comments are flooded with chinese day dreamers
Asml mistakes is asml chose the wrong master,?
.it is like you asml working with wrong mindsets people, and cause asml , to its own self destruction and downfalls ?..
Good business mindsets, never go into self isolation and destruction on ownself,?....
It is like the famous rabbit and tortoise races story,,,,?...
The rabbit very proud of itself, and his relax , and after he wake up, he suddenly realised that he is now the losser,,,,,,?..... like asml now,,,,?... all due to asml/ usa arrogantly mindsets ways of doing things,,,????...
❤
Boost the propaganda machine now😂
stange, if you not comfirm it ,why you say that people is a crime