What you find with any market news is that the optimism and pessimism are always overblown. Fear and greed. It happens all the time.
Look at Linux. For those old enough to remember, there was a time where many (including Microsoft) were worried it would destroy the company (eg [1]). There were complaints about the market destruction caused by Linux. What actually happened? Microsoft is bigger than ever even though Linux is on billions of devices worldwide.
IF DeepSeek's claims are real and this stands up, all that's happened is at worst the profit opportunity has simply move *as it was always going to do). This might be bad for OpenAI and Sam Altman but Big Tech will (IMHO) be fine.
Remember that training LLMs for chatbots, which is something people focus on, is just one narrow slice of the potential AI market. Recommendation engines, industrial/commercial applications, medicine, etc.
If there has been a aoftware breakthrough and training LLMs now costs a fraction of what it did last year, there's now an order of magnitude more potential appllications that have become economical.
Consider this: if we can do today with a model 1/10th the size of what we needed last week, what applications will there be for a model 10x DeepSeek-R1's size?
I'm also reminded of the invention of the cotton gin. This automated what used to be a highly manual process. At the time, there was concern this would diminish the need for slaves on cotton plantations. Instead the need exploded because cotton became so much cheaper [2].
Lastly, Stargate is largely meaningless. Companies spend a fortune on data centers. GPUs are just a fraction of that. A genuine software improvement just means you can do more with less.
My point is: don't panic. Unless you're an OpenAI investor, maybe.
Look at Linux. For those old enough to remember, there was a time where many (including Microsoft) were worried it would destroy the company (eg [1]). There were complaints about the market destruction caused by Linux. What actually happened? Microsoft is bigger than ever even though Linux is on billions of devices worldwide.
IF DeepSeek's claims are real and this stands up, all that's happened is at worst the profit opportunity has simply move *as it was always going to do). This might be bad for OpenAI and Sam Altman but Big Tech will (IMHO) be fine.
Remember that training LLMs for chatbots, which is something people focus on, is just one narrow slice of the potential AI market. Recommendation engines, industrial/commercial applications, medicine, etc.
If there has been a aoftware breakthrough and training LLMs now costs a fraction of what it did last year, there's now an order of magnitude more potential appllications that have become economical.
Consider this: if we can do today with a model 1/10th the size of what we needed last week, what applications will there be for a model 10x DeepSeek-R1's size?
I'm also reminded of the invention of the cotton gin. This automated what used to be a highly manual process. At the time, there was concern this would diminish the need for slaves on cotton plantations. Instead the need exploded because cotton became so much cheaper [2].
Lastly, Stargate is largely meaningless. Companies spend a fortune on data centers. GPUs are just a fraction of that. A genuine software improvement just means you can do more with less.
My point is: don't panic. Unless you're an OpenAI investor, maybe.
[1]: https://www.zdnet.com/article/microsoft-linux-is-a-threat-it...
[2]: https://www.archives.gov/education/lessons/cotton-gin-patent