Trump 2.0 has tons of common sense and is addressing many “self-inflicted” issues such as power generation.
So Deepseek stole data from OpenAI and trained their models on smuggled Nvidia chips from Singapore?
Who would have thought?
.
The idea that lies repeated often enough become true is known as the illusory truth effect. It’s a cognitive bias that causes people to believe something is true if they hear it repeated multiple times, even if it’s false. For example, claims by some well known CEOs and political influencers…
…
Read carefully how NVDA conduct business.
…
Trump is a President because he is realistic and has common sense. He doesn’t let his ego clouded his judgement.
Enough said.
…
Those who didn’t watch the interview of Aravind Srinivas, should watch it. Pay close attention to Aravind’s responses to Bose’ questions.
…
Which companies are most hurt by Deep Seek?
OpenAI, Microsoft 
Valuation of OpenAi crashes.
MSFT’s stake in OpenAI crashes.
Sam’s big ideas like Stargate becomes dumb, and going from non-profit to for-profit would be difficult.
Who egos are most hurt?
Well, 
Huh? Sure you read carefully, Gene? Few words, plenty said.
- Fully export control compliant
- Test-Time Scaling once thought is irrelevant can be used to create new models
- Training once thought has plateau because cost of next frontier has become prohibitive. Now can continue to scale upwards. That is, for the same dollar, can scale frontier that was once thought few years away.
- NVDA has the right solution for inference. Expect sale of GPUs for inference to increase.
- Total sale of NVDA chip is expected to increase because sale of GPUs for training remain healthy while sale of GPUs for inference would increase.
MeetKevin makes more comments on Deep Seek. He opines OpenAI’s tantrum and David Sacks’ allegation as non-issue.
Bosa opines that Microsoft is sour grape 
…
Mark Zuckerberg embraces Deep Seek…
Zuckerberg noted that “there’s a number of novel things they did we’re still digesting” and that Meta plans to implement DeepSeek’s “advancements” into Llama.
…given their competitive total cost of ownership and now with DeepSeek showing LLM developers how to build state-of-the-art models for a fraction of the price. With their GPUs coming in at a lower cost than Nvidia’s, DeepSeek means AMD can now offer comparable performance at much lower prices. This is huge for AI. This is big for AMD.
Disclosure: Own 100 shares of AMD.
Brad’s wild speculation is cost of developing o1 is $15million. Suddenly, from hundred of millions to billion dollars to $15million only? Is he also implying that Deep Seek was trained on 2048 H800s and not the wild speculation by another CEO to be 50,000 H100s?
Deep sip.
As AI transforms the real estate sector, its potential to optimize valuations, predict market trends, and uncover investment opportunities is immense. By leveraging these advancements, VillaTerras is positioned to help clients make data-driven decisions, identify high-growth areas, and maximize land value, ensuring long-term success in a rapidly evolving market.
So…VillaTerras - can your AI tell me what this big ole rock 3 miles from my desert getaway is REALLY worth? BTW - the lot has no electricity and likely never will.
https://www.zillow.com/homedetails/1-S-Viewcrest-Trl-Marana-AZ-85658/2061338730_zpid/
Media always made misleading headlines. US companies having regional HQs and datacenter shouldn’t be written as firms in Singapore. Intentionally vague.
Despite getting a bargain, pretend to be on the losing end
得了便宜卖乖
David Sachs (AI Czar in White House), based on his talks to many industry pundits, High Flyer and Deep Seek, has:
10,000 H100s (acquired before the export control ban)
10,000 H800s
30,000 H20s
Possible that $5.6 million is true as it is for final run of training and not the full cost that would include prior runs, etc.
Chamath quoted two great innovations born out of “necessity is the mother of invention”. GRP (instead of PPO) and PTX (instead of compiling through CUDA).
(CUDA is no longer a moat?)
Microsoft hosting R1… wow! Embraces Deep Seek!
David Sachs, "Everybody in SV do distillation.”
(When a Chinese app did it, is called stealing?)
Jason McCabe Calacanis has no sympathy over OpenAI’s accusations of DeepSeek using distillation.
Chamath opines that he won’t be surprised China would develop an alternative to ASML and leapfrog NVDA.