Anthropic AI Buying Microsoft Maia Chips - NVIDIA is Dead
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
AI Token Maxxing Failed at Uber - OpenAI is Dead
Eli the Computer Guy
22.2k views
Huawei New 1.4nm SmartPhone Chips to Beat NVIDIA - China Building Future Without USA
Eli the Computer Guy
32.8k views
Microsoft Cancels Anthropic AI Claude Subscriptions for Developers - AI Is Too Expensive
Eli the Computer Guy
30.9k views
Anthropic’s Claude AI Deletes PocketOS Production Database and Backups - Founder is an Idiot
Eli the Computer Guy
30.4k views
Chinese 14nm Chips Beat NVIDIA 4nm Silicon - China AI Architecture Superior to USA Tech
Eli the Computer Guy
24.4k views
Anthropic confirms software engineering is NOT dead
ThePrimeTime
341.6k views
Deepseek V3.2 Beats GPT-5 and Gemini 3 Pro - Chinese AI Destroying US Tech
Eli the Computer Guy
24.6k views
Meta Layoffs 600 AI Employees -- Mark Zuckerberg is Popping AI Bubble
Eli the Computer Guy
24.9k views
Huawei AI Beating NVIDIA with Atlas 950 SuperCluster -- USA Keeps Losing Tech War with China
Eli the Computer Guy
31.1k views
China Bans NVIDIA AI Hardware -- Chinese Not Addicted to USA Tech
Eli the Computer Guy
26.9k views
Top Comments (10)
The other Big Tech companies were never going to be happy with all the AI profits going to Nvidia with their ridiculous 70% margins.
Seems like everybody has given up on finding that gold and decided they needed to be the ones making the shovels and picks...
this all reminds me of Enron but instead of one company it's an entire segment of business!
Remember when the record industry tried to stop mp3? LOLZ.
NVIDIA committed hara-kiri
Basically, there's a reason most businesses fill their fleet with toyotas, not BMWs
The corporate con job must escalate at any cost! 😂
Hadn't seen u in a long time glad to see your out here teaching people
Eli is right! My IT career started with Cisco being the only game in town. Their early ads “Empowering the Internet Generation” showed how dominant they were and “Nobody got fired for buying Cisco” Nowadays, even major enterprises are running mixed networks running JUNIPER, HP, Cisco , ARISTA etc. This is a good analogy and maybe foreshadowing what is happening to AI accelerators.
The main thing needed to run llm is a huge memory bandwidth. The type of compute operations in most llms is pretty simple matrix vector ops with different data types and dimensionnality... A dedicated TPU with the right limited feature set would work pretty well if you give it 1tb/s bandwidth
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
The other Big Tech companies were never going to be happy with all the AI profits going to Nvidia with their ridiculous 70% margins.
Seems like everybody has given up on finding that gold and decided they needed to be the ones making the shovels and picks...
this all reminds me of Enron but instead of one company it's an entire segment of business!
Remember when the record industry tried to stop mp3? LOLZ.
NVIDIA committed hara-kiri
Basically, there's a reason most businesses fill their fleet with toyotas, not BMWs
The corporate con job must escalate at any cost! 😂
Hadn't seen u in a long time glad to see your out here teaching people
Eli is right! My IT career started with Cisco being the only game in town. Their early ads “Empowering the Internet Generation” showed how dominant they were and “Nobody got fired for buying Cisco” Nowadays, even major enterprises are running mixed networks running JUNIPER, HP, Cisco , ARISTA etc. This is a good analogy and maybe foreshadowing what is happening to AI accelerators.
The main thing needed to run llm is a huge memory bandwidth. The type of compute operations in most llms is pretty simple matrix vector ops with different data types and dimensionnality... A dedicated TPU with the right limited feature set would work pretty well if you give it 1tb/s bandwidth