>
"They are ALL dead... 1.8 Million of them killed in Ukraine" Col. Douglas MacGregor | Reda
Hospital food proven bad scientifically – and may undermine health
Electron beam solves Teflon trash problem with super-efficient zaps
Sam Altman Again Claims AGI is Solved and Huge Norway and Other Data Center for Superintelligence
The mitochondria are more than just the "powerhouse of the cell" – they initiate immune...
Historic Aviation Engine Advance to Unlock Hypersonic Mach 10 Planes
OpenAI CEO Sam Altman Pitches Eyeball-Scanning World ID to Bankers
New 3D-printed titanium alloy is stronger and cheaper than ever before
What is Unitree's new $6,000 humanoid robot good for?
"No CGI, No AI, Pure Engineering": Watch Raw Footage Of 'Star Wars'-Style Speeder
NASA's X-59 'quiet' supersonic jet rolls out for its 1st test drive (video)
Hypersonic SABRE engine reignited in Invictus Mach 5 spaceplane
"World's most power dense" electric motor obliterates the field
The Wearables Trap: How the Government Plans to Monitor, Score, and Control You
Hyping his data center agreement which was exaggerated. Exaggerated empty superintelligence. What they were actually talking about 10K B300s in a few months (building and 20MW power, and install the chips). They will have 230 MW and another 90K chips 2026 maybe)
10K is 10% of the number XAI Grok 3 GPUs colossus 1 end of 2024, but with 7X better than H100 B300 chips.
75K H100 equivalents in a few months. 750K H100 equivalent end of 2026.
XAI Colossus 1 already has had 400K H100 equivalent 2 months ago. XAI Colossus 2 has 2.7 million H100 equivalents now and will have 5-6 million H100 equivalents by end of 2025. OpenAI Norway is over 2 years behind XAI.
10,000 B300s is about 300 Exaflops. There is 30 petaflops (50% more than a B200) per Nvidia B300. Getting to 100,000 B300s would be 3,000 Exaflops. Every builder except XAI has had trouble getting more than 30,000 Nvidia GPUs to work together with on unified (coherent) memory which is needed for effective AI pre-training.
Sam repeats OpenAI's progress toward AGI but emphasizes a pivot to superintelligence. He talks about the need for massive physical infrastructure (e.g., chips, servers, cables, and energy) to make it a reality. OpenAI's official announcement of Stargate Norway a new AI data center project in the Narvik region of northern Norway.
Stargate Norway will be OpenAI's first major data center in Europe. It will eventually have a 100K GPU supercluster built in partnership with Nscale (a London-based data center firm). Norway has cheap hydroelectric power. They expect 100,000 Nvidia GPUs (graphics processing units) by the end of 2026, with potential for further expansion. It is a $1-billion facility and really needs about $3-4 billion to get to 100,000 GPUs and $6-8 billion for 200,000 GPUs. The centre is to be built near Narvik in northern Norway. Nscale and Aker will jointly own Stargate Norway with 50% stakes each, investing about $1 billion for the initial phase of the project, which will consume about 20 megawatts of electrical power capacity and have OpenAI as its first customer. Aker said it was working with local power producer Nordkraft to secure access to enough electricity to allow Stargate to increase its consumption to 230 MW and eventually add a further 290 MW. They will use B300s. 20 Megawatts only allows about 10,000 chips to be powered up.