Microsoft says Maia 300 racks will hold 96 chips and use ~167 kW
Microsoft says Maia 300 racks will hold 96 chips and use ~167 kW of power and plans to roll out "multi-gigawatt deployments" of the chip starting next year
TL;DRMicrosoft's custom AI chip Maia 300 fits 96 units per rack, launching mass production next year.
Why it matters: Reduces reliance on Nvidia; could lower AI infrastructure costs and shift GPU market dynamics significantly.
Deploy decision models with llama.cpp on Amazon SageMaker AI
TL;DRAWS now lets you run llama.cpp models on SageMaker with structured decision outputs.
Run multi-node inference workloads using Ray LLM Deep Learning
TL;DRRay LLM containers let you split large model inference across multiple GPUs on Kubernetes.
Latest
Clearing chat history helps AI agents find better solutions
Clearing chat history helps AI agents escape bad solution loops and find better answers.
Microsoft plans to scale its own Maia AI chips to gigawatt levels
Microsoft is building its own AI chips to power products at massive scale by 2027.
What happens to Google and Apple accounts when the owner dies?
Apple and Google have policies for accessing or memorializing accounts after death.
Bambu Lab officially teases its first 3D scanner
Bambu Lab is building a 3D scanner to complement its popular 3D printers.
Citroën has brought back the iconic 2 CV
Citroën revived the classic 2CV as an electric concept vehicle.
Amazon In Talks To Buy AI Startup Decart For Around $7 Billion
2Amazon may buy AI startup Decart for $7 billion to speed up model optimization.
Jev-like decision models are flooding the AI market
Researcher accuses TypeSafe of copying their AI model; now similar models flood the market.
Smart just unveiled a new ultra-compact EV at the Paris Motor Show
Smart released a tiny electric car with good battery range.
AMD unveils Helion racks for AI agents at OCP summit
AMD is releasing new server racks designed for AI workloads starting November.