GPU
A powerful processor, originally for graphics, now the workhorse chip of modern AI.
A GPU, or Graphics Processing Unit, is a chip originally designed to render video-game graphics. It turned out that the same ability to do many calculations in parallel is exactly what training and running AI models needs, and GPUs have become the workhorse hardware of modern AI.
Because demand for AI has exploded, powerful GPUs are expensive and often in short supply, and securing enough of them is a strategic priority for every major lab. They are a big part of why AI is so capital-intensive, and why running models on your own hardware, even a capable laptop or Mac, is an appealing, cost-controlled alternative for smaller workloads.
Alongside specialised chips like Google’s TPU, GPUs form the physical foundation of the compute that powers the whole field.
-
Anthropic Hired the Man Who Built Google's AI Chips, and It Is Not Subtle About Why
-
The Same Model Scored 30 Percent Alone and 100 Percent Inside Nvidia's Agent System
-
Anthropic Left Claude Alone With 15 Protein Targets for 48 Hours, and the Lab Results Came Back
-
Microsoft Has Spent 280 Billion Dollars on AI. The Guardian Asked Where the Chips Are.
-
Stop Asking Whether AI Is a Bubble. Start Reading the Lease Agreements.
-
The Creator of Redis Made a Chinese Video Model Run on a Mac. Europeans Are Not Licensed to Use It
-
Mojo Reached 1.0, Three Years After Promising to Make Python Fast
-
Nvidia Will Cover Part of the Loss If Its Own Chips Lose Value. That Is How You Raise 500 Billion Dollars.
-
Claude Code Turns On Auto Mode by Default, and the Safety Numbers Are Not What You Would Guess
-
AMD Buys Taalas, a Startup That Bakes an Entire AI Model Into the Chip
-
Meta ships Muse Code, a coding agent that keeps working after it crashes
-
Someone got DeepSeek V4 Flash running in production on a single AMD card and published every patch it took
-
Alibaba's Qwen3.8-Max spent 16 days writing a tool by itself, and the weights go public next week
-
Mira Murati's lab shrank its own model to a quarter of the size and lost one point
-
Microsoft says it will stop chasing the frontier and build small specialist models instead
-
OpenAI cuts GPT-5.6 Luna prices by 80 percent as the AI price war heats up
-
A $500 training run made a 9B model beat every frontier model at one boring job
-
Kimi K3's weights are finally public, and they weigh 1.4 terabytes
-
Ollama quietly made Macs a better place to run open models
-
Microsoft and Mistral team up to build AI in Europe, and keep the data there
-
AMD and Cerebras team up to make AI answers come back faster
-
OpenAI Says GPT-5.6 Sol Trained Its Own Smaller Sibling
-
Grok 4.5 Arrives at a Third of the Price, and That Might Matter More Than Benchmarks
-
Claude Science: Anthropic builds an AI workbench that keeps research data on your own machines