Google developing new chip to run Gemini AI model more efficiently: report
The hardware, dubbed Frozen v2, is slated for launch in 2028
News - TPU
The latest news on networking, cloud, and AI infrastructure
-
Google unveils Virgo, the scale-out fabric ditching three-layer networks to unify AI accelerators
Stars align for a celestial-named fabric that promises 40% lower unloaded latency and four-times bandwidth per accelerator
-
Google in discussions with Marvell for development of two inference chips: report
Follows announcement that Broadcom will develop Google TPUs until 2031
-
Broadcom secures deal with Google for next-gen AI racks and custom TPUs
SEC filing confirms networking component deal, but reports suggest inference-optimized iterations of Google’s custom silicon could face delays
-
Broadcom gulps AI-infused silicon revenues, touts superior supply chain planning
A healthy drink that CEO Hock Tan said is not impacting VMware
-
Top 5 hardware stories of 2025: Broadcom-OpenAI, AWS doubles down on custom & Nvidia marches on
AWS doubles down on custom silicon and the looming memory crisis to come
-
Broadcom bulges with Anthropic, AI orders
And a new unnamed XPU customer is teased
-
Google claims its next-gen TPUs offer ‘more compute’ than the world’s fastest supercomputer
Claude developer Anthropic spoilt for inferencing choice as Ironwood adds to AWS Rainier options
-
Google unveils seventh-generation TPU, Ironwood
Comes in both 256 and 9,216 chip configurations
-
Amazon's Kuiper nabs Google custom chip director to build space silicon
Maruthy Vedam goes from TPUs to satellite semiconductors
-
Google makes 6th generation TPUs generally available
Trillium can now be accessed by Google Cloud customers
-
Google Cloud gives Y Combinator startups dedicated Nvidia GPU cluster
And access to TPUs
-
Apple to use Google's TPUs for AI instead of Nvidia GPUs
Apple Intelligence rolls out beta test this week
-
Trillium: Google promises 4.7× compute boost for latest TPU
Sixth generation processor is also 67 percent more energy efficient
-
Google unveils Axion, its first Arm-based CPU
Also reveals updates to TPU AI chip
-
Chinese national arrested for theft of Google AI data center trade secrets
500+ documents on data centers and TPUs allegedly stolen by software engineer
-
Google settles $1.67bn TPU patent infringement lawsuit
Settlement reached on day closing arguments were due to be heard
-
Google fights $1.67bn lawsuit against alleged AI chip patent infringement
Trial is expected to last two weeks
-
Training Google's Gemini: TPUs, multiple data centers, and risks of cosmic rays
What we can glean from research papers, public comments, and talking to Google Cloud's CEO
-
OpenAI appoints former Google TPU lead as head of hardware, hiring for "experts in data center facility design"
Chip engineer Richard Ho leads new division at the generative AI startup to help "influence roadmaps for hardware partners for data center networks, racks, and buildings"
-
Anthropic to use Google TPU v5e chips to train generative AI models
Chip now generally available in Google's Cloud
-
CyrusOne offers 300kW-per-rack AI data center design
Company says Intelliscale offers higher density in a smaller footprint, through multiple liquid cooling techniques