We have one- and three-year commitment pricing for Cloud TPUs that give substantial discounts off the more ad hoc hourly pricing .
AI HARDWARE
-
New AI Chip Launch Generates Excitement for Future Applications
By
–
It's a great chip! Congrats! We're looking forward to using it for many things later this year!
-
Google Colab Offers Free TPU Access for ML Research
By
–
Many of them are still happily serving internal and external uses! For example, there's free access to some kind of TPUs on Colab, to enable external researchers without easy access to accelerators to run accelerated ML workloads: https://
research.google.com/colaboratory/f
aq.html
… -
Harold Sinnott at NVIDIA GTC25 Conference
By
–
Harold Sinnott NVIDIA #GTC25 (
@haroldsinnott
.bsky.social) https://
bsky.app/profile/harold
sinnott.bsky.social/post/3llwi7r6sg22j
… -
Telecom Companies Partner with NVIDIA to Build Sovereign AI Factories
By
–
Telecom companies are collaborating with NVIDIA to become sovereign #AI factories. Three key steps are needed: Build AI infrastructure Develop a national foundation model Drive consumption of AI infrastructure Read the ebook: https://
nvda.ws/4iVokzk -
Cerebras Delivers 19x Faster Code Generation with Llama 4
By
–
Subsecond code generation in action. 👀
— Cerebras (@cerebras) 9 avril 2025
Same prompt. Same output. Just 19x faster. Cerebras powers the fastest @AIatMeta Llama 4.
Get Access now: https://t.co/isthZ2Xau4 pic.twitter.com/7xzU7Tevf8Subsecond code generation in action. Same prompt. Same output. Just 19x faster. Cerebras powers the fastest @AIatMeta Llama 4. Get Access now: https://
cerebras.ai/build-with-us -
Meta Llama 4 Maverick achieves fastest inference speed SambaNova
By
–
You know we always come through with the speed 🚀@AIatMeta's #Llama4 Maverick is now available on SambaNova Cloud with the FASTEST inference!
— SambaNova (@SambaNovaAI) 9 avril 2025
✈️ 655 t/s — independently verified by @ArtificialAnlysYou know we always come through with the speed @AIatMeta
's #Llama4 Maverick is now available on SambaNova Cloud with the FASTEST inference! 655 t/s — independently verified by @ArtificialAnlys -
Google Unveils Ironwood 7th Generation TPU System
By
–
Great to see @sundarpichai kicking off #GoogleCloudNext in Las Vegas by talking about our Ironwood, our 7th generation TPU system (29X more power efficient / operation than our first TPU pod system, and 3600X more performance). I'm excited to join @SabaMugazambi on Friday at
-
NVIDIA Google Cloud Deploy Agentic AI On-Premises Gemini
By
–
Announced at #GoogleCloudNext: NVIDIA and @GoogleCloud are bringing #agenticAI on-prem with support for Google Gemini models—powered by NVIDIA Blackwell and Confidential Computing for top-tier performance and security. https://
nvda.ws/4loZnhh Hear more from our CEO Jensen Huang. -
Llama 4 Achieves Record 2611 Tokens Per Second on Cerebras
By
–
Llama 4 is now live on Cerebras!
– We broke the perf chart again at 2,611 tokens/s
– 19x faster than the leading GPU cloud
– Only API in the world with <1s total response time
Try now: https://
inference.cerebras.ai