eFinder

eFinder

Home › Article

Google unveils chips for AI training and inference in latest shot at Nvidia

Cnbc · · 498 words · By Jordan Novet
open_in_new Read the original article: https://www.cnbc.com/2026/04/22/google-launches-training-and-inference-tpus-in-l…

fact_checkFact-Check Results

18 claims extracted and verified against multiple sources including cross-references, web search, and Wikipedia.

schedule Pending 8
check_circle Corroborated 4
verified Verified By Reference 2
info Single Source 2
help Insufficient Evidence 2
check_circle
“Google is separating those tasks into distinct processors, its latest effort to take on Nvidia in AI hardware.”
CORROBORATED
Multiple web search results confirm Google is splitting its TPU tasks into separate chips for training and inference to compete with Nvidia.
menu_book
wikipedia NEUTRAL — Google AI is a subsidiary of Google DeepMind dedicated to artificial intelligence (AI). It was announced at Google I/O 2017 by CEO Sundar Pichai. This division has been expanded to its reach with rese…
https://en.wikipedia.org/wiki/Google_AI
menu_book
wikipedia NEUTRAL — Gemini (also known as Google Gemini and formerly known as Bard) is a generative artificial intelligence chatbot and virtual assistant developed by Google. It is powered by the family of large language…
https://en.wikipedia.org/wiki/Google_Gemini
menu_book
wikipedia NEUTRAL — Claude is a series of large language models developed by American software company Anthropic which were first released in 2023. The Claude models are used in chatbots and for AI-assisted software deve…
https://en.wikipedia.org/wiki/Claude_(language_model)
+ 3 more evidence sources
check_circle
“Google said Wednesday that it's making the change for the eighth generation of its tensor processing unit, or TPU.”
CORROBORATED
Web search results specifically mention the TPU 8t (training) and TPU 8i (inference) as part of this architectural shift.
menu_book
wikipedia NEUTRAL — Google Tensor is a series of ARM64-based system-on-chip (SoC) processors designed by Google for its Pixel devices. It was originally conceptualized in 2016, following the introduction of the first Pix…
https://en.wikipedia.org/wiki/Google_Tensor
menu_book
wikipedia NEUTRAL — A neural processing unit (NPU), also known as an AI accelerator or deep learning processor, is a class of specialized hardware accelerator or computer system designed to accelerate artificial intellig…
https://en.wikipedia.org/wiki/Neural_processing_unit
menu_book
wikipedia NEUTRAL — Tensor Processing Unit (TPU) is a neural processing unit (NPU) application-specific integrated circuit (ASIC) developed by Google for neural network machine learning. Tensorflow, Jax, and PyTorch are …
https://en.wikipedia.org/wiki/Tensor_Processing_Unit
+ 3 more evidence sources
check_circle
“Both chips will become available later this year.”
CORROBORATED
Web search results explicitly state that Google says both the training and inference chips will be available later this year.
menu_book
wikipedia NEUTRAL — Amir Salek is an electrical engineer and technology executive known for his leadership in custom silicon design and artificial intelligence hardware development. He founded and led Google's custom sil…
https://en.wikipedia.org/wiki/Amir_Salek
menu_book
wikipedia NEUTRAL — Google AI is a subsidiary of Google DeepMind dedicated to artificial intelligence (AI). It was announced at Google I/O 2017 by CEO Sundar Pichai. This division has been expanded to its reach with rese…
https://en.wikipedia.org/wiki/Google_AI
menu_book
wikipedia NEUTRAL — Tensor Processing Unit (TPU) is a neural processing unit (NPU) application-specific integrated circuit (ASIC) developed by Google for neural network machine learning. Tensorflow, Jax, and PyTorch are …
https://en.wikipedia.org/wiki/Tensor_Processing_Unit
+ 3 more evidence sources
verified
“In March, Nvidia talked up forthcoming silicon that can enable models to rapidly respond to users' questions, thanks to technology obtained in its $20 billion deal with chip startup Groq.”
VERIFIED BY REFERENCE
While Wikipedia confirms Groq is an AI accelerator company and Nvidia is a GPU leader, no provided evidence mentions a $20 billion deal between Nvidia and Groq.
menu_book
wikipedia NEUTRAL — Groq, Inc. is an American artificial intelligence (AI) company that builds an AI accelerator application-specific integrated circuit (ASIC). The architecture was originally introduced as a Tensor Stre…
https://en.wikipedia.org/wiki/Groq
menu_book
wikipedia NEUTRAL — Nvidia Corporation ( en-VID-ee-ə) is an American multinational technology company headquartered in Santa Clara, California. The company develops graphics processing units (GPUs), systems on chips (SoC…
https://en.wikipedia.org/wiki/Nvidia
menu_book
wikipedia NEUTRAL — Steven Cliff Bartlett (born 26 August 1992) is an English entrepreneur, investor and podcaster. In 2014, he founded Social Chain, and in 2017, started The Diary of a CEO podcast, which Spotify Wrapped…
https://en.wikipedia.org/wiki/Steven_Bartlett_(businessman)
+ 3 more evidence sources
check_circle
“Google is a large Nvidia customer, but offers TPUs as an alternative for companies that use its cloud services.”
CORROBORATED
Web search results confirm Google Cloud offers both Nvidia GPUs and its own TPUs to customers.
menu_book
wikipedia NEUTRAL — Anthropic is an American artificial intelligence (AI) company headquartered in San Francisco. It has developed a range of large language models (LLMs) named Claude and focuses on AI safety. Anthropic …
https://en.wikipedia.org/wiki/Anthropic
menu_book
wikipedia NEUTRAL — Generative artificial intelligence (AI), sometimes abbreviated as GenAI, is a subfield of artificial intelligence that uses generative models to generate text, images, videos, audio, software code (vi…
https://en.wikipedia.org/wiki/Generative_AI
menu_book
wikipedia NEUTRAL — Tensor Processing Unit (TPU) is a neural processing unit (NPU) application-specific integrated circuit (ASIC) developed by Google for neural network machine learning. Tensorflow, Jax, and PyTorch are …
https://en.wikipedia.org/wiki/Tensor_Processing_Unit
+ 3 more evidence sources
verified
“Apple has included neural engine AI components in its in-house iPhone chips for years.”
VERIFIED BY REFERENCE
Wikipedia explicitly states that the Neural Engine was first introduced with the A11 Bionic chip in 2017 for the iPhone 8 and X.
travel_explore
web search NEUTRAL — Neural Engine is a series of AI accelerators designed for machine learning by Apple. Neural Engine was first introduced with the A11 Bionic system-on-a-chip, used in the iPhone 8, iPhone 8 Plus and iP…
https://en.wikipedia.org/wiki/Neural_Engine
travel_explore
web search NEUTRAL — 1. Introduction: Why Future iPhone Chips Matter for Education Apps. 1.1 The hardware-driven renaissance in mobile learning. Mobile chips are no longer just for UI animations; they are driving on-devic…
https://equations.live/maximizing-performance-with-apple-s-f…
travel_explore
web search NEUTRAL — M5 is Apple’s next-generation system on a chip built for AI, resulting in a faster, more efficient, and more capable chip for the 14-inch MacBook Pro, iPad Pro, and Apple Vision Pro.
https://www.apple.com/newsroom/2025/10/apple-unleashes-m5-th…
info
“Microsoft announced a second-generation AI chip in January.”
SINGLE SOURCE
The provided web search results for Microsoft are generic landing pages and do not contain information about a second-generation AI chip announcement in January.
travel_explore
web search NEUTRAL — It’s all here with Microsoft account Your Microsoft account connects all your Microsoft apps and services. Sign in to manage your account.
https://account.microsoft.com/account
travel_explore
web search NEUTRAL — Create your Microsoft account to access various services and features.
https://signup.live.com/
travel_explore
web search NEUTRAL — Explore Microsoft products and services and support for your home or business. Shop Microsoft 365, Copilot, Teams, Xbox, Windows, Azure, Surface and more.
https://www.microsoft.com/en-us
info
“Last week, Meta said it's working with Broadcom to develop multiple versions of AI processors.”
SINGLE SOURCE
The provided evidence for Meta consists of general company descriptions and does not mention a collaboration with Broadcom for AI processors.
travel_explore
web search NEUTRAL — Learn more about Meta and stay updated on our role in social technology, virtual reality, augmented reality, and the future of human connection.
https://www.meta.com/about/
travel_explore
web search NEUTRAL — By logging in, you can navigate to all business tools like Meta Business Suite, Ads Manager and more to help you connect with your customers and get better business results.
https://business.facebook.com/
travel_explore
web search NEUTRAL — Meta Platforms, Inc. (doing business as Meta) is an American multinational technology company headquartered in Menlo Park, California. Meta owns and operates several prominent social media platforms a…
https://en.m.wikipedia.org/wiki/Meta_Platforms
help
“In 2015, the company started using processors it had designed for running AI models, and began renting them to cloud clients in 2018.”
INSUFFICIENT EVIDENCE
No evidence was provided or found in the search results to verify the specific dates of 2015 for internal use and 2018 for cloud rental.
help
“Amazon Web Services announced the Inferentia chip for handling AI requests in 2018, and unveiled the Trainium processor for training AI models in 2020.”
INSUFFICIENT EVIDENCE
No evidence was provided or found in the search results to verify the announcement dates for Inferentia (2018) and Trainium (2020).
schedule
“DA Davidson analysts estimated in September that the TPU business, coupled with the Google DeepMind AI group, would be worth about $900 billion.”
PENDING
schedule
“Google did say the training chip enables 2.8 times the performance of the seventh-generation Ironwood TPU, announced in November, for the same price, while performance is 80% better for the inference processor.”
PENDING
schedule
“Nvidia said its upcoming Groq 3 LPU hardware will draw on large quantities of static random-access memory, or SRAM, which is used by Cerebras, an AI chipmaker that filed to go public earlier this month.”
PENDING
schedule
“Google's new inference chip, dubbed TPU 8i, also relies on SRAM.”
PENDING
schedule
“Each chip contains 384 megabytes of SRAM, triple the amount in Ironwood.”
PENDING
schedule
“Citadel Securities built quantitative research software that draws on Google's TPUs”
PENDING
schedule
“all 17 U.S. Energy Department national laboratories use AI co-scientist software built on the chips, Google said.”
PENDING
schedule
“Anthropic has committed to using multiple gigawatts worth of Google TPUs.”
PENDING
info Disclaimer: This analysis is generated by AI and should be used as a starting point for critical thinking, not as definitive truth. Claims are verified against publicly available sources. Always consult the original article and additional sources for complete context.