- The 79
- Posts
- Google revealed Ironwood: A new AI accelerator chip
Google revealed Ironwood: A new AI accelerator chip

Hi everyone! Hereās what you need to know about AI today:
š WordPress launched a powerful AI website builder for free
š Googleās Ironwood AI chips are here to compete with Nvidia
š Google adds a ton of new AI features to its Workspace and Vertex AI cloud platform
and many more!
š§ Did someone forward you this email? Subscribe here for free to get the latest AI news everyday!
Read time: 5.1 minutes

WORDPRESS
WordPress introduces free AI website builder

Source: WordPress
Whatās going on: WordPress, the worldās largest and most famous web content management system, has launched a new AI-powered website builder, available at no cost to all users. This tool simplifies website creation by allowing users to describe their needs through a chat-style interface, after which the AI generates a fully functional site in minutes.
What does it mean: The tool handles everything, layout, text, and images; eliminating the need for coding or drag-and-drop design. The goal is to make website building accessible to anyone, regardless of technical skill, while leveraging WordPressās established platform, which powers a significant portion (43.5%) of the web.
More details:
The process starts with logging into WordPress.com and accessing the āCreate Newā option, where you can select āBuild it with AI.ā From there, a conversational prompt will guide you to specify details like the siteās purpose, name, and content preferences. The AI then produces a draft, complete with placeholder images and text tailored to the input.
Users can refine the output by requesting changes through the same chat interface, such as swapping images or rewriting sections, though more advanced edits still require manual tweaks via WordPressās standard tools.
While this new AI builder is free, itās tied to WordPressās ecosystem, meaning users need an account, and premium features like custom domains or hosting may still incur costs.
Interested in the details and want to try it out? Visit this page.
Google unveils Ironwood AI accelerator chip

Source: Google
Whatās going on: Google has introduced Ironwood, its seventh-generation Tensor Processing Unit (TPU). This new AI accelerator chip is designed specifically for inference, the process of running AI models to generate outputs. This is Googleās first TPU optimized for this task. Ironwood will be available later in 2025 for Google Cloud customers in two configurations: a 256-chip cluster and a massive 9,216-chip cluster, aimed at handling large-scale AI workloads.
What does it mean: Ironwood reflects Googleās decade-long investment in custom silicon, offering a viable alternative to Nvidiaās dominant GPUs. While itās exclusive to Googleās ecosystem, accessible only through its cloud or internal teams, itās a strategic move to meet the growing demand for inference computing in AI applications. The chipās scalability and power efficiency address the rising energy concerns in AI infrastructure, though Google hasnāt disclosed its manufacturing partner. As competition intensifies with Amazon, Microsoft, and Meta advancing their own AI chips, Ironwood strengthens Googleās position in the AI hardware space.
More details:
With a peak performance of 4,614 teraflops per chip and a total of 42.5 exaflops in its largest setup, Ironwood positions Google to compete directly with industry leaders like Nvidia.
Each Ironwood chip comes equipped with 192GB of dedicated RAM and a bandwidth of 7.4 terabits per second, supported by an advanced Inter-Chip Interconnect network delivering 1.2 terabits per second bidirectional speed.
This architecture minimizes latency and boosts efficiency, aided by a specialized SparseCore for tasks like recommendation systems and advanced ranking.
Google claims Ironwood doubles the performance-per-watt of its previous Trillium chip, thanks to optimized design and liquid cooling.
Ironwood integrates with Googleās AI Hypercomputer and Pathways software, enabling it to manage complex models like Gemini and AlphaFold, extending its utility beyond AI to fields like finance and scientific research.

ā TSMC may face a $1 billion fine as the US investigates how its chips, made for Chinese firm Sophgo, ended up in Huawei's AI processors, potentially violating export control regulations.
šø Artisan, AI sales agent startup known for its controversial "Stop Hiring Humans" campaign, raised $25 million after experiencing early-stage challenges like customer churn and flawed AI email pitches, but has since improved its technology and is expanding its AI agent offerings while also increasing its human workforce.
š° Tessell, a multi-cloud database-as-a-service startup, secured $60 million in funding to expand its market presence, develop AI-powered database management, and further simplify data management.
šŖ Google Workspace is receiving many AI-powered upgrades, including Workspace Flows for automating multi-step processes, podcast-style overviews in Google Docs, "Help me refine" for writing improvements, "Help me analyze" for Google Sheets, "Take notes for me" in Google Meet, Gemini chatbot integration in Google Chat, Veo 2 model integration in Google Vids, and new data residency controls.
š¤ Samsung is integrating Google's Gemini AI into its Ballie home robot, allowing users to interact with the robot using audio and video for tasks like outfit suggestions and health recommendations, with plans to launch Ballie in South Korea and the US in the first half of 2025.
ā Google is enhancing its Vertex AI cloud platform with updates to its media-generating AI models, including the preview release of its Lyria text-to-music model, enhanced editing and customization options for its Veo 2 video creation model, a generally available voice-cloning feature powered by Chirp 3, and improved performance for its Imagen 3 image generator.
šØāš» Google's Gemini Code Assist is adding new "agentic" capabilities that allow it to deploy AI agents to handle complex programming tasks like app creation and code transformation, putting it in closer competition with rivals like GitHub Copilot.


AI + Preparing for a talk
Outline a 3-month plan to improve my public speaking skills for a 15-minute keynote, including practice techniques, feedback methods, and confidence-building exercises.
Gemini 2.5 Pro's answer

Pallet - Voice AI Engineer
Striim - Senior Software Engineer, AI/ML
Extend - Senior Data Scientist
Moloco - Data Scientist II, Product Analytics
Thank you for staying with us like always! If you are not subscribed, subscribe here for free to get more of these emails in your inbox! Cheers!