• The AI Furnace
  • Posts
  • AI Furnace Newsroom: Google on a Roll with Gemini Flash, OpenAI Launches its App Store, China Steals Blueprint for NVIDIA AI Chips

AI Furnace Newsroom: Google on a Roll with Gemini Flash, OpenAI Launches its App Store, China Steals Blueprint for NVIDIA AI Chips

The AI Furnace weekly newsletter. Your destination for the latest news, innovations, opportunities, and product launches in AI.

In partnership with

Welcome to this week’s AI Furnace Newsroom

In today’s insights we cover:

  1. Google debuts Gemini 3 Flash - a bargain version of Gemini 3 Pro

  2. ChatGPT goes live with ChatGPT App Store

  3. China steals secret formula for NVIDIA AI chips

  4. Amazon makes $10B OpenAI bet

  5. CIOs get closer to CEOs as AI takes centre stage in enterprise

Read time: 5 mins

💡 Furnace Insights

Google

Google launches Gemini 3 Flash - a lighter weight, just-as-intelligent version of their Gemini 3 Pro model that actually costs 40% less.

The model, while not technically as good as Gemini 3 Pro, is competitive across all frontier models and in some cases even beats them (e.g. it beat Claude Opus 4.5 in the SWE-2 benchmark). This low-cost version can fit straight into tight loops like coding agents, product experiments, search, and planning, where waiting even a second feels clunky. Spelling a big change in enterprise AI and in scenarios where AI is ripe to take over smart workflows.

However, the most important story here is that this model is now embedded at scale across all of Google’s non-AI products, Gemini apps, Search AI Mode, APIs, and enterprise tools, including Google Search…effectively cannibalizing their flagship product.

Why you ask? Perhaps Google sees the writing on the wall and wants to remain the dominant force behind advertising, their largest revenue generator.

OpenAI

Source: OpenAI

OpenAI has rolled out an app directory inside of ChatGPT across web and mobile. Apps connect services like Dropbox, Spotify, Apple Music, and DoorDash via chat.

Developers can submit apps, with monetization starting through external links. Once authorized, apps can be called with an @ mention inside chats to pull files, manage playlists, or plan shopping.

You can now treat ChatGPT like a mini app platform where you ask and it does, instead of bouncing between screens and logins. This is getting us astonishingly close to a personal AI assistant that we rely on and is embedded into our daily lives to help us carry out everyday tasks with ease.

Nvidia

Source: Reuters

It seems China has got their hands on something that Washington has been trying to prevent for many years: a prototype of a machine capable of producing the cutting-edge semiconductor chips that power artificial intelligence, smartphones and weapons central to Western dominance, Reuters reports.

A company called ASML based in Netherlands is the owner of proprietary machinery called ‘Extreme Ultraviolet Lithography’ (EUV) - a crucial part of the AI chip manufacturing process. It is responsible for imprinting nanometer-sized transistors to enable the likes of OpenAI, Google, and Anthropic to create frontier intelligence.

The company has faced no competition due to the incredibly complex and expensive process required to perform EUV, allowing them to enjoy a 100% monopoly. That is, until now.

It is said that China hired ex-ASML engineers and reverse-engineered the entire process, building an identical prototype of ASML’s machine. If true, China would be the first entity to present a viable competitor to one of the most important companies on earth today, setting them on a path to building better chips than Nvidia.

Amazon

Source: The Information

Amazon is in talks with OpenAI regarding a whopping $10 billion investment as the AI chip wars continue to heat up.

According to reports, the deal would see OpenAI use Amazon’s latest Trainium3 chips to train future GPT models, potentially cutting training costs by as much as 50%. If finalized, the partnership would underscore how capital, compute, and model development are increasingly bundled together as hyperscalers compete to anchor the next generation of AI platforms.

Deloitte

Source: Deloitte

CIOs are getting closer to their CEOs as generative AI pushes technology from a back-office function to a core driver of enterprise strategy. A Deloitte survey shows that 65% of CIOs now report directly to the CEO, up from 41% a decade ago as boards increasingly rely on technology leaders to drive growth, efficiency, and transformation.

That proximity is translating into ambition. Deloitte found that 67% of CIOs now say they want to become CEO, outpacing other senior technology roles and signaling how central the position has become to value creation across the business.

For executives, the message is clear: as AI reshapes how companies operate, the CIO role is evolving into a proving ground for top leadership - not just a steward of systems, but a builder of strategy, scale, and revenue generation.

📈 AI Venture Deals of the Week

  • Endra, a AI-powered MEP design platform, raised its $20 million Seed round.

  • Arcads.ai, an AI marketing video creation platform, raised its $16 million Seed round.

  • Datalane, a GTM data/knowledge graph for the local economy, raised its $22 million Series A.

  • Galatek, an AI-driven automation startup for life sciences and semiconductor manufacturing, raised its ~$30 million Series A.

  • Finny, an AI-powered prospecting platform for financial advisors, raised its $17 million Series A.

  • Chai Discovery, an OpenAI-backed biotech startup, raised its $130 million Series B.

  • Lovable, an AI “idea-to-app” builder for rapid prototyping, raised its $330 million Series B.

  • Ankar, an AI-driven IP patenting platform that unifies the entire patent lifecycle into one secure workflow raised its $20 million Series A.

  • PolyAI, an AI voice assistant platform for enterprise call centers raised its $86 million Series D.

⚒️ New AI Product Launches You Don’t Want to Miss 

  • Ray3 Modify, by Luma AI, a Dream Machine video-to-video editing tool that reimagines footage with improved consistency plus keyframe controls and character reference for precise scene/element changes.

  • FunctionGemma, by Google, a Gemma 3 270M variant fine-tuned for native function calling so lightweight on-device agents can translate natural language into executable API actions.

  • Qwen-Image-Layered, released its new image model that decomposes a picture into multiple RGBA layers so each layer can be edited independently (move/recolor/resize) for “inherent editability.”

  • Disney Research Hub, using ML to train a real, self-roaming robot Olaf.

  • Mistral OCR 3 by Mistral AI, an OCR model for high-fidelity document understanding (including forms, tables, and handwriting) that outputs clean text/structured JSON via API or Mistral’s Document AI Playground.

Upcoming Events 📅

Interested in meeting the Who’s Who of AI?

The AI Hot 100 Summit is back for its third edition on May 7th, 2026 in New York City. The only AI Summit built for real connections where startup innovation meets enterprise implementation. Join over 500 AI Executive Leaders, Founders and Investors for a jam packed day of enterprise AI lightning talks, innovation showcases, and curated networking.

Get ready to meet and greet speakers and attendees from OpenAI, NVIDIA, Haleon plc, Legora, Anthropic, Cursor, Eleven Labs, Adobe, Walmart, Insight Partners & more.

🗓️ May 7, 2026

📍New York City

📢 Want to partner? Reach out to get your brand in front of 25k+ AI executives, entrepreneurs, researchers, investors, and AI leaders.