🍦 Nvidia Swallows Groq for $20B
Nvidia eliminates its biggest threat, Lovable jumps to $6.6B in 5 months, and your doorbell now has an AI receptionist
FOMA AI News
December 25, 2025
Scooped from 45+ sources. Read it fast before it melts.
⚡ If you only have 1 minute...
Nvidia Acquires Groq for $20B: Historic AI chip deal as AI shifts from training (Nvidia’s stronghold) to inference (where they face pressure from Groq, AMD, Cerebras)
Lovable raises $300M at $6.6B valuation: This was $1.8B just five months ago. At what point do we sit the VCs down, hold their hands, and gently ask what’s going on at home?
OpenAI raising $10B from Amazon: Which backs Anthropic. Which backed by Google. Microsoft backs both. Nvidia’s in all of them. Somewhere there’s a detective pinning photos to a corkboard with red string trying to figure out who’s actually competing with who
💰 Market Moves
[ACQUISITION] Nvidia Acquires Groq for $20 billion
Historic AI chip deal, as AI shifts from training (Nvidia’s stronghold) to inference (where they face pressure from Groq, AMD, Cerebras)
[FUNDING] Lovable raises another $300 million at a $6.6B valuation
This was $1.8B just five months ago. At what point do we sit the VCs down, hold their hands, and gently ask what’s going on at home?
[FUNDING] OpenAI are raising $10B from Amazon
Which backs Anthropic. Which backed by Google. Microsoft backs both. Nvidia’s in all of them. Somewhere there’s a detective pinning photos to a corkboard with red string trying to figure out who’s actually competing with who
[ACQUISITION] Cursor acquired Graphite
Because devs couldn’t review code fast enough. So now AI writes the code. AI reviews the code. Can you human developers at least click the big “merge” button or do we need to acquire someone for that too?
[ACQUISITION] Flipkart bought Minivet AI
To turn product photos into AI videos. Can’t wait to watch a deepfake kettle rotate for 45 seconds straight
[FUNDING] Databricks raised $4 billion a Series L
Yes, Series L. Most companies IPO by Series C. Databricks looked at the alphabet and said “we’re not done yet”
[FUNDING] Yann LeCun confirmed his new startup AMI (Advanced Machine Intelligence)
Seeking a $5B+ to build the AI he’s been insisting everyone else should’ve built all along. He’s not CEO (someone has to actually run it while he posts), but he’s executive chairman of telling everyone “I told you so.”
[ACQUISITION] ClickUp acquired AI coding startup Codegen
Cursor rival to build their agents, for an undisclosed amount, which will probably be disclosed by an intern on LinkedIn next week.
[FUNDING] Figure AI raised another $1B
Now at $39B valuation for AI humanoids. With this much money, the robot boots up thinking it’s the Terminator. While in reality I imagine him waking up, looking up at the head of R&D, which hands him a note the reads: “You fold laundry”🤣
[ACQUISITION] Intel are buying AI inference company SambaNova for $1.6B
Fun fact though, Intel’s CEO? yeah he is also SambaNova’s Executive Chairman. AI companies are like that Spiderman meme where they all pointing at each other
🛠️ Tech & Product
[LAUNCH] Google dropped Gemini 3 Flash
It’s faster than Pro, cheaper than Pro, and beats 2.5 Pro. At this point Google is just competing with itself
[LAUNCH] OpenAI released GPT-5.2-Codex
With “improved long-horizon work,” which is nerd for “it works 8 hours straight without asking if you’re still watching.” kinda like that Netflix feature, just without refactoring your entire living room while you’re asleep
[LAUNCH] OpenAI launched ChatGPT Images
It makes precise edits, and generates images up to 4x faster than before, its also available in the api as gpt-1.5-image
[LAUNCH] ClickUp launched “Super Agents”
AI employees you can @mention just like real teammates. They work 24/7, have “infinite memory,” and one prompt spins up an entire team. HR is gonna have a really weird time explaining this on the next all-hands
[LAUNCH] OpenAI introduced new personalization options for ChatGPT
Allowing users to adjust enthusiasm, warmth, emoji use, and formatting
[LAUNCH] Manus launched Manus 1.6
This release introduces three major leaps forward: Manus 1.6 Max, their most powerful agent yet, Mobile Development, for building beyond the web, and Design View, for interactive image creation.
[LAUNCH] OpenAI has launched app directory
Allowing users to search, discover, and install apps, developer submissions are also now open
[LAUNCH] Kling shipped v2.6
With three upgrades: (1) upload custom voices so AI characters speak consistently across videos, (2) automatic audio generation (dialogue, sound effects, ambient noise), and (3) better motion tracking for dance/martial arts
[LAUNCH] Zoom launched AI Companion 3.0
Analyzing your Zoom meetings, Team Chat, and connected apps (Gmail, Drive, OneDrive) to automatically generate daily summaries, draft follow-ups, and surface key insights
[LAUNCH] Luma AI launched Ray3 Modify
Which lets you film an actor once, then change the video’s background, actors clothes, or entire visual style with a a single prompt
[LAUNCH] Tiny Fish launched Mino
A web agent API that logs into websites for you, handling passwords, CAPTCHAs, and auth barriers to access data behind login walls that normal scraping can’t reach.
[LAUNCH] Anthropic launched Claude in Chrome beta
With Claude Code integration, workflow recording, scheduled tasks, planning mode, multi-tab workflows, and browser control from desktop
[LAUNCH] xAI released the “Grok Voice Agent API”
Real-time speech-to-speech at $0.05/minute (half the price of OpenAI), sub-1-second response time, multilingual, can call tools, search the web and X
[LAUNCH] DoorDash is testing Zesty
A new app that replaces “scrolling through 47 TikToks about brunch spots” with an AI that just tells you where to go. It pulls from multiple sources and has Instagram-style social features. No ordering yet, just discovery
[LAUNCH] ElevenLabs now integrates with WhatsApp
Letting businesses deploy AI voice agents that can answer both calls and texts from customers through Whatsapp
[LAUNCH] Google Labs launched “CC”
An AI productivity agent that reads your Gmail, Calendar and Drive to send you a daily “Your Day Ahead” email with meeting prep, drafted replies and task summaries
[LAUNCH] Lovable launched direct ChatGPT integration
So you can brainstorm an app idea in ChatGPT, tag @Lovable, and it builds you a working full-stack web app straight from the chat.
[LAUNCH] Meta launched SAM Audio
An AI model that lets you isolate any sound from audio by typing what you want, kinda like Photoshop but for sound,and they’re already testing it with hearing aid manufacturers
[LAUNCH] Langflow 1.7 was released
Featuring upgrades to MCP transports with Streamable HTTP, new agent options, and additional components for flows
[LAUNCH] xAI released “Collections”
A RAG-as-a-service API that handles all the embedding, indexing, and vector search infrastructure for you, just upload docs and query them with semantic, keyword, or hybrid search.
[LAUNCH] Vercel AI SDK 6 is out of beta
With everything developers have been duct-taping together manually: ToolLoopAgent (handles the full tool execution loop), tool approval (human oversight before your agent does something stupid), MCP integration that’s actually stable, and DevTools to see why your agent decided to call weather() 18 times in a row
[LAUNCH] Firecrawl launched “Agent”
You don’t need URLs anymore. Just tell it what data you’re after and Agent searches the web, figures out where to look, clicks through pages, and comes back with a dataset
[LAUNCH] Amazon has launched a new built-in AI shopping assistant called Rufus
It knows what you bought before, can reorder past items, automatically add products to cart, and even find stuff outside Amazon to buy for you
[LAUNCH] YouTube launched “Playables Builder”
Where Gemini 3 turns text descriptions, images, or videos into actual playable games that viewers can play right inside your channel. No coding, no game dev skills, just feed it prompts and it generates working mini-games in minutes
🗣️ Industry Chatter
[RUMOR] Meta is developing two new AI models codenamed Mango and Avocado
Mango will handle visual tasks like image/video generation and recognition (targeting H1 2026), while Avocado focuses on coding. This marks Meta’s push into multimodal AI beyond just language models
[DEMO] Unitree’s G1 humanoid robots performed synchronized dance routines
At a concert, demonstrating coordinated multi-robot movement control. The $16k robots executed choreographed sequences together, showing progress in robot-to-robot coordination for entertainment and industrial applications
[LAUNCH] Amazon launched “Greetings” for Ring doorbells
An Alexa+ feature where conversational AI chats with visitors when you’re not home, answering questions about delivery instructions or taking messages. It’s essentially an AI receptionist for your front door
[RUMOR] The US Air Force is upgrading its experimental X-62 F-16
With advanced radar and sensors specifically designed to test AI-controlled wingmen. The jet serves as a testbed for autonomous fighter systems that will fly alongside human pilots
[BETA] Google is testing a ‘Lecture’ mode in NotebookLM
That can generate approximately 30-minute, single-voice audio lectures in various languages
[NEWS] YouTube started mass-deleting channels
That use AI to generate fake movie trailers for non-existent films. The crackdown targets channels creating deceptive content that racks up millions of views by promoting fake sequels and crossovers
[LEGAL] Google is suing SerpAPI
A scraping company used by ChatGPT, Cursor, and Perplexity
[RESEARCH] Anthropic ran an experiment
Giving Claude access to a vending machine to manage
[LAUNCH] Alibaba’s Qwen models can now clone voices
From three seconds of audio and generate expressive speech from text, Qwen claims the model achieves a lower error rate than competitors like Elevenlabs or MiniMax
🍦🔥 Dare to summon the bot?
Our Ice Cream mascot is desperately looking to make new friends. Share us on LinkedIn and tag @FOMA AI News. We guarantee a snappy comment (or a light roast) on your post within 24 hours.
Warning: Bot has zero chill.