Welcome to the AI chat podcast. I'm your host, James Shafer. Today on the show, we're going to be covering the top 10 AI news stories of the week in rapid fire. Before we get into that, if you want to try using AI to build a tool for you, you can go to AIbox.ai, my own startup, describe the tool you're imagining, and have our AI build it out.
We'll connect all of the different AI models together, fill in the prompts, build the UI, and make something incredible for you. If you want to check it out, there's a link in the description to AIbox.ai. I roast Apple for AI all the time, and I think they've turned the tables with iOS 26. So instead of chasing giant cloud models like opening out your Google, Apple is betting really big on local AI.
That's tiny models that run directly on your iPhone. There's no subscription fees. There's no server costs. There's no sending your private data to the cloud.
A lot of benefits if you can pull it up. Developers are already jumping in. Look up now, generates examples and word origins right on your phone using this technology. Moneycoach, they're analyzing your spending and auto tag transactions using these kind of internal AI models.
Day one, another app suggests titles and deeper writing prompts for your journal using it. And crouton breaks down messy recipes into set-by-step instructions. And even sign every or easy sign, sign easy. Can summarize contracts instantly without touching the internet.
These models aren't as powerful as GP4 Gemini, but there is genius in all of this, and that's Apple's AI is invisible, private, and free. So these are features that just quietly make your app better. Apple isn't trying to win AI arms race. They're trying to make AI disappear into your phone.
NVIDIA just pulled off a $5 billion power move, and they bought their biggest rival or huge stake in it, 4% of Intel instantly becoming one of Intel's largest shareholders. But not just about the money. This is about survival in the AI chipboard. Here's the play.
NVIDIA and Intel will now co-develop multiple generations of data centers and PC chips. Intel is going to build a new X86 CPU customized for NVIDIA's AI infrastructure. And for customers, they're making something even more powerful chips that combine Intel CPUs with an NVIDIA's RTX GPUs in a single package. So what's the translation on this?
NVIDIA graphics baked straight into future PCs. If you want to build custom tools with AI code, no code tools, check out my startup, AIbox.ai. So this comes right after NVIDIA just logged another record breaking quarter while Intel's been struggling. Layoffs fail project, shrinking market size.
But with this partnership, Intel gets a lifeline. NVIDIA gets to lock-in dominance before AMD catches up. This isn't just a deal. It's reshaping the entire AI hardware market.
And the crazy part, NVIDIA just turned its old rival into its new secret weapon. Google Ventures just doubled down on a tiny startup that's beating AWS and Azure at their own game. It's called Blacksmith. It is a dev tool company.
And it just raised $10 million. This is only four months after they did their seat round, by the way. And this is because they're making it possible to ship code way faster at a fraction of the cost. So here's how they do it.
Instead of renting expensive cloud servers like AWS, Blacksmith runs CI, CD on bare metal gaming-grade CPUs. That means that there's basically two times a faster of a build and up to 75% cheaper on Compute costs. And switching over takes just one line of code. Since May, Blacksmith has gone from $1 million in annual recurring revenue to $3.5 million.
And they have over 700 customers. They have a team of just eight people, so really small. And they have a bunch of really big customers, including Superbase, Veed, and Ashby. This is why Google Ventures isn't just invested again.
They're actually betting on Blacksmith. And they believe that they basically could redefine how software ships in the AI era. The bottom line of all of this is that if AWS is the cargo ship, Blacksmith is more like a Formula One racing car. And right now, I feel like speed is everything.
Google just turned Chrome into an AI super-assisted. And it might make search engines feel less obsolete. So starting now, Gemini is built straight into Chrome for US users. And it means instead of just Googling, you can ask Chrome to summarize pages, compare tabs, or even rewrite recipe to make it gluten free, all without leaving a site.
So the idea I know, it gets even crazier, though. So Chrome will remember what sites you visited last week, organize multi-tab sessions like trip planning, even fetch specific parts of YouTube videos for you. If you want to build with AI tools, check out mystartupai. So here's the real game changer.
Google is rolling out a Gentic browsing. It's actually going to act for you. So it could book your haircut or your groceries. It could be resettling or resetting your breach password, which personally I would love, and even protecting you from AI-generated scams.
And you'll also be able to type natural questions right into the Chrome search bar, and keep your conversation going with follow-ups like you would with chat GPT. So it isn't just upgrading the browser, they're building an AI co-pilot for the internet. And I am very excited about this. Google just made it official.
Your AI can now swipe your card without asking you. They launched the Agents Payment Protocol AP2 a system where bots negotiate buy and bundle deals for you. Here's the twist. This isn't just Google.
Over 60 banks and merchants including MasterCard, and XPayPilot, Coinbase, and Metamask, are all back in it. That means that your AI could shop in dollars or crypto, and receipts will be cryptographically signed. Why this matters? Shopping's no longer you versus Amazon's algorithm.
It's your AI agent versus the seller's AI. They're battling in milliseconds over bundles, upgrades, discounts, and yes, AP2, you know, that was fully automated purchases. If you give your bots strict spending rules, you can get it to buy stuff for you. You could say, plan your trip for $1,800, and your agent will go, hey, go with airline bots.
It's going to negotiate the hotel upgrades, and it's going to book all of it without you having to do anything. But here's a catch. Whoever owns the protocol owns the rails of Converse. It's going to be interesting to see, you know, who is going to decide what is purchased online.
Opening Eye just signed a $300 billion five-year deal with Oracle and Oracle Stock Price, obviously exploded. But here's the twist. Most assumed that the AI cloud race was between AWS, Google, and Microsoft, but Oracle, the so-called legacy dinosaur, just secured the biggest AI infrastructure contract, literally ever handling 4.5 gigawatts of compute for OpenAI. This matters because it shows how desperate OpenAI is for raw power.
They're already burning billions committed. They have $60 billion a year on Compute, and another $10 billion on custom chips with Broadcom, all of the rate in $10 billion in annual recurring revenue. But Compute without entering the world, it's going to be a huge amount of $10 billion a year on Compute. And then, you know, the real thing is that the energy is basically worthless, and nobody knows whether electricity for this deal is actually going to exist.
So, Simulnet has personal bets in nuclear and solar startups, but OpenAI itself is staying asset-light. They're letting Oracle handle the dirty work, and this keeps the valuation really shiny while it leaves the looming question. Can even $300 billion buy enough power to keep opening eyes growth alive, or does AI's biggest deal yet run head-first into an energy wall? Google's being sued for stealing the spotlight from Rolling Stones and Billboard publisher, Penci Media.
They just found the lawsuit accusing Google of using their journalism to power AI overviews, the summaries that keep readers on Google instead of clicking through. But here's the twist. Publishers already tolerate Google crawling their sites in exchange for traffic, but Penci says that Google has now changed the bargain. They're forcing them to give up content for AI training, and summaries or they're going to lose visibility in the search results.
The only opt-out that they have is not working. So, why does this matter? Penci is basically claiming that its site saw major drops in clicks once the AI overviews rolled out, which were getting ads, subscriptions, and affiliate revenue. While Google insists that summaries drive more diverse traffic, the lawsuit says that there is no evidence to back this up.
And this isn't just another copyright. If Penci wins, it could've decided whether Google gets to rewrite the rules of the open web, or whether publishers fight back against this AI land draft. So, the real battle isn't over-clixed, it's over-weather journalism, fuels AI, or as some dramatic are saying, is going to get completely erased by it. NVIDIA just got completely banned from China, and Huawei is now trying to take over.
At the Huawei Connect Conference, the company unveiled their SuperPOD Interconnect. This is a system that links up to 15,000 GPUs, including their own, ascend AI chips into one massive AI cluster. So, it's kind of like China's answer to NVIDIA's envy link. It is basically the back one that lets GPUs talk to each other at really fast speeds.
And Huawei's chip may not be as powerful as NVIDIA's, but clustering thousands together could give China the compute power it needs to train next-gen AI models. And here's the timing bond. Just one day earlier, China banned its tech companies from buying NVIDIA's hardware, including servers specifically designed for the Chinese markets. So, now Huawei isn't just competing with NVIDIA, they're positioning themselves as the literal only AI chip option left in China.
And this isn't just a product launch, it's definitely a geopolitical chess move, and it seems like China definitely had a little bit of hand on the scale to try to push Huawei up to the very forefront of AI chips in China. CodeRabbit is halfway to Unicorn status. This is an AI-powered code review platform. It just reads $60 million at a $550 million valuation.
And here's the problem. Developers are basically using these AI coding systems like GitHub, CoVilot, and Cloud Code. And it turned out tons of code fast. My startup, AI box included.
But a lot of it is buggy, messy, and sometimes unusable. And that means that basically engineers are spending hours trying to fix it. CodeRabbit is reviewing code like a teammate. So basically they're going to flag bugs, or even going to cut the number of humans needed for review by about half.
Companies like Chag to Group on and Work 3 are already using them. They have, I think, 8,000 businesses using this. And for example, CodeRabbit is growing 20% month over month and already pulling in $15 million in annual recurring revenue. NVIDIA's VC arm even just joined in.
And I think if you want to see a bullish sign, NVIDIA getting on board would be that. And I'd be writing the code too fast for humans to keep up with, which is a problem. But now it seems like startups like CodeRabbit are cashing in by babysitting the bots. The AI race for Avatar is just leveled up.
DID, which I've talked about a lot, is a startup behind all of those talking AI avatars that you've seen everywhere. And they just acquired Berlin based simple shows, this is a beauty video creation platform. And they have over 1,500 enterprise clients, including Microsoft, McDonald's, and eBay. So I think this isn't just about tech, it's about distribution.
Simple shows, SaaS platform has a massive customer base. And DID instantly expanded its reach in training, marketing, and sales videos. Here's the kicker, DID wants to make avatars interactive. So you can imagine pushing pause on a training video to ask the AI instructor a question, or being quizzed in mid presentation by an AI avatar that feels really human.
Of course, the competition is definitely heating up. There's some these, there's soul machines. And even Google and McKinsey are building their own AI avatar tech. But with this acquisition, DID really positions themselves as a serious contender in the enterprise avatar world.
But the question is, will AI avatars replace human presenters, or will they just become another corporate gimmick? Thank you so much for tuning into the AI Chat Podcast. If you prefer episodes that are a deep dive on one particular topic or interviews, we also have lots of that content on the show. So just feel free to take whatever you prefer from the show.
Make sure to go check out AIbox.ai. If you want to get access to the top 40 different AI models all in one place, we'll have to pay subscriptions to all of the different platforms for $20 a month. You can go check out AIbox.ai. We have an amazing AI model playground that people love.
We love to have you as part of it. Thank you so much for tuning into the podcast today. And I hope you have a fantastic rest of your day.