EP 610: Microsoft and OpenAI change partnership, publishers partner against AI scraping, Oracle’s $300B AI play and more AI News That Matters

Topics Covered in This Episode:

1. OpenAI and Oracle $300B Cloud AI Deal
2. Oracle Joins Big Four Cloud Providers
3. Anthropic Claude Launches File Creation Feature
4. Claude File Creation for Enterprise and Teams
5. Microsoft Integrates Claude Into Office 365
6. Microsoft Copilot Adds Multi-Model AI Support
7. Apple Loses Senior AI Executive to Meta
8. OpenAI Adds Anthropic’s MCP Protocol to ChatGPT
9. Major Publishers Launch Real Simple Licensing (RSL) Standard
10. RSL Targets AI Web Scraping and Licensing
11. Microsoft Building In-House AI Chip Clusters
12. OpenAI–Microsoft Partnership Restructuring and Future IPO
13. OpenAI Revenue Share with Microsoft Declines

Keywords:

Microsoft and OpenAI partnership, OpenAI, Microsoft, Oracle, $300 billion AI deal, Oracle Cloud Infrastructure, OCI, hyperscaler investment, generative AI demand, enterprise AI, cloud providers, Amazon Web Services, Microsoft Azure, Google Cloud, big four cloud, Claude, Anthropic, Claude file creation feature, coding workflows, AI documentation, software development, managed project files, AI-enabled teams, sandboxed execution, chatbot integration, Microsoft Copilot, Office 365 AI, ChatGPT, Claude 4, model integration, dynamic model selection, performance optimization, AWS, AI governance, Apple AI leadership, Apple Siri departures, Meta AI hires, media publishers AI licensing, Really Simple Licensing, RSL, web scraping, AI content licensing, AI training data, legal compliance AI, monetization AI data, revenue sharing AI, publisher partnerships AI, AI protocol, Model Context Protocol, MCP, developer mode ChatGPT, Zapier, AI connectors, API integration AI, read-write access AI, AI workflow automation, Microsoft AI chip clusters, self-sufficiency AI, frontier models AI, nonbinding agreement, public benefit corporation, PBC, OpenAI IPO potential, SoftBank Stargate partnership, AI data centers, AI legal gray area, AI model economics, AI business strategy, vertical video generation AI, Meta licensing AI images, Perplexity AI, government AI access, on-device AI, AI browser, custom AI reports, ByteDance AI image model, Google Nano Banana, AI ads, Albania AI minister.

Podcast Transcript

Jordan Wilson [00:00:16]:
One of the landmark partnerships that has defined the generative AI age is changing. The another thing that's changing, the way that large language models may be getting their information in the future after some major publishers have partnered together in a $300,000,000,000 play in AI by Oracle. Maybe we didn't see a huge new model released this week from Google or OpenAI, but we have some partnership news that is technically going to impact tens of millions of people and how they work. So we're gonna be covering all of that today on everyday AI and a whole lot more. Thanks for tuning in, and thanks for joining us. What's going on y'all? My name is Jordan Wilson, and welcome to everyday AI. This is your daily livestream podcast and free daily newsletter helping everyday business leaders like you and me not just keep up with AI because there's a ton going on, but how we can make sense of all of these updates and behind the scenes happenings and take something away to grow our companies and our careers. So if that's what you're trying to do, it starts here with the unedited livestream podcast.

Jordan Wilson [00:01:28]:
But to take it to the next level, you gotta go to our website y'all. It's @youreverydayai.com. So there, two things you gotta do. One, sign up for the free daily newsletter where each day we recap the livestream podcast like we're going over now. But, also, we keep you in, in line and up to date with everything that's happening in the world of AI, not just what we're going over in today's show. So, yeah, most Mondays, we go over, kind of this special segment, the AI news that matters. So you don't gotta waste hours every single week or hours every single day trying to figure out how these things are gonna impact your line of work. Just join us on Mondays.

Jordan Wilson [00:02:05]:
We do our no BS straight to the point recap. So with that livestream audience, I see you there. Thanks for joining us, but let's get straight into it. I'm in a new place. You can see, you know, in in Wisconsin now, so not my normal Chicago, but thanks for everyone else for tuning in live. Jay, joining in from West Virginia. Angie, good morning. Marie joining from New Jersey, Fort Lauderdale.

Jordan Wilson [00:02:33]:
Joe says he's not a Florida man. Alright. Let's get into it y'all. Our first piece of AI news, $300,000,000,000. That's a big one. Right? That's a number you don't see floated around a lot. So let's talk about this new $300,000,000,000, deal between OpenAI and Oracle. Technically, kinda new, but kind of, amending and updating their previous agreement.

Jordan Wilson [00:02:58]:
So Oracle has reportedly signed a five year computing power commitment with OpenAI worth about $300,000,000,000, driving a 359% surge in Oracle's reported future, contract revenue this quarter and a $317,000,000,000 increase in backlog. So if this deal is fully utilized and there is some contingencies that it might not be an actual $300,000,000,000 contract in the long run, but the deal could put Oracle's cloud revenue runway near a trillion dollars by 2030. A figure that underscores how generative AI demand is reshaping where hyperscaler, investment flows. So Oracle Cloud Infrastructure or ACI, well, now you gotta say there's a big four in cloud. You can't not include Oracle Cloud or OCI anymore. So joining Amazon Web Services, Microsoft Azure, and Google Cloud. So now for the big AI cloud providers, the big three have officially become a big four, with Oracle. You can't make an argument against them with this $300,000,000,000 play.

Jordan Wilson [00:04:07]:
So according to reports, OCI's approach, which is identical regions, tight capacity planning, and engineering first delivery was a key reason that OpenAI opted in alongside Oracle's strengths in data movement and high performance infrastructure. Skeptics, however, note the financial bar is steep. So for OpenAI to fully fund a $300,000,000,000 commitment over five years, it would need roughly $60,000,000,000 in annual revenue minimum, right, just to make this thing work, and that's around six times its current reported run rate. So, So, yeah, some people are doing the math and the math isn't necessarily mathing, but obviously, there's some other AI news that could shape how OpenAI may be able to raise funds and they may be able to bring money in aside from just their annualized, revenue. So for Oracle database fusion and ERP customers, Gartner says that this is a net win with Oracle expected to keep investing in core apps while offering access to expanded GPU capacity. And also the broader market implication is faster AI forever. Right? Now that Oracle has this big partnership in place, building off of their previous partnership, with OpenAI and Stargate, This means that all the other cloud providers are gonna have to continue to keep up and push each other, push each other. So, ultimately, what this means, especially for enterprise companies that are looking to take advantage of generative AI, you know, we always talk about, oh, running things in the cloud.

Jordan Wilson [00:05:40]:
Well, for the most part, there had been three big cloud providers, and now Oracle knocking on the door and saying, yeah. Now there's a big four. Alright. Speaking of new new features, at least, that I think are actually pretty big. So anthropic has rolled out a new file creation feature for Claude that can generate, modify, and manage project files, making it more useful for coding tasks, documentation, and data workflows. So the feature supports end to end tasks such as scaffolding repo repos, editing code, writing config files, and producing cloud artifacts, which can speed up software and content development for teams. So Quad operates with controlled external access via an allowed white list of domains, including api.antropic.com, github.com, and other places where you can kind of create these files. So right now though, unfortunately, this isn't going out to all plans.

Jordan Wilson [00:06:41]:
So if you're logging in, yeah, this is, this new file creation feature has actually, grabbed a lot of headlines the last few days, but a lot of people are like, wait. I don't have this. That's because right now, this feature is only for team and enterprise accounts, and team and enterprise administrators have to first enable or disable the feature for their organizations, to be able to use it. So for other users, including free pro, you know, the $20 month and the $100, plan, max apparently didn't even get this file right away. So public sharing also will be disabled, for conversations that involved file creation. So enterprise users though that do have, their hands on this and Teams users, but on the enterprise side, users get sandboxed execution. So environments, are never shared between users supporting cleaner separation of workflows and easier auditability. So Anthropic says that it continuously tests and red teams, this features and recently updated their documentation to reflect this.

Jordan Wilson [00:07:43]:
So, like, what does this mean? Well, essentially, most large language models by default have real problems creating files. Right? You have to actually, either sometimes just get lucky. Right? Whether you're working with, Copilot, chat g b t, etcetera, or you have to know a little bit about prompt engineering using the right mode, etcetera. So, you know, as an example, you might have to call out, and have, chat g p t use advanced data analysis mode and not just use the default g p t five in order to create a file. So now this is a new feature that Claude has started to roll out, across teams in enterprise accounts, which is eventually gonna be adapted by everyone because it's going to have to be. You you know, I think earlier on, a lot of people weren't necessarily as concerned about creating files inside of a large language model because they were like, okay. Well, you know, I'll just copy and paste it. But as large language models obviously get more robust, get more features and functionality, right, get the ability to create PowerPoints.

Jordan Wilson [00:08:49]:
Right? So, you know, ChatGPT agent mode can do a lot of these things by default. They can create, Excel sheets now, PowerPoint now, but agent mode is extremely slow and a little clunky right now. So now this is pretty big news with Anthropic Claude rolling this out to their broader chat. Right? Just normal Claude being able to create PowerPoints, docs, Excel sheets, etcetera. Pretty big news, but don't like Anthropics rollouts not giving access to all paid users. So, unfortunately, this might be something if you're just on the base, $20 a month Claude plan. I don't know. You might have to wait a couple months, a couple of quarters, but, hopefully, it's a little faster rollout than that.

Jordan Wilson [00:09:30]:
Speaking of Claude, well, Microsoft is adding Anthropics Claude to their Office three sixty five mix. Pretty big news here, and we have more news, I think, on our last news story with Microsoft and OpenAI, but Microsoft definitely shaking it up just as OpenAI shook it up with their Oracle partnership kind of going outside of the, Azure cloud. So right now Microsoft is widening its AI lineup inside of Office three sixty five. Now routing tasks between Anthropics Claude and OpenAI's ChatGPT models to improve performance and reduce reliance on a single provider. So now Microsoft will begin to integrate Claude models alongside ChatGPT across their office products, including Word, Excel, and PowerPoint with Microsoft dynamically choosing models based on demand, workload, and task fit. I'm not a big fan on that. I think it's better to allow users to choose which model that they want to use. So I'm not a big fan of that.

Jordan Wilson [00:10:39]:
And one of the reasons why, right, is you wanna be able to, work large language models into your daily workflow. So if you're constantly, you know, using Microsoft Copilot, and using their, obviously, AI features to help across Word, Excel, and PowerPoint, you're gonna get drastically different results, which is gonna require drastically different inputs, whether it's using any of the GBT five models versus using the Claude four family. So although I think it's great that Microsoft is starting to incorporate other models, not a huge fan with Copilot doing this dynamically. So a little bit more on this partnership. So the partnership with Anthropic runs over AWS, meaning that Microsoft will pay AWS, and that one's interesting. Right? So Microsoft having to pay one of their competitors. So AWS is one of, Claude's big backers, and also that's how Claude runs. So Microsoft is having to pay AWS to access Claude even though they have their open AI models running on their own Azure architecture.

Jordan Wilson [00:11:40]:
So Microsoft, though, says there should be no change to current Copilot or Office three sixty five pricing even though their pricing is increasing with the new claud inclusion, signaling a focus on capability gains rather than a short term monetization strategy for, Microsoft. So, yeah, it's not like Microsoft is using this as a key feature to try and go and get more, users across their Copilot, enterprise suite. They're just, you know, hoping for more reliability, a little bit more versatility as well. So Microsoft said that performance was a key driver, and, reportedly, they said that Claude Sonnets four family, was a little bit better in helping generate slides, spreadsheets, and PDFs directly in chat and has shown advantages on office specific tasks like more polished PowerPoint design and automated Excel functions. So this move, it's not necessarily very unique within Copilot because Microsoft and their Copilot product had pretty much been just running on the GPT family and the o series family of models when there was an o series family of models. But, essentially, Microsoft had been exclusively over the life of their front end, Copilot just been offering OpenAI's models. But, it follows their GitHub Copilot multi model approach, which they've already been using for a lot a long time on their AI coding platform Copilot. They let developers pick from chat g b t, x AI, Gemini, and Claude suggesting maybe this might mean a future where Microsoft offers many different, models inside their product stack.

Jordan Wilson [00:13:21]:
I do think or gonna hope if they go there, I do hope that means that they'll allow, users to actually select which models make the most sense. And this shift comes amid an uneasy but still ongoing Microsoft and OpenAI relationship despite Microsoft holding a reported 49% equity stake in OpenAI with a total of about $13,000,000,000 invested in the company over the years. But like I said, that partnership is definitely changing. Another one walks out the door. Yes. Apple has lost another one of their senior AI leaders. And this one, again, another one not looking good for Apple. So according to Bloomberg, Robbie Walker, one of Apple's most senior AI executives and the former head of Siri, is leaving the company next month, raising fresh questions about Apple's already cautious approach to artificial intelligence.

Jordan Wilson [00:14:27]:
So, Bloomberg reports that Walker has led Apple's answers, information, and knowledge team since April and previously ran Siri. So the news lands as Apple actually trails by a lot, by a huge distance, their biggest rivals on consumer AI features in their Apple intelligence, including the built in chat GPT tie in, has rolled out extremely slowly and has carried with it a bunch of lawsuits for essentially Apple promoting all these Apple intelligence features that just don't exist. So and we've seen that Apple has had to form a partnership with one of its biggest rivals in Google for a future smarter version of Siri. So Apple did not immediately respond to this big news, but internal leadership, shifts continue, as, Apple has lost just about all their big AI names and leading AI researchers. So recent talent losses to Meta have intensified pressure as Bloomberg has reported that, Roming Ping, who is Apple's top executive for AI models, plus other researchers such as Mark Lee and Tom Gunter have joined Meta's superintelligence labs teams. So the timing, like I said, follows last week's Apple's product event where they introduce new iPhones and a slimmer iPhone while also just not really talking a whole lot about AI, and that's probably a good thing because they would just be talking about other companies that they're having to pay in order to provide AI to their consumers. Another one that's kind of a strange partnership. Right? We've had a lot of strange partnerships, in AI news this week.

Jordan Wilson [00:16:20]:
Right? We just talked about Microsoft having to pay Amazon to use Claude even though they're partnered with OpenAI, OpenAI, and Oracle. And now we have OpenAI and Anthropic, because OpenAI has officially added in Tropic's very popular model context protocol directly inside ChatGPT. So OpenAI did release something called developer mode, this week, which allows people to enable, the model context protocol or MCP directly inside of ChatGPT. So this gives, people the access and ability to run their own MCP servers and also let ChatGPT run, read, write commands from within a chat. So OpenAI did even admit themselves that this is, quote, unquote, powerful but dangerous with safeguards and guidance needed to reduce risks. So this is why they, kind of rolled this out in what they're now calling a developer mode. So you have to go into your chat GPD settings, depends on what plan you're on. Obviously, if you're on enterprise teams, a paid plan, etcetera, but you have to first enable, developer settings, and then you are able to use the MCP protocol within chat GBT, which is actually a huge deal.

Jordan Wilson [00:17:39]:
Right? So in the same way that you can kind of think, Zapier. Right? Zapier can connect you if you, you know, use Zapier's, you know, custom GPT or something like that. It can connect you to 6,000 or 7,000 different pieces of software. Well, MCP is the same thing. Well, Zapier has an MCP that you can use inside of chat GPT. Wow. That's a lot of acronyms. But this also opens up to just all MCP, servers.

Jordan Wilson [00:18:06]:
And we probably should do, a show on MCP now, especially, since ChatGPT is supporting it. So I don't know. Live stream audience, Spotify people, just comment MCP, like, yes or no if you want the show. I've said for months, I'm like, yeah. We'll do one eventually, but I think this might be, the the thing that's really gonna push MCPs into the spotlight. Even though Microsoft Copilot started to, support them natively inside of Windows, I actually think it is ChatGPT's 700,000,000 weekly active users that could thrust MCPs into the real spotlight here and not not even for good reasons. I think this is actually what's gonna open the door for a lot of, prompt injections, malicious attacks that for the most part, you haven't really seen inside of Chatt GPT. So although this does open up a wide range of capabilities when it comes to what you can do and what you can accomplish.

Jordan Wilson [00:19:06]:
Right? Essentially, giving read write access to anything on the Internet now, but that's actually can be bad because you're getting read, write access to anything on the Internet within your ChatGPT account. So, this is a door that definitely swings two ways, but I think this is actually gonna be the thing that is going to make MCPs something that's not just for developers or dorks like me anymore. So if you're like, what the heck is an MCP? You can kinda think of it like a universal port for AI tools. Right? How there's, you know, a USB, you know, free or something on your computer and on your iPhone and on other devices. Well, MCP is kind of like that. It's like a universal port for AI tools. So developers though can also expose functions through these MCPs such as checking inventory, updating records in your back end tools that you use, and even processing payments through different MCP servers. So, safety controls though require explicit confirmation for any right action with full JSON payload review.

Jordan Wilson [00:20:12]:
Tools labeled read only hint are treated as read only while others are blocked until approval. So open AI though is trying to center their approach on safe in chat usage with guardrails while anthropics documentation focuses on MCP as open infrastructure for enterprise scale systems and a growing ecosystem for prebuilt servers. So, y'all, this is actually big. Right? Because you can do like I said, you could do a lot of these things already by going through Zapier. But number one, you have to pay a lot of money. Number two, it can be a little slower, essentially, if you're using, an intermediary. So this is kind of like how websites have APIs to talk to each other. So MCPs kind of, are like in, in an API type protocol for websites.

Jordan Wilson [00:21:06]:
So it can allow different AI systems to talk to different AI systems and then different AI systems to talk, to different back ends. Right? So think if your company has, you know, a CRM and an ERP and they have all of these different systems, well, now Chad GPT can probably be able to run off those because there's, probably already a publicly available, MCP tool that will do it, or you can build your own MCP tool. So, again, the whole read write giving chat g b t read write access to, in theory, an unlimited number of systems. This really does change what you can do with chat g b t. It changes. So if you've used chat g b t's newer, connectors so it's kind of like a connector, and and MCP, although they are, technically, from a technical standpoint, much different built much different from a functionality statement, or from a functionality viewpoint. They're kind of the same. Right? So, ChatGPT, a couple of months ago rolled out these connectors.

Jordan Wilson [00:22:08]:
So, you know, Gmail for, for an example. ChatGPT can always just know your Gmail, and it can go in and read your Gmail. It can't send anything. Right? So it's the same way. These different MCP servers are gonna be able to access just about any data that you want, any dynamic data that your company uses. So huge news there. Alright. This one could be even bigger.

Jordan Wilson [00:22:32]:
So I'm excited about this one. A lot of like, a lot of people know, my background is in journalism. So I've I've been talking about for now almost three years about how the future of media and publishing, can go, and this is a huge new partnership that could influence that. So, now you have some of the world's biggest publishers that have come together. So including, companies like and publishing companies like Reddit, Yahoo, Medium, Quora, People, O'Reilly, Wikihow, Zipf Davis, all coming together to support a new open standard called really simple licensing or RSL. So this sets paid terms for how AI systems like chat g p t, Gemini, Copilot, Claude, etcetera, how they can access and crawl and use content on your websites. So the RSL standard adds licensing instructions to these different robot dot text in, essentially, these AI bots that crawl different websites, and it tells them and embeds the terms in different web material, books, videos, and datasets, enabling large language model systems to be charged via a subscription, a pay per crawl fee, or a pay per inference when an AI references their work. So, this initiative is organized by the RSL Collective led by RSS co creative Eckhart Walder and former ask.com CEO Doug Leeds, who aims to create a scalable business model that defines rights and compensation across the lab.

Jordan Wilson [00:24:19]:
I'm actually gonna try to get, I was talking with Doug, yesterday. So gonna try to get Doug on the show. If you guys think it would be an interesting show, I would love it, just to talk about how this might impact the future of work because this is big because this could, in theory, obviously, keep a lot of publishing companies in the game longer, but also create a much more equitable, kind of, compensation package for all of these companies that they're you know, we've seen the stories about how, you know, the traditional web is dead and how major publishers now because of large language models, for the most part, just scraping their websites and using all of their information without attribution, without compensation. You've seen a lot of these companies over the past couple of years lose up to sixty, seventy, 80% of their, website users. And a lot of times, that means losing 60 to 70 to 80% of their revenue. So you've seen these huge layoffs across the industry. So this new RSL or real simple licensing, agreement could be huge. So RSL differs from individual licensing deals struck by publishers.

Jordan Wilson [00:25:27]:
Right? So we've seen, you know, Reddit as an example have, you know, multimillion dollar, annual partnerships with some of the big large language model makers. So RSL is different, from these individual deal, deals because it offers any website or creator a standardized plug in way to get paid without negotiating separate contracts. So on enforcement, RSL by itself cannot block bots, but the collective is working, with Fastly. So only AI crawlers with a valid license ID essentially. Right? Saying, hey. You're only gonna be able to crawl websites if you're registered through there. So then publishers would need to, we'll need other providers to build similar gatekeeping tools. Right? So this is one of those things it's only truly gonna work if all of the big names get on board.

Jordan Wilson [00:26:22]:
Right. But from what I've read, it's fairly free and easier for small, indie, and individual content creators, to essentially start using this technology. So the collective says that members will share legal enforcement costs, comparing the approach to ASCAP or ASCAP in music, yet acknowledging that AI training data and web scraping remain in a legal gray area amid ongoing lawsuits involving all the big names such as Ready, Getty Images, and other major publishers. So for companies and creators, this is big and it could open up a new revenue stream from AI usage and provide clear compliance rules for data access at scale. So y'all, I've been saying this literally for almost three years. I said there's only three ways that this works out in the long run because the cat's already out of the bag and this is going to take years, if not decades to legislate or judicate in the courts. Right? Especially here in The US. So the three options for, big publishers are, well, number one, you die.

Jordan Wilson [00:27:29]:
Number two, you sue. Or number three, you partner. Right? And there hasn't been a good way to partner aside from if you're one of the biggest websites in the world. But what about for everyone else? So this is a great way, for it looks like a great way. So I'm gonna be investigating more and hopefully, bringing the, RSL team on the Everyday AI show soon to talk about this. But this seems like a great way for all different companies, to choose the third option there because there's literally no other option. Because eventually, especially if you if your company creates content to get leads, right, which is what for the most part, whether you're in b two b, whether you're in b two c, whether you're a a small, you know, solo creator or you're a Fortune 500 company, Companies put content on their website to attract their ideal client and customer and eventually bring them on as a client, as a paying client or customer. So there's obviously been a weird awkward reckoning going on, I think, especially in 2025 as we've seen, kind of this shift toward large language models essentially being an answers engine.

Jordan Wilson [00:28:36]:
Right? And not just relying on old, scraped content. Right? So many, so many of the large language models, Chat GPT, even Claude now, Gemini, etcetera. They're using just real time web scraping to provide answers. So a lot of times, they're not users aren't going to these websites. They're not signing up for these, you know, services or, you know, being able to get retargeted, you you know, by visiting a website. So this is huge, and I can't emphasize enough. Even though this might be a small thing, I think this could be a big win, for the media, for journalism, for content. Because here's the other thing, large language models are eating from the hand.

Jordan Wilson [00:29:17]:
Like, they're eating the actual hand that is feeding them. Right? Because you've had so many large, enterprise media companies that have had to lay off hundreds or thousands of employees. And when you go back and look at it, it's like, yo. It's their work that actually allow these large language models to be as smart and as robust as they are. Alright. I'll get off my my soapbox on that one, and let's move on. I think we have two more, two or three more big stories here. So let's get to it.

Jordan Wilson [00:29:46]:
So next, Microsoft, according for to reports, is moving to cut reliance on OpenAI as they plan significant investments in their own AI chip clusters to reduce dependency on OpenAI according to Business Insider reports. So this is signaling a push for self sufficiency while keeping their partnership in place. So, Microsoft AI CEO Mustafa Suleiman told employees reportedly that the company should be able to build world class frontier models in house and use other models pragmatically, underscoring a strategy to play a close second to OpenAI in three to six months. So what does this mean? Well, essentially, Microsoft is going to be building its own chip clusters to build more powerful models to reduce their reliance on OpenAI. So Microsoft is testing third par party models in Copilot. Aside from that earlier partnership that we talked about, with Claude in, Infropix Claude in, Microsoft Office products inside Copilot, but Microsoft is also testing other third party models in Copilot and building its own off frontier models aiming to diversify suppliers and lower costs for enterprise customers who rely on Microsoft three sixty five in Azure. So a little bit more on the Microsoft in OpenAI, but this is now our second AI news story of the day, and we haven't even gotten to the big one yet that shows that Microsoft and OpenAI technically are third. Sorry.

Jordan Wilson [00:31:26]:
They're kind of diversifying. So OpenAI, or sorry. Microsoft is now, starting to build its own chip clusters to build more powerful models. We have the story earlier about Oracle and OpenAI's potential $300,000,000,000 cloud AI deal and also Microsoft starting to integrate and implement, anthropic quads models inside of Copilot. Alright. Moving on. And this is this is our big one, y'all. This is the big one.

Jordan Wilson [00:31:59]:
Actually a two part story here, two parts to it. So first OpenAI in Microsoft have agreed to move forward with a new partnership. Their old partnership is gone, and this is pretty big especially if your company is an OpenAI enterprise. So OpenAI and Microsoft signed a nine nonbinding agreement to let OpenAI convert its for profit arm into a public benefits corporation or PBC, while the OpenAI nonprofit side keeps control of this entity and gains a stake valued at over a $100,000,000,000 according to reports. So this matters because the new PBC public benefits, corporation structure could unlock large new funding rounds for OpenAI, make an eventual IPO possible, or allow, OpenAI, the chat g p t maker to go public and still keep mission focused control in place, which could speed product development and change pricing across widely used AI tools. So the agreement comes in the form of an m MOU or a memorandum of understanding, and it's not final. And both sides says they are working on a definitive contract and need approvals from both California and Delaware attorneys general before any transition takes effect. So Brett Taylor, who's OpenAI's board chair, said the nonprofit will continue to control operations and also hold the PBC stake, but other terms were not yet disclosed.

Jordan Wilson [00:33:31]:
So like I said previously, Microsoft had reportedly invested, between 13 and $14,000,000,000 for a 49% claim in OpenAI. So this is a big shift here. So Microsoft, under this new agreement, would remain a preferred tech partner and primarily cloud a primary cloud provider under the current deal, but OpenAI is reducing reliance by, again, reportedly committing $300,000,000,000 to Oracle like we just talked about and also partnering with SoftBank on this Stargate data center. So months of tense talked included. So, yeah, this has been going on for many months. We've been reporting on it. These new partnership talks between OpenAI and Microsoft, and they, included disputes over rights to tech from Windsurf, an AI coding startup that OpenAI sought to buy, but the acquisition kind of collapsed because of, this kind of, agreement in the current, partnership between Microsoft and OpenAI. And then, Windsurf essentially was aqua hired by Google, and then much of the remaining team moved to Cognition.

Jordan Wilson [00:34:40]:
So, reportedly, this, the current partnership had also been a problem for OpenAI, in, acquiring other large AI companies that may not fit under the current agreement or that may be a direct competitor to Microsoft. So OpenAI's nonprofit over profit governance, which enabled the 2023 firing and the quick rehiring of CEO Sam Altman remains intact and would carry over into this new structure. So this is big because if regulators approve and contracts are finalized, like I said, you could expect faster capital raises, a clear path for OpenAI to go public. Right? So being able to, you know, go in your April and buy OpenAI shares, but also this would allow OpenAI to probably have faster acquisitions, right, which could ultimately change the product drastically that 700,000,000 people are using on a weekly basis. Alright. So a little bit more on the revenue side. So, technically, a separate story here but related. So according to the information, OpenAI projects that its revenue share, with Microsoft will fall from nearly 20% in 2025 to only about eight percent by the end of the decade, a shift that could reshape the economics of OpenAI.

Jordan Wilson [00:36:09]:
So this new report says this change could let OpenAI retain more than $50,000,000,000 of additional revenue that they did not, project through 2030, giving it critical capital to cover massive training and inference costs as their usage scales. So, yeah, reportedly, Microsoft had about 20 percent of a revenue share, and this is kind of outside, but we don't know if it's exactly outside or in addition to, that 49% equity stake. But according to this new report from the information, the rev share side would fall from 20% to 8% by the end of the decade. So a pretty big drop off there. So but that would mean, OpenAI having a greater ability to scale and ramp up, usage as well. So the companies, like I said, are finalizing that new partnership, terms following the, new nonbinding MOU that enabled OpenAI's corporate restructuring. So Microsoft, though, still gains strategically with a significant equity stake moving forward, and integration of OpenAI's models across Copilot and the right of first refusal, sorry. The right of first refusal on OpenAI's computes, keeping workloads anchored to Azure if they wish.

Jordan Wilson [00:37:32]:
Alright. Sheesh. That was a lot. Yeah. Doug was saying, is it hot take Tuesday already? Not yet. Alright. But there's more. Let's go over quick rapid fire here of what's new and what's next.

Jordan Wilson [00:37:53]:
Alright. So these are some kind of news stories, some rumors, some new features, but there was a ton of smaller updates that could end up being big for you or your business. So we're gonna go over them very quickly. Here we go. Bullet point. Let's go. So Claude rolled out memory finally, but only for enterprise and team plans. Claude also rolled out private chats for all users.

Jordan Wilson [00:38:21]:
Perplexity, unveiled the free access to Perplexity for the US government. Meta signed a $140,000,000 licensing deal with Black Forest Labs for AI images after a similar partnership with Midjourney. So Meta going all in on third party AI image models. Google's v o three can now generate vertical videos, and that's actually important. So keep an eye on social media reels being flooded with Google v o three videos if they weren't already, because now v o three can generate vertical, whereas before it could only generate landscape. OpenAI reversed course and is keeping their standard voice mode after they previously said they would get rid of it in lieu of their new advanced voice mode. So standard voice mode is sticking around. Perplexity raised $200,000,000 at a $20,000,000,000 evaluation.

Jordan Wilson [00:39:09]:
Microsoft added scripted audio mode, which looks pretty cool. I played around with it to its co, Copilot audio expressions. AI agent startup, Genspark, released an updated AI browser with on device AI, which looks really good. And it might give, Chromes or sorry, not Chromes, Chromium based Comet from Perplexity a run for their money on a top, agentic browser. Replit introduced agent three, which looks extremely impressive, an agent that can run for up to two hours at a time. Thinking Machine Labs, run by former OpenAI exec, Mira Murad launched their connectionism's blog and shared code to make large language model inference more deterministic and less, generative. OpenAI, yeah, we got a lot. OpenAI is testing new checkout features for its depth desktop application where they'll eventually offer shopping straight within chat g b t.

Jordan Wilson [00:40:12]:
AI agent startup Manus added connectors like Gmail, Google Calendar, Notion, Zapier, and others to its product. ChatGPT may soon roll out parental controls, which I think will be, sorely needed. Google rolled out a new feature inside Canvas called select and ask, making editing much easier inside Canvas. You can just select an, an element in Canvas and talk to Gemini and have it change it. Chat GPT, earlier in the week announced branching of conversations. I think a lot of people are going nuts over this. I'm not really gonna use it a whole lot because there's some problems with carrying over context that I think people aren't really understanding, but that's fine. Notebook l m added custom reports and new audio overview, options, which we covered late last week in episode six zero eight.

Jordan Wilson [00:40:59]:
Make sure to go check it out. It's actually really, really good. The TikTok parent company ByteDance released AI image model seed RAME four point o, which is really good. A lot of people are saying it's better than Google's nano banana. I don't think it is. Benchmarks don't say it is, but it is by far, number two. So huge new player on the scene. If you haven't seen c dream four point o yet, from ByteDance, pretty big.

Jordan Wilson [00:41:26]:
Speaking of Google's nano banana, Adobe has already rolled it out, in Photoshop. So pretty big news there if your company is heavy Photoshop users. Google is rolling out ads in its AI mode, and, Albania Albania became the first country to endpoint an AI bot as an official government minister. Wild. So much happening this week in AI news. I hope this was helpful. If it was, don't be greedy. Go tell someone about this.

Jordan Wilson [00:41:56]:
If this was helpful, you know, a lot of people are like, hey, Jordan. This is this is my cheat code. You know, listening to everyday AI and having you do all my work for me, it's great. What can I do to help you? Tell someone about it. Alright. If you're listening to the podcast, appreciate it if you click that follow or subscribe button on, Spotify or Apple Music. If you could, leave us a rating. But, also, if you're listening on the live stream, click that little repost button.

Jordan Wilson [00:42:21]:
Share this with your network. We try to do this every single day. I understand how hard it is to keep up with everything in the world of AI, and AI, if it hasn't already, is going to impact all of us and how we work. So join us on Mondays. If you don't really care about anything else, if you just wanna know what's new, what's next, that's what we do for you. Tuesday's kind of our hot take Tuesday, coming with a a a hot take this week, so make sure you go check out our newsletter. I'm gonna put out a poll, let you guys decide what's gonna be on our show tomorrow. So another reason you need to go to your everydayai,uh,.com and sign up for the free daily newsletter if you haven't already.

Jordan Wilson [00:42:58]:
So thank you all for tuning in. Hope to see you back tomorrow and every day for more everyday AI. Thanks, y'all.