Nvidia and OpenAI Up the AI Stakes with $100B Deal

0:00Today on the AI Daily Brief, a new$100 billion 10 gigawatt deal between OpenAI and NVIDIA. Before then, in the headlines, the agentic era has officially begun at Citigroup. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

0:23Hello, friends. Quick notes before we dive in. First of all, thank you to today's sponsors, Notion, Superintelligent, Agency.org, and Robots and Pencils. And to get an ad-free version of the show, go to patreon.com slash ai daily brief. Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. We kick off today with a bit of a case study, this one from Citigroup via the Wall Street Journal, and it is very clear that for Citigroup, the agentic era has arrived. The bank has been steadily working on their proprietary AI platform over the past two years, and this month, the agentic capabilities of that system will go live.

0:58Users will be able to direct agents to complete tasks that access multiple systems within the company, with CTO David Griffiths saying, A couple of years ago, you could do agentic things with the early versions of the models that were available then, but they weren't always very reliable. They weren't always very good at invoking tools. But they are now. Griffiths explained that the new system is capable of things like researching a specific client, building a profile of them based on public information and internal data sets, and then translating a report into foreign languages all from a single prompt.

1:28The previous system was capable of each of these functions individually.

1:41In other words, the difference here is about autonomy. How many touchpoints do you require a human for versus just setting an agent on a task and letting it go figure out how to accomplish that task? The initial pilot of the agentic system will roll out to 5 ,000 Citigroup employees over a four to six week period. In other words, 5 ,000 Citi employees are about to get agent employees. Citigroup says that they will keep a close eye on the cost-to-value ratio of the new functionality. That said, most of the tasks that are being enabled at this stage are not super complex, long-running tasks. Most of the functions will be completed in only a few minutes.

2:17So for the moment, at least, the company isn't all that worried about cost. They also are struggling to estimate costs just given how fast things are changing. Griffiths said the company had baked hard cost limits into the systems, but since the cost of model usage is coming down so quickly, it's very hard to come up with up-to-date and accurate return on investment calculations. If and as Citigroup's agents become capable of running for longer, hours-duration-type tasks, cost is expected to be more of a factor. When asked whether at that point the agents would start impacting the workforce, Griffiths gave a surprisingly candid assessment.

2:48Does it mean we need less people? I don't know. It certainly means that we would get a lot more done, and we'll see how the workforce evolves with that massive boost of capacity that we're getting here. And this is, of course, the question. Does a company choose to reinvest the savings and efficiency from agents into cost-cutting, i.e. removing people, or into growth, i.e. trying to accomplish new things that weren't possible before? Obviously, those of you listening know which one I think is the better bet in the long term. Now, speaking of enterprise AI, one of the things that's become abundantly clear over the last few months is how hard the last mile implementation of AI systems really is.

3:25Outside of the simple assistance that you can hand over to your people and hope they do well with, AI systems, particularly agentic systems, aren't the type of thing that you just install and let go. They involve a lot of customization, connecting to particular data sources, feeding in information about SOPs and workflows, integrating agentic flows with existing human workflows. There's just a lot of work at the point of actual work. Luckily for enterprises, there are many companies who are trying to help with some aspect of that, and one of them is the latest AI unicorn. Distill AI has announced a Monster Series B, raising$175 billion at a$1.8 billion valuation.

4:01The company, which was founded by former Palantir employees, has scaled up dramatically over the past year. Their last round was raised in November, raising$20 million at$200 million, so this is a 9x valuation jump inside of 12 months. The company aims to help Fortune 500s become AI native. Co-founder and CEO Arjun Prakash said, The companies that win in the AI era are those that are willing to reimagine how they operate, not just what tools they use. AI is forcing enterprises to move beyond silos. Our belief is that the next chapter of AI leadership won't be won by models alone, but by operationalizing AI at scale in enterprises and transforming how they work.

4:35You would not believe, my friends, the number of times that I've had conversations with Silicon Valley's most vaunted investors who sit perched in their sandhill towers, who have said things with a straight face. Like, well, don't you think enterprises are just going to figure it out for themselves? Sure, this is an important problem for the next couple of years, but won't they just get it then? Let's just say I'm glad to see indications that some of that thinking may be going away as the very complex reality of enterprise AI deployments comes clearer into focus. Moving now to the tool slash feature launch section of the headlines, Perplexity has launched an email assistant.

5:11The announcement says that the assistant drafts replies, organizes messages, schedules meetings, and more. The feature is available as a plugin for Gmail and Outlook for all Perplexity Mac subscribers, meaning that Perplexity is keeping this feature exclusive to their$200 a month subscription for now. Now, email assistants have been one of the most common use cases getting deployed. They're starting to be added as a built-in feature for both Microsoft and Google, and premium standalone versions of that tool remain popular as well. Certainly, the premium end of the market is where Perplexity is trying to play.

5:38If you read their announcement post, they're basically treating email as a small proxy for a larger personal assistant play. The company writes, Email is more than a message center. Your inbox contains your professional memory, your relationships, calendaring, and coordination. Meanwhile, your outbox is your productivity and your reputation. The most successful workers and thinkers hire personal assistants for their email instead of relying on AI tools and algorithms that handle rote tasks. There's too much at stake. Perplexity's email assistant is more than a tool and available to anyone. We've talked a lot recently about context engineering and context orchestration.

6:11And basically, Perplexity is saying is that the place that context lives for most people is in their email. One tweet that I thought was interesting about this came from Dmitry Shevelenko, Perplexity's chief business officer. He wrote, The product we are launching today, which at first glance may seem pedestrian, has generated a remarkable volume of woe moments the last few weeks of hourly use. So much alpha in elegant augmentation of core workflows. Now, of course, he is the CBO of this company. He hasn't intended to get people excited. But frankly, it was that tweet which captured my attention before I even saw the announcement.

6:43And that, plus perplexity's argument that email is the home of context, has me paying attention perhaps a little bit more than I otherwise would have. Another big theme we've recently been exploring is evaluations and benchmarks. And it turns out OpenAI is hiring for an applied evals team. The job ad reads, We're hiring product-minded engineers to design and build evals and harnesses that capture real-world quality for advanced AI systems. This role sits at the center of model improvement. The systems you design will directly shape how models behave, accelerate their reliability, and raise the standard for what customers expect.

7:15Now, like I said, evals have been a big emerging theme over recent months, and they are no longer just about measuring how well a model performs, but are also playing an active role in reinforcement learning post-training. If you can define a particular goal and design a reward function around it using an eval, RL training methods can develop reliable model behavior to achieve that goal. Earlier in the month, Shyamal Anakat, who is building the team, posted, Our goal is to build the world's best evals for the economically valuable tasks our customers care about most. Broadly, the art of evals is figuring out how to ask the right questions.

7:45But asking the right questions is really hard. In any domain, it takes real deep-lived experience to define what truly good looks like. This work is super important, and we want to scale it. One more quick note on OpenAI. The rumor mill continues to swirl about those compute-intensive features coming soon, according to Altman. And while Sora 2 is still a leading candidate, Rasser X captured an updated guess, writing, OpenAI may soon unleash its internal reasoning model, the one that hit golden math and coding purely through reasoning no tools. Whispers say it will be pro-only with ultra-limited usage unless you pay for extra credits.

8:17If true, we're looking at the first taste of true general reasoning sold as a scarce resource. With this kind of exclusivity, accelerate progress or gatekeep the future? One more today before we move over to the main episode. Oracle is doubling down on AI infrastructure with a C-suite shakeup. The company announced on Monday that CEO Safra Katz would be stepping aside after more than a decade in the role. She will take a new position as the executive vice chair on the company's board. In her place, the board has appointed Clay Magyorek, whose name I'm sure I'm mispronouncing and I will learn as they move into this new role, and Mike Cecilia as co-CEOs.

8:50Clay came across from AWS in 2014 and founded Oracle's cloud engineering team. He has served as the president of Oracle's cloud infrastructure division for the duration of his tenure. Mike has held a series of different roles at the company since he joined via acquisition in 2008. Since June, he has held the position of president of Oracle's industries division, and during an investor call to announce the move, Cecilia was described as responsible for much of Oracle's software stack. In a statement, Kat said, Today, Oracle is recognized as the cloud of choice for both AI training and inferencing.

9:18I'm very proud of that. Oracle's technology and businesses have never been stronger, and our breathtaking growth rate points to an even more prosperous future. At this time of strength is the right moment to pass the CEO role to the next generation of capable executives. Those two obviously come to power at a very big moment for the company, which is at the epicenter of these swirling discussions around AI infrastructure buildouts and bubbles, which is of course the subject of our main episode, to which we will now head. Chatbots are great, but they can only take you so far. I've recently been testing Notion's new AI agents, and they are a very different type of experience.

9:52These are agents that actually complete entire workflows for you in your style. And best of all, they work in a channel that you already know and love because they are purpose-built Notion super users. Notion's new AI agents completely expands the range of what Notion can do. It can now build documents from your entire company's knowledge base, organize scattered information into organized reports, basically do tasks that used to take days, and get them complete in minutes. These agents don't just help with work, they finish it. Getting started with building on Notion is easier than ever. Notion agents are now your very own super user to help you onboard in minutes.

10:25Your AI teammates are ready to work. Try Notion AI for free at the link in our show notes. Today's episode is brought to you by Superintelligent. Now, one thing that we are having a lot of conversations with folks about is the fact that for some of you, your fiscal year is coming to an end. And that means two things. One, it means planning and thinking about what you're going to do in the next year. And two, it means using up those last of budgets so you don't lose them. If you are an enterprise that happens to find yourself in that situation, Superintelligent would love to help on both fronts.

10:56We are moving increasingly towards an annual AI planning model where we map out how you can create an action map of your organization's agent opportunities that represents an executable backlog of AI and agent use cases that you can deliver on over the course of the next year. Additionally, for those end-of-year budgets, we have worked out deals with a number of partners where we can pre-lock in general implementation packages even before you figured out exactly what use cases are going to require them. If you'd like to learn more about superintelligence agent readiness audits and this new end of fiscal year plan, visit us at bsuper.ai, click get started, and make sure to use the word fiscal somewhere in the description.

11:33Shape the future of enterprise AI with Agency, A-G-N-T-C-Y. Now an open source Linux foundation project, Agency is leading the way in establishing trusted identity and access management for the Internet of Agents, the collaboration layer that ensures AI agents can securely discover, connect, and work across any framework. With agency, your organization gains open, standardized tools and seamless integration, including robust identity management to be able to identify, authenticate, and interact across any platform. Empowering you to deploy multi-agent systems with confidence, join industry leaders like Cisco, Dell Technologies, Google Cloud, Oracle, Red Hat, and 75-plus supporting companies to set the standard for secure, scalable AI infrastructure.

12:13Is your enterprise ready for the future of Agendic AI? Visit agency.org to explore use cases now. That's agntcy.org. AI isn't a one-off project. It's a partnership that has to evolve as the technology does. Robots and pencils work side by side with clients to bring practical AI into every phase. Automation, personalization, decision support, and optimization. They prove what works through applied experimentation and build systems that amplify human potential. As an AWS-certified partner with Global Delivery Centers, Robots and Pencils combines reach with high-touch service. Where others hand off, they stay engaged.

12:50Because partnership isn't a project plan, it's a commitment. As AI advances, so will their solutions. That's long-term value. Progress starts with the right partner. Start with Robots and Pencils at robotsandpencils.com slash AI Daily Brief. Welcome back to the AI Daily Brief. This week kicked off with a bang as NVIDIA and OpenAI, two of the most important and iconic companies in this new AI era, announced a massive$100 billion deal around data and compute. Let's talk first about what was announced and then what it all means. There are a couple parts of the deal. One part of the deal is that NVIDIA will invest up to$100 billion in OpenAI.

13:30And the other is that OpenAI has signed a letter of intent to deploy at least 10 gigawatts of NVIDIA systems to train and run their next generation of models. For reference, that's somewhere in the ballpark of 4 to 5 million GPUs, which is around 25 % of the current total data center capacity of the United States. In announcing the deal, Sam Altman wrote,

14:01Now, of course, this will happen in stages. It's not like Jensen will wire$100 billion to Sam and they'll break ground on a 10 gigawatt facility or anything crazy like that. But still, the sheer size of the deal is catching attention. Here's how Sam Altman described the deal in an interview on CNBC. Jensen said, building this infrastructure is critical to everything we want to do. Without without doing this we cannot deliver the services people want We can't keep making better models And now that we really see what's on the near-term horizon of how good the models are getting the new use cases that are being enabled What people want to do this is like the fuel that we need to drive improvement to drive Drive better models to drive revenue everything so this is this is helping us get to a world along with our partners at Stargate Microsoft Oracle where we can build out increasing amounts of infrastructure to deliver on what the world is demanding out of these services.

14:51Now, I want you to pay close attention to that word demand, as it is one that we're going to come back to in just a little bit. Now, when it comes to reactions, one strand, the frankly obvious strand, was in full force, which is, of course, the hand-wringing around the circularity of AI revenue and the argument that this is all just emblematic of this big AI bubble. You had some people use a simple visual metaphor, like the outlet plugged into itself. Others have diagrams to show the flow of compute and investment to pay for that compute. Others described it in words. AI entrepreneur Sully Omar captured that Twitter virality with his version of this take, which has been around every time we see a deal like this.

15:25Sully writes, So let me get this right. Oracle says OpenAI committed$300 billion for cloud compute. Oracle's stock jumps 36%. Oracle runs on NVIDIA GPUs, has to buy billions in chips from NVIDIA. NVIDIA just announced they're investing$100 billion into OpenAI. OpenAI uses that money to pay Oracle, who pays NVIDIA, who invests in OpenAI. Some tried to connect the dots to the dot-com bubble. Cashi writes,

16:14seen this slow of demand before and it was all of a sudden moment. Some also tried to draw the connection to recent comments from Mark Zuckerberg on a podcast where he said, if we end up misspending a couple of hundred billion dollars, I think that is going to be very unfortunate, obviously. But what I'd say is I actually think the risk is higher on the other side. If you build too slowly and then superintelligence is possible in three years, but you built it out assuming it would be there in five years, then you're just out of position on what I think is going to be the most important technology that enables the most new products in innovation and value creation in history.

16:41Now, as you might imagine, that quote got lots of people saying, yeah, but didn't he say this about the metaverse? Investor Ross Hendrick writes, the top will put an end to the AI CapEx bubble, not the other way around. When did Zuck stop torching billions on the ill-fated metaverse? When the share price collapsed by 75%. No one knows when the top will occur, but that's what will put an end to this madness. And indeed, that's the way that one interpretation sees this as madness. To use the famous Greenspanian phrase, a big bucket of irrational exuberance. In an article titled, AI Investments Carry Whiff of Vicious Circles Past, Reuters wrote, Artificial intelligence is spawning some genuinely concerning financial decisions.

17:18All this mutual backscratching is reminiscent of the bubble from a generation ago. Then, as new entrants as well as more established companies borrowed more than a trillion dollars to instar telecommunications cables across ocean floors and beyond. Equipment suppliers, including Cisco, Lucent Technologies, and Nortel Technologies, enjoyed fast top-line growth and valuations. The problem was that the revenue underpinning the boom wasn't all that cracked up to be. Except the issue with that Reuters piece is that it wasn't written this week. It was written all the way back in 2023, right around the one-year anniversary of ChachiBT.

17:48I do not think that it is a problem to question the sustainability of these deals or to question the market's interpretation of events. In fact, I think it's absolutely critical to do so. However, these clever made-for-Twitter quips are not the analysis we need. All of these pithy posts simply willfully ignore the fact that unlike all those other booms, this one is anchored in real demand and real revenue. Revenue that ultimately comes back from people using the technology and adopting it in an incredibly rapid clip. Remember I told you to keep track of that word demand? That's the key thing here.

18:25OpenAI's revenue has grown from a million back in July of 23, or about a million and a half when that previous Reuters article was written, to 12 billion today. Anthropic this year alone has grown from$1 to$5 billion. And that revenue is not based on fake deals. It's based on individuals and companies deciding to open up their wallets and pay for a service that they find somewhere between incredibly useful and totally transformational. Investor Martin Bradstreet writes, This would be a circular Ponzi scheme, except we're missing the part where Nvidia is trading GPUs for OpenAI equity because they expect it to make ridiculous amounts of revenue from the rest of the global economy as it is currently doing.

19:00OpenAI has 3x the revenue of Palantir and double the growth rate. And the reality is, for the first time in some time, with this financing, I am actually seeing a bit of a counter-narrative. A vibe shift that is neither the make money during musical chairs of Wall Street, nor is the clever bubble posters on X, but is instead asking the question, what if this is real? What if we are actually just watching the beginning of a new global economy being built? Garrison Murado writes, I don't think we are giving nearly enough attention to the scale of what is happening before our eyes with OpenAI. The Kobeisi letter writes, The AI revolution is way bigger than you think.

19:35The NVIDIA OpenAI deal is colossal and its implications will be widespread. The investment in compute will flow downstream and spur incremental investment in various components of the AI revolution. Electricity is now the world's most crucial commodity. Writing in his financial newsletter on Monday, Anthony Pompliano argued that there are two major factors likely to drive stocks much higher. The first is the skyrocketing money supply, but the second was a fundamentals-based concentration on the largest U.S. tech stocks. He concluded that piece, Plenty of people are betting on the end of the party similar to the dot-com bust of 2000.

20:06But I wouldn't be so confident. Things in motion tend to stay in motion, and these large-cap companies are driving record profits and revenue growth year over year. Maybe, just maybe, these businesses are actually becoming more valuable at a rate we have never seen before. Time will tell. And I think this is the logical fallacy that we're dealing with here. Just because things are growing as fast as they are, and faster than we've seen before, does not mean that that growth is not real. Balaji Srinivasan writes, My explanation is the legacy economy is being sunset in favor of the internet economy.

20:37Since the 2008 financial crisis, every transaction and every communication has moved online, but we're still at the foot of the mountain. The next step is internet economies, communities, cities, presidencies. The world is becoming internet first. The point, again, is not to argue that we shouldn't be having concerned conversations about this. Markets are momentum machines, and even when developments are real, they can become over-exaggerated. But the critical conversation we deserve is not the one that we're having. And in the meantime, the world's biggest companies are completely convinced that there is nothing more important to be spending their time and resources on than this particular build-out.

21:11So, over-exuberant or not, the future looks like more energy, more compute, more deals, and ultimately, more AI. And that's going to do it for today's AI Daily Brief. appreciate you listening or watching as always and until next time be safe and take care of each other peace