Think about the assumptions we make every single day. Beat, we spent decades thinking AI was just a chat bot. Yeah, just a simple digital box. Right, you type a prompt, you get text back. But what happens when that chat bot finally stands up? It starts acting like a digital employee. Exactly. or when it autonomously solves 80 -year -old math problems. These are puzzles that baffled the absolute brightest human geniuses. It completely changes our relationship with this technology.
I mean, the ground is shifting right beneath our feet. Welcome to this deep dive for you today. We are looking at a stack of incredibly fresh sources. Really exciting stuff. They map out the next era of artificial intelligence. We are going to trace the massive hardware shift happening right now. We'll explore the mind bending billions trading hands between top labs. We'll examine the tools turning AI into proactive workers. And finally, we'll look at a historic breakthrough
in autonomous reasoning. It is a fascinating blueprint of what comes next. So to truly understand how AI is getting smarter. We have to start physically. Right at the foundation. Yeah. We have to look at the underlying hardware. And we really need to follow the money moving behind it. Right. And the numbers here are absolutely staggering. NVIDIA just posted another record -breaking financial quarter. What are we looking at? They reported $81 .6 billion in revenue. That is up 20 % from
the previous quarter. That is a massive leap in just... Three months. Yeah, it really is. Their data center business alone hit $75 .2 billion. Wow. That shows infrastructure demand is still incredibly strong. The physical foundation is still being rapidly built. I want to pause on the financial flexing here. They approved $80 billion in share buybacks. What does that signal to the broader tech market? Well, it signals massive internal confidence in their future growth.
They are telling the market they have plenty of cash. Right. But what's even more interesting is their venture capital strategy. They hold $43 billion in private startup stakes right now. Exactly. They bought $18 .5 billion this quarter alone. That is wild. They aren't just selling chips to random companies anymore. They are actively funding the exact startups that buy their hardware. It creates this massive self -sustaining financial flywheel. Yeah, it does. That's brilliant, but
it can't accelerate forever, right? No, probably not. Forecasts predict $91 billion for the next quarter, but their overall growth rate may slow to 12%. Plus, their China export revenue remains highly uncertain right now. That brings us to the biggest pivot in their strategy. Jensen Huang just announced a new $200 billion market. This is huge. It centers entirely around a new CPU called Vera. This is a fundamental shift in computing architecture. Vera is built specifically for
agentic AI workflows. Let's define that term for everyone listening. Agentic AI is AI that takes actions independently to achieve a goal. Perfect definition. Beat. Think of this shift like a busy corporate office. The GPUs are the creative brainstormers in the room. Right. They generate huge ideas and process heavy data. Exactly. But CPUs like Vera are the actual project managers. They're the ones executing the complex daily tasks. That analogy is spot on. NVIDIA isn't
just selling GPUs to researchers anymore. They want to power the full AI stack entirely. From massive data centers down to proactive autonomous agents. Huang says Vera has already generated $20 billion in sales. That is just in this current year alone. It proves the market demand. for agentic hardware is real. Future AI agents will deeply need these CPU -like systems. Why is that? Because they need them to use external tools efficiently. They have to run sequential tasks
without constant human intervention. How does NVIDIA stepping into CPUs affect the rest of the industry? Well, they aren't just selling individual processing parts anymore. They want to own the entire operational architecture. Right. They are building the infrastructure for autonomous digital workers. So they own the brains and the hands now. Exactly. They are the foundation of the new economy. But if they are supplying the brains and the hands the energy required to power,
that must be astronomical. Oh, it is. Who is actually footing that massive computing bill? The operational costs are becoming almost incomprehensible today. Let's look at Excel. for a very clear example. Okay. They burned $6 .4 billion last year. That was on just $3 .2 billion in revenue. The math there is incredibly brutal to look at. It is. And a recent SpaceX IPO filing revealed something even wilder. What was it? It detailed Grok's plan for a multiple trillion parameter
model. A trillion parameter model has a trillion adjustable internal data connections. Yeah. That ambitious plan will make the compute bill substantially bigger. But here is the wildest twist in our sources. I know what you're going to say. Anthropic is reportedly paying XAI directly for compute power. They are paying $1 .5 billion to Sex Silence per month. Yeah. That is for access to Colossus Compute through May 2029. Whoa. Imagine scaling to a billion queries. The power required to run
an entire data center just for one model. That is the reality of an industrial revolution scale. You are paying for miles of servers and gigawatts of power. It is a staggeringly expensive physical footprint. Yet despite these massive infrastructure costs, Anthropic expects its first profitable quarter very soon. Right. Their revenue is... jumping to around $10 .9 billion. That is a massive milestone for the entire tech industry. It proves the underlying business model can actually work.
It dropped right alongside rumors of an open AI IPO. The market is maturing incredibly fast right now. Extremely fast. How does a company possibly survive burning billions a year? It's a massive infrastructure land grab right now. The sheer utility of the resulting models is the price. Right. That utility will eventually outweigh the massive upfront cost. We see this clearly with Anthropic finally hitting profitability. We must spend billions today to make trillions
tomorrow. Exactly. The initial capital expenditure is the only way forward. So what incredible utility justifies paying a billion dollars monthly? It is the fundamental shift from generating text to taking action. We are looking at the toolkit for these new digital workers. The ecosystem of agentic tools is exploding right now. And it completely changes how we interact with software. EXA just raised $250 million. That is at a $2 .2 billion valuation. Yeah. They are simply building
a search engine for AI agents. This is a crucial piece of the puzzle. Agents don't browse websites like human beings do. Right, they can't see the screen. Exactly. They can't navigate flashy pop -ups or read layout designs. They need raw... perfectly structured data to operate effectively. So EGZA provides that exact structured data layer for them. Yeah. And developer demand for this is just huge. Google is also making massive updates to their tool ecosystem. Gemini Omni is generating
entire new video scenes now. It's amazing. It uses mixed inputs to create high quality outputs. You can combine images, audio, video, and text seamlessly. Yeah. It is completely changing the creative production pipeline forever. Google AI Studio is taking this even further today. It turns simple text prompts into native Android apps. The traditional barrier to software coding is rapidly disappearing. Then there's anti -gravity 2 .0 to discuss. It is turning AI into a real
developer workspace. This one is wild. They feature parallel agents and background coding tasks seamlessly. We should definitely explain what parallel agents actually do. Parallel agents are multiple AI programs working on tasks simultaneously. Beat, I still wrestle with prompt drift myself. Oh, same here. That's when AI slowly loses focus on your original instructions. So the idea of parallel agents managing whole workflows feels wild. It is a lot for human beings to adapt to.
We're moving from prompting to managing a digital staff. That is probably why Andrew Nehring released a free course. It is a two and a half hour masterclass on AI prompting. Very timely. It is designed to help humans keep up with these tools. Prompting is quickly becoming a necessary professional management skill. We also have to talk about proactive actions. Gemini Spark handles proactive actions across Google products natively. Right. It can manage tasks under your direction entirely
in the background. But having proactive agents doing things. automatically raises massive concerns. You need serious security frameworks before you deploy them. Absolutely. OpenAI is adopting Google's SynthID watermarking for generated images. They are also building a public verification checker very soon. We really need to know what's real and what's AI. We are seeing movement on the regulatory front, too. The European Commission
released draft guidance on high -risk AI. It clarifies exactly which systems fall under stricter rules. Now we have impending U .S. executive orders coming soon. OpenAI and Anthropic are discussing voluntary frameworks as well. They are talking about model sharing directly with the government. It's all moving so fast. So how do we prevent chaos if agents are acting proactively? We have to rely on strict verification systems. The strict necessity of Google SynthEye is a
vital start. Yeah. We also need those incoming EU and US regulatory frameworks. Trust requires built -in digital watermarks and strict guardrails. Right. And without those guardrails, the... enterprise utility drops to zero. We've talked extensively about the physical hardware and the money. We explored the powerful new agentic tools being built. But the ultimate test of this era is pure reasoning. Can these machine learning models actually think deeply? This is the most profound
question in the field today. OpenAI just announced a truly historic mathematical breakthrough. GPT -5 .5 autonomously created a completely original geometric proof. It's incredible. It disproves a famous geometry conjecture by Paul Erdos. He proposed this specific riddle back in 1946. This mathematical problem stood unsolved for nearly 80 years. Mathematicians strongly believed the best solution would involve square grids. But the AI model found an entirely new geom... construction.
Right. Crucially, this wasn't a specialized math tool at all. It was a general purpose reasoning model exploring the logic. It worked through the complex logical steps completely on its own. Wait, hold on. I have to push back a little bit. Okay, go ahead. Didn't this exact same thing happen late last year? OpenAI faced intense public skepticism back then. They did. People found the models were just surfacing existing human work. They weren't actually solving brand new
geometric problems. Why should we believe this time is any different? Yeah, and that skepticism was completely justified back then. But this is exactly why this specific moment matters. How so? OpenAI came incredibly prepared for the public pushback this time. They brought the receipts? They absolutely did. They didn't just drop a press release. They subjected the model's output. to rigorous human peer review. They included supporting remarks from highly respected mathematicians
today. People like Noga, Alon, and Melanie Wood reviewed the work. And Thomas Bloom, who runs the Erdin's Problems website itself. These experts validated that the logical steps were genuinely novel. This proves the model generated completely new mathematical logic. Wow. It didn't just memorize old data from the internet. Does this mean AI is now better at math than humans? It's not really about being quote unquote better than us. It is about AI being capable of genuine logical
discovery. It works right alongside human experts to find objective truth. AI is a peer now, not just a calculator. Exactly. It fundamentally changes how scientific discovery will happen. Sponsor. Let's bring all these incredible pieces together for you. The era of the simple chatbot is truly over today. From NVIDIA's Verichips to billion -dollar compute infrastructure deals, the entire tech world is fundamentally re -architecting
for proactive agents. We are arming general -purpose reasoning engines with massive hardware capabilities. They are smart enough to solve 80 -year -old math puzzles autonomously, and we are giving them tools to operate natively inside workflows. Right. They are quickly becoming our digital colleagues and active project managers. It is a profound shift in how humans interact with machines. I want to leave you with a final provocative
thought today. If a general purpose AI model can look at an 80 -year -old geometry problem and find a completely new angle that human geniuses missed for decades, what longstanding assumptions in your own industry is an AI agent about to completely rewrite? It is a thrilling, slightly terrifying question to consider. Thank you for taking this deep dive with us today. ODTRO Music.