"The Age of AI Agentics Is Here" - Nvidia CEO
17:04 Share

On today's AI Daily Brief, NVIDIA's CEO declares the age of AI agentics is here at CES. And before that, in the headlines, Anthropic is raising new money at a $60 billion valuation. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes.

Welcome back to the AI Daily Brief Headlines Edition, all the daily AI news you need in around five minutes. To the surprise of no one, the trend of mega financing rounds for AI companies appears to be continuing into 2025, with the latest round coming together seemingly for Anthropic.

According to people familiar with the investment round, say that the company is talking about raising about $2 billion at a $60 billion valuation. The round appears to be being led by Lightspeed Ventures and represents a big jump in its valuation. The company's last price round was at $18 billion and was led by Menlo Ventures. Basically, ever since OpenAI raised its mega round in October that valued it at $157 billion, it's been open season for fundraising for these huge foundation model firms.

Originally, we had seen numbers in the $40 to $50 billion range for Anthropic, but then Elon's XAI raised it around that valuation. And I kind of assumed that just based on market forces, Anthropic might get a bump from that.

Sure enough, that seems to be the case. Anthropix revenue is definitely behind OpenAI's. OpenAI is looking at something like $3.7 billion, while sources suggest that Anthropix annualized revenue is around $875 million. One thing that is notable, however, is that Anthropix has significantly increased its market share when it comes to enterprise buying. It definitely appears to be scooping some of that business away from OpenAI, and so perhaps investors are betting on the trajectory rather than the current state.

Of course, the biggest difference between Anthropic and OpenAI is not just their enterprise sales, but the fact that ChatGPT is at this point largely synonymous with AI for consumers, and that shows up in their revenue numbers as well. A lot of the discourse was about what Anthropic is going to use this money for. Chubby writes, I mean, nice and all, but it's about time for Anthropic to release a reasoning model in Opus 3.5 or 4. Sonnet 3.5 is good, no doubt, but OpenAI seems to have a good advantage now.

Think broadly, there's a sense that maybe the next shoe to drop in terms of model updates will be Anthropic, but of course, we'll have to wait and see. Some folks think that this valuation actually makes OpenAI look like an even better investment. Dolly Bolly writes, Anthropic being valued at one third of OpenAI makes OpenAI look like a long. Ultimately, I still think we're in a situation where there are an extremely small number of companies that are legitimate contenders to get to AGI and beyond, and the valuations for that are fairly uncapped because of it.

Next up, another one of the big themes for 2025 is, of course, going to be infrastructure build-out, and that has a significantly geopolitical dimension as well.

One of the big questions is how comfortable the U.S. government is going to be with U.S. companies investing in and partnering in the Middle East. The Gulf states are quite literally between the U.S. and China, having strong trade relationships with both sides. This has been a point of contention, flaring up in places like Microsoft's minority investment in G42. And so it was interesting to see that President-elect Donald Trump has announced the deal going the other way with a Gulf firm spending billions of dollars in the U.S.,

In a press conference at Mar-a-Lago yesterday, property development mogul Hassan Sajwani said, It's been amazing news for me and my family when Trump was elected in November. We've been waiting four years to increase our investment in the U.S. to very large amounts of money.

According to the announcement, Sajwani is planning to invest $20 billion to build new data centers across the U.S., and Trump said that he was actually pledging at least that amount, saying, quote, they may go double or even somewhat more than double that amount of money. The first phase of the multistage investment will focus on building facilities in Arizona, Illinois, Indiana, Louisiana, Michigan, Ohio, Oklahoma, and Texas.

Now, for a sense of scale, Microsoft intends to spend about $80 billion on AI data centers this year, with half of that being deployed in the U.S. This deal seems to be somewhat emblematic of the way Trump intends to approach AI infrastructure policy. One of the ideas that has percolated from Trump's social media account was a proclamation that anyone investing a billion dollars in the U.S. would be fast-tracked through environmental regulation and permitting. The goal seems to be to put private sector investors in a position to replace and exceed government subsidies through things like the CHIPS Act.

During a recent interview, Sam Altman discussed why this plan could make sense, commenting, There's a real opportunity for the Trump administration to do something much better than the CHIPS Act as a follow-on. I don't think the CHIPS Act has been as effective as any of us hoped. The thing I really deeply agree with President Trump on is, it is wild how difficult it has become to build things in the United States. Power plants, data centers, any of that kind of stuff. I understand how bureaucratic cruft builds up, but it's not helpful to the country in general.

This investment from the Emirati billionaire also echoes calls from Microsoft President Brad Smith. In a blog post published last week, Smith said, the most important U.S. public policy priority should be to ensure that the U.S. private sector can continue to advance with the wind at its back. He also called for a focus on exporting American AI to allies and friends, particularly through partnerships in the Gulf states.

Speaking of infrastructure build-out, two more quick stories before we get out of here. Microsoft intends to invest $3 billion in AI and cloud infrastructure in India over the next two years. CEO Satya Nadella announced the plans at an event in Bengaluru on Tuesday. He said the company will also provide AI training to 10 million people in India. Nadella said, The investments in infrastructure and skilling we are announcing today reaffirm our commitment to make India AI first and will help ensure people and organizations across the country benefit broadly. The diffusion rate of AI in India is exciting.

Amazon Web Services, meanwhile, says it plans to invest at least $11 billion to expand infrastructure in Georgia. In a press release, the company wrote, AWS is proud to expand our operations in Georgia to help drive the next generation of cutting-edge technologies such as AI. The announcement comes just eight months after the company committed $11 billion to building out data centers in Indiana. Interestingly, Georgia is quietly going through an AI infrastructure boom. According to a report from GovTech, data center construction grew by 76% last year in the metro Atlanta area.

I expect that we will hear lots and lots of this type of infrastructure build-out story this year. But for now, that is going to do it for today's AI Daily Brief Headlines Edition. Next up, the main episode. Today's episode is brought to you by Vanta. Whether you're starting or scaling your company's security program, demonstrating top-notch security practices and establishing trust is more important than ever.

Vanta automates compliance for ISO 27001, SOC 2, GDPR, and leading AI frameworks like ISO 42001 and NIST AI risk management framework, saving you time and money while helping you build customer trust. Plus, you can streamline security reviews by automating questionnaires and demonstrating your security posture with a customer-facing trust center all powered by Vanta AI.

Over 8,000 global companies like Langchain, Leela AI, and Factory AI use Vanta to demonstrate AI trust and prove security in real time. Learn more at vanta.com slash nlw. That's vanta.com slash nlw.

If there is one thing that's clear about AI in 2025, it's that the agents are coming. Vertical agents by industry, horizontal agent platforms, agents per function. If you are running a large enterprise, you will be experimenting with agents next year. And given how new this is, all of us are going to be back in pilot mode.

That's why Superintelligent is offering a new product for the beginning of this year. It's an agent readiness and opportunity audit. Over the course of a couple quick weeks, we dig in with your team to understand what type of agents make sense for you to test, what type of infrastructure support you need to be ready, and to ultimately come away with a set of actionable recommendations that get you prepared to figure out how agents can transform your business.

If you are interested in the agent readiness and opportunity audit, reach out directly to me, nlw at bsuper.ai. Put the word agent in the subject line so I know what you're talking about. And let's have you be a leader in the most dynamic part of the AI market. Welcome back to the AI Daily Brief. This week is the great big CES show over in Las Vegas. It's a chance for particularly consumer tech companies to show off their latest innovations as well as where they're headed.

It's kind of a weird event in the sense that a huge number of people go every year, and there are always some interesting announcements, but it's never really the home base of some world-changing announcement. I don't know why, but it's just not the place that people choose to really make their biggest announcements of the year. Perhaps it's just timing. January is kind of a weird time. Still, this year, the big keynote was from NVIDIA's Jensen Huang, and there was so much announced, it is going to be difficult to get through in a single show.

Still, we are going to try. The TL;DR is that NVIDIA is attacking basically every vertical in AI, including robotics, autonomous driving, agents, consumer devices, and much more.

The announcements were in fact so wide-ranging that at least one publication has called this NVIDIA's Apple moment. Right at the core of Jensen's presentation was this declaration that the age of AI agentics is here. For those of you who are not watching, I've got an image up of this great slide that he had showing an up-and-to-the-right graph that goes through perception AI, degenerative AI, to agentic AI, to physical AI.

Under agentic AI, he references coding assistance, customer service, and patient care. Basically, Jensen described the AI sector as heading into a steep exponential curve as agents come online and provide a stepping stone towards a fully robotic world. Over the shorter term, he said, quote, AI agents are the new workforce. The IT department of every company is going to be the HR department of AI agents in the future.

Jensen also mentioned that NVIDIA's AI chips are improving, quote, way faster than Moore's Law. He said that the current model of chips is 30 times faster than the previous generation, adding, we can build the architecture, the chip, the system, the libraries, and the algorithms all at the same time. If you can do that, then you can move faster than Moore's Law because you can innovate across the entire stack. Explaining why this is so crucial for an AI industry struggling to hit profitability, he said, Moore's Law was so important in the history of computing because it drove down computing costs.

The same thing is going to happen with inference where we drive up the performance and as a result, the cost of inference is going to be less. You might remember our discussion yesterday around how Sam Altman was saying that OpenAI's $200 a month subscription is actually not profitable. The presentation itself got rave reviews.

Salesforce CEO Mark Benioff breathlessly posted, Rarely am I so captivated and inspired by a CEO's keynote that it holds my attention for two hours. But Jensen Huang's indefatigable energy makes him one of the most inspiring and visionary leaders among tenured and authentic CEOs.

His ability to innovate while delivering amazing products I genuinely want to buy is remarkable. A true masterclass in running a great tech company. I was particularly struck by his framing of AI's evolution, from perception to generative to agentic and ultimately to physical. Truly inspiring. Still no less breathtaking was the gigantic range of products and services that NVIDIA announced. And so even though I've just said that in general, CES isn't the place for big announcements, this was truly a full-scale product launch.

Now for our purposes, and given the theme of the presentation, I think the most important products was NVIDIA's first full family of agent models. Called NEMOTRON, these models come in three sizes to cover use cases from edge device inference to tasks that require frontier models. Last year, NVIDIA gave a sneak preview of a 70B version of NEMOTRON, which outperformed the same class of models from OpenAI and Anthropic on agentic tasks. The new family of models will also come in two different varieties.

a text-based model for language tasks, as well as a vision-based model optimized for physical AI projects. The models are impressively fully open sourced, with the text models based on a fine tune of Meta's Lama models. This means that NVIDIA will be competing on price with models that cover the full spectrum of agentic use cases. And as a total aside, we could have a very long conversation here around the commoditization of AI models that feels like it's only going to accelerate in the agentic age. Alongside the model family, NVIDIA also released a set of blueprints for AI orchestration.

These are designed to guide agents through specific tasks and can aid in coordination of multiple agents. NVIDIA has partnered with several AI orchestration companies to build blueprints that will be available on NVIDIA's enterprise platform, including Langchain, Llama Index, and Daily. One example created by Crew.ai handles code documentation to ensure repositories are easy to navigate. Another example is a PDF-to-podcast blueprint that produces results similar to Google's Notebook LM.

Rev Liberadian, VP of Omniverse and Simulation Technology at NVIDIA said, That's a lot of words, but basically the goal here is to make agents actually work and be as quick to deploy as possible.

One of the more impressive blueprints shared allows agents to analyze video. A complex pipeline of integrated services can deliver high-quality video analysis 30 times faster than watching it in real time. This has a huge range of extremely valuable applications, both in updating systems with AI and devising new processes with this novel ability. The low-hanging fruit is monitoring of industrial processes. There's currently more than 1.5 billion enterprise cameras deployed worldwide that record 7 trillion hours of video per year. Only a fraction of that actually gets analyzed.

At the same time, manufacturers lose trillions of dollars annually to product defects that could have been spotted earlier using an AI-enhanced camera. This is just one specific example, but one where you can see the immediate tangible ROI right away.

Another marquee announcement from NVIDIA was a small footprint AI workstation called Project Digits. Around the same profile as a Mac Mini or the size of a hardback book, the device contains the company's Grace Blackwell hardware. Huang said the device, quote, runs the entire NVIDIA stack. All of NVIDIA software run on this. It's a cloud computing platform that sits on your desk. It's even a workstation if you'd like it to be. The device can be used as a standalone computer running NVIDIA's Linux-based OS or plugged into a primary Windows or Mac PC.

It can put out a petaflop of AI compute for prototyping, fine-tuning, and running AI models. NVIDIA claims a single unit can run inference for models up to 200 billion parameters. Not quite enough to run the largest version of Meta's Lama model, but plenty of power to run mid-sized modern LLMs locally. A pair of devices can also be networked to deliver inference for models with up to 405 billion parameters. The device will be available in May at a price point of $3,000, which while making it a prosumer product, certainly doesn't make it inaccessible for people.

Huang said, placing an AI supercomputer on the desks of every data scientist, AI researcher, and student empowers them to engage and shape the age of AI.

NVIDIA also launched a range of AI gaming tools. There was an automated producer agent that can manage live Twitch streams, as well as AI-driven in-game bots. On the automotive side, NVIDIA has revealed new hardware and software platforms for autonomous driving, along with a giant list of partnerships. Companies from Toyota to Uber will build up their self-driving capabilities using NVIDIA's technology, which has now been certified by tough European regulators. NVIDIA is also partnering with Aurora and Continental to deploy driverless trucks at scale by 2027.

There was also a vision-enabled desktop assistant complete with a very impressive rendered avatar with voice mode. And that's all without mentioning the new range of consumer GPUs built on Blackwell architecture and a new data center chip with 72 Blackwell processors on the same wafer. The super chip was really something to see. Jensen kind of looked like Captain America carrying this thing across the stage like a shield. Now, in terms of response, the Neuron newsletter referred to this as NVIDIA's Apple moment, commenting, "'Remember when putting a personal computer on every desk and in every home changed the world?'

NVIDIA is more or less trying to do the same thing with AI supercomputers. Now to me what seems clear, and this has been coming for a while, is that NVIDIA is clearly running the playbook similar to Google from the early 2010s, where they shift focus from their one thing to try to capture as many related verticals as possible.

If NVIDIA can deliver on this bulky roadmap, it will have a competitive platform available for basically every use case of AI development, from language agents to industrial robots. They're also leaning heavily into open source, which will likely drive a massive price war among AI labs. All of this, of course, drives NVIDIA's core business of selling AI chips. The TLDR on their strategy? Make deploying AI in every aspect of life as friction-free as possible to drive demand for AI chips through the roof.

Now, funny enough, the implications kind of seemed lost on Wall Street. Bloomberg reported the event under the headline, NVIDIA's slides after unveiling leaves investors wanting more. I actually think that this is a total misunderstanding of what happened. NVIDIA's stock was down 6% yesterday, but I think that has much more to do with unrelated macroeconomic shocks than it does to do with investors not liking what Jensen had to say in Vegas. Indeed, stock analysts were at least a little less myopic, with

with Goldman Sachs writing that the quote, string of announcements at a minimum highlight the company's ability to innovate at industry-leading speed across hardware and software, as well as its robust partner and customer ecosystem. Ultimately, what this was, was an unveiling. NVIDIA has been inching away from being just a chip company to being an AI-first tech giant, and that couldn't have been clearer in this presentation. We'll wrap there. However, that's going to do it for today's AI Daily Brief. Appreciate you listening or watching as always. And until next time, peace.