Jensen Huang, the CEO of Nvidia (NVDA +2.34%), has been talking a lot about agentic artificial intelligence (AI) in recent weeks.
He's right to do so, as agentic AI represents the next leap in AI technology.
No matter how advanced modern AI programs might seem when you interact with them, the way they function is, in essence, identical to how an ordinary computer program does.
You input a prompt, and the AI outputs an answer to that prompt. It's functionally no different from using a word processor. You input key presses on your keyboard, and your computer outputs letters on your screen.
However, with an agentic AI program, you could give it general instructions and have it interact with the internet on your behalf.
While the technology is still in its infancy, Google's parent company, Alphabet (GOOG +2.16%)(GOOGL +2.25%), has already emerged as an early leader in it.
Image source: Getty Images.
Somewhere, beyond the sea
Subscribers to Google's $250/month AI ultra plan get access to Project Mariner, the company's experimental AI agent.
Project Mariner is fully integrated into Chrome and can interact with websites on behalf of its human supervisor through it. For example, Project Mariner can purchase your tickets to a sporting event or concert, or even buy groceries for you online.
It still can't interact with the physical world on your behalf, it can't even interact with the internet outside of one browser, but it is far closer to what many of us likely imagined when AI programs first started coming onto the scene in 2022.
There are competitors, of course, both OpenAI and Anthropic offer Operator and Computer Use, respectively. But I think it's Alphabet that has the most potential for a few reasons.
Google, Google that for me
I'll start with Alphabet's edge over OpenAI and Anthropic. Neither of them has turned a profit yet.
Now, both companies have plans to achieve profitability, and Anthropic is far closer to achieving it with a stated goal of 2027 to 2028, but neither one will rival Alphabet anytime soon.
For instance, Anthropic is projecting $70 billion in annual revenue by 2028. Alphabet generated $113.8 billion in Q4 of 2025 alone, which represented 18% growth over Q4 2024, and it managed a net profit margin of 32.81%.
Put simply, Alphabet has way more resources to throw into its AI program than either of the two most prominent companies focused on the industry.
And the meteoric rise of Google Gemini, Alphabet's answer to Anthropic's Claude and OpenAI's ChatGPT, is further proof of Google's rising dominance in the AI space.
Back in 2023, ChatGPT controlled a 50% share of the Enterprise Large Language Model (LLM) market. Meta controlled 16%, Anthropic had 12%, and Google Gemini was sitting at a paltry 8%.
Fast forward to the end of 2025, and ChatGPT's market share has fallen to 27%, and it's likely to soon be overtaken by Google Gemini, which has surged to 21% market share. Meta, meanwhile, has lost half its market share and fallen to 8% while Anthropic's Claude has grown to 40% market share.
But Alphabet is set to profit from Anthropic's rise as well.
NASDAQ: GOOG
Key Data Points
From cyberspace to reality
Unique among its AI peers, which all rely on Nvidia's graphics processing unit (GPU), Alphabet is building its own AI hardware, the tensor processing unit (TPU), which it designed in collaboration with Broadcom (AVGO 1.69%).
I'll spare you the technical details, but there are some key differences in role and cost that mean the TPU and GPU are not necessarily interchangeable. But the TPU does represent one of the first real competitors to Nvidia's hardware dominance.
And Anthropic announced late last year that it planned to add up to one million TPU chips to its hardware through 2026, or about one gigawatt of computing capacity. So, even when its apparent rivals win, so too does Alphabet, which is an enviable position for a company to be in, wouldn't you say?
So, given Google's emerging AI dominance and the fact that it's one of only a handful of companies to bring an agentic AI to market, albeit in an experimental prototype capacity, I think it's well poised to be a frontrunner in the agentic step of AI's evolution as a technology.
