If you thought the AI boom peaked with chatbots and image generators, think again. According to Nvidia CEO Jensen Huang, we’re only just getting started — and the next wave could make today’s tools look like dial-up internet.
At Nvidia’s 2025 GTC keynote, Huang unveiled the company’s vision for Agentic AI — a powerful new generation of artificial intelligence capable of executing complex, multi-step tasks autonomously. And while it’s still in its early innings, investors are already starting to pay attention.
What is Agentic AI?
Unlike traditional generative AI models like ChatGPT, Agentic AI doesn’t just generate content based on a prompt. It can act on that prompt — making decisions, taking actions, and even executing a sequence of tasks on a user’s behalf.
Think of it this way: instead of asking an AI tool to “find a flight to New York,” you could ask it to “plan a three-day business trip to New York with meetings, hotel, and transport.” An agentic model wouldn’t just generate text — it would book the flight, reserve the hotel, draft your calendar, and schedule your Uber.
Huang told GTC attendees this kind of capability could require 100x more computing power than current AI systems demand. That could reshape everything from chip design to how data centers are built and powered.
Real-World Use Cases
Agentic AI isn’t science fiction — in fact, the foundational tools already exist. But Huang sees a near future where these systems move beyond niche productivity apps and into massive industries, including:
- Travel & Personal Services: Booking multi-leg trips, coordinating meetings, and managing schedules autonomously
- Logistics: Managing supply chains, scheduling warehouse operations, and even optimizing delivery routes in real-time
- Enterprise Automation: Acting as digital assistants that don’t just suggest actions — they take them, across CRM, sales, HR, and customer service platforms
- Robotics Integration: When combined with Nvidia’s robotics initiatives, agentic AI could enable autonomous machines capable of adapting to real-world conditions on the fly
The Power Problem: Why the Infrastructure Isn’t Ready
The leap to agentic AI doesn’t just require smarter models — it requires a lot more power. Literally.
Huang estimates that agentic AI will need 100 times the computing resources of today’s leading generative tools. That demand has implications far beyond Nvidia’s chip architecture — it’s a challenge for data centers, power grids, cooling systems, and cloud platforms.
The hyperscalers are already responding. According to recent earnings calls:
- Amazon Web Services (AWS) plans to spend over $105 billion in 2025, the majority going toward tech infrastructure to support AI
- Google Cloud (Alphabet) is ramping up to $75 billion in capital spending, focusing on servers and networking
- Microsoft plans to invest $80 billion in its fiscal 2025, nearly doubling its AI-enabled data center capacity
Those numbers underscore how serious the infrastructure needs are — and how much capital is flowing into companies building the foundation for AI at scale.
How to Invest: Stocks and Sectors to Watch
Agentic AI is still in development, but the market is already shifting in anticipation of its demands. Here are some places investors may want to look:
Semiconductors
Nvidia (NASDAQ: NVDA) remains the clear leader, with its just-launched Blackwell chips already driving billions in booked sales — and a next-gen architecture, Rubin, expected to be 14x more powerful and coming in 2026. If agentic AI takes off, Nvidia is likely to be at the center of the storm.
Other chipmakers with exposure to AI compute or memory bandwidth needs:
- AMD (NASDAQ: AMD) – strong presence in GPUs and AI accelerators
- Marvell Technology (NASDAQ: MRVL) – builds networking chips for data centers
- Broadcom (NASDAQ: AVGO) – infrastructure chips and custom silicon for hyperscalers
Power & Cooling Infrastructure
The power draw from future AI workloads could overwhelm legacy systems. Companies enabling efficient energy use, heat dissipation, and data center scalability could benefit:
- Vertiv Holdings (NYSE: VRT) – power and thermal management for data centers
- Eaton Corp (NYSE: ETN) – industrial power infrastructure and electrical systems
- Schneider Electric (EPA: SU) – global leader in energy management and automation
Cloud Providers
Big cloud platforms are not just end-users of AI chips — they’re also building proprietary AI solutions. Consider:
- Microsoft (NASDAQ: MSFT) – Azure is a key Nvidia partner and major AI investor
- Alphabet (NASDAQ: GOOGL) – Google Cloud’s TPU hardware and AI software suite are central to the stack
- Amazon (NASDAQ: AMZN) – AWS is the largest cloud provider and major buyer of Nvidia hardware
Bottom Line
Agentic AI is more than just a buzzword. If it delivers on its promise, it could radically expand what AI is capable of — and what it demands from the systems that power it. For investors, this is the kind of inflection point that doesn’t come around often.
It’s early, but the infrastructure buildout is already underway. The companies laying the groundwork for this shift — from chipmakers to cloud giants — could be the ones leading the next wave of growth.
