AI Is Creating More Winners Than Investors Can Track
Models, agents, chips, and infrastructure are advancing at once. Finding the winners is only the beginning.
Listen to the audio version of this article (generated by AI).
There was a time, just a few years ago, when AI investing felt easy.
You could buy Nvidia (NVDA), the hyperscalers, or the companies wiring up the world’s data centers. And then… you could basically stop thinking. The AI boom did the rest.
That simple playbook worked spectacularly.
But it’s no longer the right approach.
Meta (META) just released an open-weight AI model capable of running on an ordinary laptop. Google Maps can now order food, hunt for hotels, and carry out errands on your behalf. Microsoft (MSFT) is reportedly preparing another generation of custom AI chips. And Nvidia is organizing some of the world’s top AI labs around a shared family of open models.
Four developments in four different parts of the AI economy.
Together, they show how many new ways there are to invest in the AI boom.
Gone are the days when AI investing was centered on one chipmaker, one cloud platform, or one kind of technology. The boom is spreading – into personal devices, consumer agents, custom silicon, open-model ecosystems, networking, memory, power, and the software connecting all of it.
That is excellent news for long-term investors.
It also creates a problem.
An investor can understand every one of these trends, pick several good stocks, and still build a bad portfolio.
Finding winners is no longer the hardest part.
Figuring out how they fit together is.
One AI Boom, Several Different Trades
Glimmer Brings More AI Onto the PC
Start with Meta.
This week, the company released Muse Glimmer, a compact open-weight model designed to handle coding, administrative work, and other agentic tasks while running on a standard laptop or PC. Mark Zuckerberg paired the launch with a sweeping vision for “personal superintelligence,” where individuals can run powerful AI systems without depending entirely on a handful of centralized providers.
That pushes the AI trade onto the device.
If capable models can run continuously on consumer hardware, demand spreads beyond giant cloud clusters. AI PCs need better processors, more memory, larger storage systems, stronger connectivity, and efficient power management. The model may run locally, but an entire hardware stack has to support it.
Google Maps Moves From Navigation to Action
Then there is Google Maps.
What began as a navigation product evolved into a local-search engine. Now Google is turning it into something closer to a consumer agent.
Its latest Ask Maps features can help users order food, search for hotels that match specific preferences, find local events, and personalize results using information from other Google services.
Maps is beginning to steer the transaction itself, pulling cloud inference, payments, local-commerce software, restaurant technology, digital advertising, and the businesses inside Google’s distribution network into the trade.
Microsoft Wants More Control of the Chip Stack
Microsoft’s reported Maia 300 plans point to another corner of the market.
According to recent reporting, Microsoft could unveil its next-generation AI accelerator as early as September. The company has already spent years developing proprietary silicon to reduce costs, gain more control over its infrastructure, and lessen its dependence on outside chip suppliers.
Maia changes more than Microsoft’s chip bill.
A custom chip needs an architect. It needs a foundry. It needs advanced packaging, high-bandwidth memory, networking, power systems, cooling equipment, and racks capable of turning silicon into usable compute.
A hyperscaler designing its own accelerator does not remove the supply chain. It rearranges who gets paid.
Nvidia Is Building More Than Hardware
And Nvidia is pushing into yet another layer.
The company formed the Nemotron Coalition with Mistral AI, Cursor, LangChain, Perplexity, Black Forest Labs, and several other leading AI developers. The group is building open frontier models trained on Nvidia’s DGX Cloud, with the first shared foundation supporting the upcoming Nemotron 4 family.
Nvidia is still selling the picks and shovels.
Now it is helping organize the miners, too.
Its hardware dominance gives Nvidia a natural position at the center of an open-model ecosystem. More developers building on Nemotron means more workloads trained and served on Nvidia infrastructure.
AI Is Becoming Its Own Economy
Meta’s Glimmer is an edge-AI story.
Google Maps is a consumer-agent story.
Microsoft’s Maia program is a custom-silicon story.
Nemotron is a model-platform and developer-infrastructure story.
All four belong to the AI boom.
They do not belong in a portfolio for the same reason.
Same Boom, Different Economics
AI now has model makers, consumer platforms, chip designers, memory suppliers, network builders, power providers, and software companies helping agents carry out work.
Each group makes money differently. Each depends on different customers. And each carries a different set of risks.
A new open model may pressure premium API pricing while boosting demand for consumer GPUs. A custom chip can take share from Nvidia inside one cloud platform while creating new revenue for a foundry, an HBM supplier, and a networking company. A consumer agent can strengthen Google’s ecosystem while generating more work for payments and local-commerce providers.
That complexity comes with maturity. Capital is moving beyond the obvious names and into companies solving increasingly specific problems.
Our own results show what that can look like.
Lumentum (LITE), an optical-networking supplier that most investors once viewed as a niche component maker, is currently sitting on a roughly 645% gain from our August 2025 recommendation. Louis Navellier’s Nvidia position is up roughly 375%.
Those profits came from different layers of the same broad buildout: one from the chips doing the work, the other from the optical infrastructure moving the data.
The winners are multiplying across the AI economy.
A Collection of Good Stocks Is Not Necessarily a Good Portfolio
This is the point where AI investing gets harder.
Suppose an investor owns Microsoft, Amazon (AMZN), Alphabet (GOOGL), Nvidia, Broadcom (AVGO), Marvell (MRVL), Taiwan Semiconductor (TSM), Micron (MU), and several networking suppliers.
That may look diversified. In reality, much of the portfolio could depend on the same underlying variable: hyperscaler infrastructure spending.
If that spending ever slows, several positions may react at once.
The opposite problem can happen, too. An investor may own one exciting robotics stock, one experimental power company, and one small AI-software name. The themes are different, but the risk may be heavily concentrated in early-stage businesses with little room for execution mistakes.
Position size matters just as much as stock selection.
A profitable hyperscaler with hundreds of billions in contracted revenue should not carry the same weight as a speculative component supplier. A mature semiconductor leader should not be treated like an emerging agent platform. Two stocks operating in different industries may still depend on the same customer or capital-spending cycle.
A good AI portfolio gives every holding a job.
Some positions form the core. Others provide exposure to emerging layers of the market. Smaller allocations create room for higher-upside ideas without allowing one failed thesis to overwhelm the entire portfolio.
The goal is coherence.
That has become much harder as the number of credible AI investments has grown.
Our Success Created a New Problem
InvestorPlace’s AI research team has produced more than 200 recommendations over the past year.
That reflects the scale of the opportunity. It also leaves readers with one glaring question: What are they supposed to do with all of them?
Owning 200 stocks is not a strategy. Neither is chasing whichever recommendation happens to be newest.
Investors need to know which ideas deserve a place in the portfolio, which ones overlap, and how much capital each position should receive. That is the problem our newly rebuilt AI Revolution Portfolio is designed to solve.
The last time we did this, the portfolio more than doubled the Nasdaq’s return.
Following its December 2024 rebalance through July 23, the AI Revolution Portfolio gained 58%. Over that same stretch, the Nasdaq rose 25%, the S&P 500 gained 24.4%, and the Dow advanced 19%.
The lesson from that outperformance goes beyond any single winner. Our portfolio captured gains across multiple parts of the AI economy while organizing those positions around one coherent market view.
Rebuilding the AI Revolution Portfolio
Since that last rebalance, the market has changed again.
Models are moving onto personal computers. Agents are beginning to transact. Hyperscalers are designing their own chips. Nvidia is helping build an open-model ecosystem. New infrastructure bottlenecks are appearing as quickly as old ones get solved.
So we went back to work.
Louis Navellier, Eric Fry, and I have gone through our AI research and narrowed that sprawling universe into roughly 20 stocks we collectively believe deserve capital now.
The market is creating winners across models, agents, chips, optics, memory, energy, and infrastructure. No single recommendation can capture all of it. And simply adding more tickers does not solve the problem.
AI is creating more winners than investors can track.
Now the real edge comes from knowing which ones deserve your money, how they complement one another, and how large each position should be.
Louis, Eric, and I are about to unveil the newly rebuilt AI Revolution Portfolio.
Sign up to see the portfolio, the recommended allocations, and the thinking behind every position right here.
