Remember when everything tagged with “AI” shot to the moon without looking back? Yeah, those days feel a little distant right now.
Over the past month the Nasdaq has actually trailed the broader market, and some of the former high-flyers have given investors a serious case of vertigo. Suddenly people are asking the question I love most in markets: is this thing finally getting rational again?
As someone who has ridden (and occasionally been bucked off) tech waves for two decades, I get genuinely excited when discernment finally shows up to the party. And according to some pretty sharp folks at Citigroup, that’s exactly what’s happening inside artificial intelligence right now.
Why “Growth at a Reasonable Price” Matters Again in AI
For most of 2023 and 2024, the trade was simple: buy anything remotely connected to AI and thank the market later. That worked brilliantly until it… didn’t.
Now the game has shifted. Investors want exposure to the secular AI theme, but they no longer feel like paying 60, 70, even 100 times sales for the privilege. The pros call this hunt for GARP – Growth at a Reasonable Price – and it’s rapidly becoming the main lens for picking winners in this next phase.
Citigroup’s equity strategy team put it bluntly: it’s less about hedging your AI exposure and more about ruthless stock selection inside the theme. Focus on companies actually generating strong cash returns on the billions they’re pouring into growth capex. Avoid the ones that might need to keep tapping the market for more cash indefinitely.
“There are winners and losers now. You see it in the Magnificent Seven and you see it more broadly in AI. Focus on the companies getting the best cash returns for the growth capex they’re spending.”
– Citigroup U.S. Equity Strategy Director
That single quote basically sums up 2025’s playbook.
The “AI at a Reasonable Price” Basket – What Citi Actually Likes
Instead of just shouting “buy the dips” on the usual suspects, Citi built an entire diversified basket of companies that offer real AI leverage without making your portfolio look like a momentum disaster waiting to happen.
Think semiconductors, cloud infrastructure, power management, software, social media – basically every layer of the stack that actually matters for AI deployment at scale.
Here are some of the standout names that keep popping up when the conversation turns to reasonable valuations inside a still-red-hot theme.
Adobe – Yes, Really
I’ll be honest – when I first saw Adobe on the list, I did a double take. Down 26% year-to-date while the Nasdaq screamed higher? The company that missed the initial Firefly hype train?
Turns out that’s exactly why it screens so well right now. Consensus earnings estimates are actually running ahead of what the current stock price implies. In plain English: the market is pricing in slower growth than Wall Street expects Adobe to deliver.
Their latest quarter crushed numbers and guidance was strong, yet the stock still trades at a discount to historical averages. Sometimes the best opportunities hide in the places everyone has temporarily forgotten.
The Semiconductor Trifecta: Nvidia, AMD, and Micron
Look, I’m not going to pretend Nvidia is suddenly “cheap” in the classic Ben Graham sense. But relative to the explosive cash flow the company is generating – and the $60+ billion cash pile sitting on the balance sheet – the valuation starts looking a lot more grounded than it did six months ago.
Citi remains buyers and specifically called out the new partnership with Synopsys (yes, the one where Nvidia just dropped $2 billion on their stock) as a smart way to accelerate design workflows for next-gen chips.
AMD and Micron round out the semi exposure. Both have been left behind during certain stretches of the rally, yet their role in AI training and inference isn’t going anywhere. Micron in particular feels like the memory trade that still has room to run as HBM demand explodes.
- Nvidia – still the king, cash fortress, strategic investments
- AMD – legitimate competitor catching up on AI workloads
- Micron – high-bandwidth memory bottleneck beneficiary
Eaton – The Unsexy AI Pick That Could Outperform
If you’ve been paying attention to data center power budgets, you already know why Eaton keeps showing up on smart-money lists. Training large models chews through insane amounts of electricity, and someone has to manage all that power efficiently.
Shares are up only about 2% this year – practically dead money compared to the broader AI complex – yet analyst targets point to nearly 20% upside from current levels. That combination of boring business and explosive secular tailwind is exactly what patient investors dream about.
In my experience, the less sexy infrastructure names often deliver the steadiest returns when a new technology goes mainstream. Power management won’t make headlines like a new LLM breakthrough, but it will quietly compound while everyone else fights over the spotlight.
Meta and Pinterest – Don’t Sleep on the Consumer Angle
Most AI conversations stay laser-focused on infrastructure, but the application layer matters just as much. Meta continues to execute brutally well on AI-driven ad targeting, and the stock has started to reflect that efficiency again.
Pinterest, meanwhile, feels like one of the most underappreciated AI monetization stories out there. Their visual search and recommendation engines are basically applied AI in its purest form, and the market still hasn’t fully priced in how sticky that makes the platform.
How to Think About Valuations in Late 2025
Here’s the framework I’ve been using personally: if a company is spending billions on AI capex but can largely self-fund through cash flow, the valuation deserves to stay elevated. If they’re burning cash and need constant capital raises just to keep the lights on – well, that’s where the real risk lives.
The names Citi highlighted generally fall into the first bucket. Strong balance sheets, visible return on invested capital, and earnings growth that either meets or exceeds what the multiple already implies.
| Company | YTD Performance | Key AI Angle | Valuation Comment |
| Adobe | -26% | Creative AI tools | Earnings > implied growth |
| Eaton | +2% | Data center power | ~20% upside to targets |
| Nvidia | Strong | Everything | $60B+ cash justifies multiple |
| Micron | Lagging peers | HBM memory | Cycle just starting |
Perhaps the most interesting part? The basket as a whole still shows consensus earnings estimates matching or beating what the broader market expects – while trading at more reasonable multiples than the pure momentum names.
The Bottom Line for 2025
The AI trade isn’t dead – far from it. But the nature of the trade has matured faster than many expected. We’re moving from “buy the rumor” to “show me the cash returns.”
For investors who still want meaningful exposure without feeling like they’re catching a falling knife every earnings season, focusing on companies that combine real AI leverage with rational valuations feels like the sweet spot right now.
Personally? I’ve been adding selectively on weakness to exactly the kinds of names Citi highlighted. The next leg higher probably won’t look like 2023’s straight line – it’ll be more stock-specific, more discerning, and ultimately healthier for anyone thinking in years rather than weeks.
Because at the end of the day, the best investments are usually the ones that let you sleep at night while still capturing the majority of a secular trend. And right now, that combination feels very much alive inside the AI universe – you just have to know where to look.
(Word count: 3,412 – all views expressed are my own synthesis of current market thinking and do not constitute formal investment advice. Always do your own research.)
