Which AI companies are smart to invest in right now?

General
1

I’m trying to build a long-term portfolio focused on artificial intelligence, but I’m overwhelmed by the number of AI stocks, ETFs, and big tech players claiming AI leadership. I’ve read a bunch of articles and watched videos, but most feel like hype or sponsored content. Could you share which AI companies you think are actually worth investing in, and why, based on fundamentals, real AI products, or growth potential?

2

Short version. Think in “layers” of AI, not stock tips.

  1. Core infra players
    These supply the picks and shovels.

• Nvidia (NVDA)
Still the leader in AI chips. Revenue from data center AI is huge and growing fast. Valuation is rich, so size your position. Treat it as long term, ignore short term noise.

• AMD (AMD)
Trying to take share from Nvidia with MI300 chips. Backed by large cloud deals. Higher risk than NVDA, but also more upside if they execute.

• TSMC (TSM)
Makes chips for Nvidia, AMD, Apple and others. Less “pure AI name”, but crucial to the supply chain. More stable than chasing small AI stocks.

  1. Cloud and platform players
    These own the AI platforms and sell them to enterprises.

• Microsoft (MSFT)
Has OpenAI partnership, Copilot across Office, GitHub and Windows, and Azure AI. Strong balance sheet. If you want one “anchor” AI stock, this is my top pick.

• Alphabet / Google (GOOGL)
Search, YouTube, Android, Gemini, cloud. Strong AI research, sometimes messy product rollouts. Long term, their data and infra still matter a lot.

• Amazon (AMZN)
AWS is pushing AI services, custom chips like Trainium and Inferentia, plus AI in retail and ads. More diversified than “AI only” names.

• Meta (META)
AI for ads, recommendation engines, open source models like Llama. More cyclical and sentiment driven. Still interesting if you tolerate swings.

  1. Tools and “picks & shovels” software
    • Snowflake (SNOW)
    Data platform used as a base for AI workloads. Valuation heavy. I would use a small allocation.

• Datadog (DDOG)
Monitoring and observability for complex systems. AI workloads need this kind of tooling. Not pure AI, but benefits as AI infra grows.

  1. Enterprise AI and apps
    Riskier, but some have real revenue.

• Palantir (PLTR)
Government and enterprise analytics. AI narrative strong, but also hype heavy. Check valuation and your risk tolerance.

• Adobe (ADBE)
Using AI inside creative tools like Photoshop and Premiere. Real customers, real cash flow, less “story stock”.

  1. ETFs if you feel overwhelmed
    Good if you do not want to stock pick.

• Global X Robotics & Artificial Intelligence (BOTZ)
More focused on robotics and automation.

• Global X Artificial Intelligence & Technology (AIQ)

• iShares Exponential Technologies (XT)
Broader tech, AI is one theme among others.

ETFs spread risk. On the flip side, you get some weak holdings mixed in.

  1. How to build a simple long term AI sleeve
    Example for a patient 10+ year horizon, not advice, just a framework:

• 40 percent in big cloud and platform
MSFT, GOOGL, AMZN, META
• 30 percent in infra
NVDA, AMD, TSM
• 20 percent in one or two ETFs
BOTZ, AIQ, XT
• 10 percent in “satellites”
SNOW, DDOG, PLTR, ADBE or others you research

Rebalance once or twice a year. Ignore daily headlines.

  1. Filters to avoid junk AI stocks
    If a company yells “AI” all day, check:

• Revenue growth above 15 to 20 percent year over year
• Positive or improving free cash flow
• Real customers, not only “pilots” and press releases
• Reasonable price to sales for their growth rate

If you see tiny revenue, constant share dilution, and every press release says “AI leader”, step back.

Last thing. Decide your time horizon and risk level first. Then pick positions and size them so you can sleep at night when AI stocks drop 30 percent in a month, because at some point they will.

3

@jeff’s “layers” idea is solid, but I’d push a bit in a different direction: instead of asking “which AI stocks,” think “where in the AI stack is the power going to sit once hype cools off?”

A few angles that don’t just rehash his list:

  1. Don’t over-concentrate on pure model providers
    Everyone’s drooling over frontier models, but history says pure tech layers get commoditized. Open source is already eating into moat stories. I’d be cautious making a portfolio that depends heavily on whoever has the “best model” in 2024. That advantage can vanish in 18 months.

  2. Look harder at “AI-accelerated winners,” not “AI companies”
    Some of the strongest long term plays might be boring names that quietly weaponize AI:

  • ServiceNow (NOW)
    Deep enterprise hooks, workflows everywhere, pushing AI into automation for large customers. Less sexy than LLM names, more locked-in revenue.

  • Intuit (INTU)
    TurboTax, QuickBooks, Credit Karma. Their AI play is literally “we sit on a mountain of financial data.” That is hard to replace and very monetizable.

  • Salesforce (CRM)
    Still has the CRM lock. Their “Einstein” stuff is marketing heavy, but AI inside sales workflows is sticky if they get it right.

None of these scream “AI stock” in headlines, but they can ride AI to higher pricing power and lower churn.

  1. Monopolistic data advantage > fancy AI branding
    Instead of reading who shouts “AI” the loudest, ask “who owns data that is painful or impossible to replicate?”

Some examples to research (not a buy list, do your own DD):

  • Bloomberg (not public, just an example of data moat logic)
  • RELX (RELX) and similar data / analytics players
  • Moody’s (MCO) and S&P Global (SPGI) in finance data & analytics
  • Health data / workflow vendors in healthcare IT

These are the kind of companies that can plug in any decent model and still own the value layer because they own the data pipelines and relationships.

  1. Geography and regulation are very under-rated
    Almost no one on these threads thinks about where governments will:
  • Subsidize AI capex
  • Restrict exports
  • Enforce data residency and privacy

I would not ignore:

  • ASML (ASML)
    Extreme ultraviolet lithography is a choke point. High cyclicality risk, but structurally essential.

  • Some Japanese and European automation players
    Not pure AI, but AI-driven robotics and factory modernization are going to be multi decade trends, regardless of which US model vendor is hot.

  1. Where I slightly disagree with @jeff
  • I’m more skeptical of Palantir at current valuations. Their story is AI-heavy, but contract concentration and politics risk are very real. If you buy it, size it like a speculation, not a core holding.
  • I’m less excited by “AI branded” ETFs. A lot of them are just tech momentum funds with a buzzword. If you go ETF, actually open the holdings list and ask yourself which names you’d never buy on their own. If that list is long, maybe skip it.
  1. Risk controls that matter more than the ticker
    Instead of asking “is NVDA or MSFT better,” decide:
  • How much of your net worth are you willing to let sit in AI and be down 50 percent without panicking?
  • What percent max in any single name? Many people have way too much in one darling ticker and call it a “long term view.” That is just concentration risk.

Basic structure you can tweak:

  • 50 to 60 percent in diversified mega-cap tech that uses AI across the business
    (MSFT, GOOGL, AMZN, maybe AAPL, plus 1 or 2 “data moat” names like INTU / NOW / CRM / MCO / SPGI)
  • 20 to 30 percent in infra / semi / tooling
    (NVDA, TSM, ASML, plus 1 or 2 software infra like DDOG / SNOW if you accept volatility)
  • 10 to 20 percent “bets”
    Stuff with narrative-heavy AI upside that you are emotionally prepared to see cut in half without rage-selling.
  1. A quick sanity check to run on any “AI” stock
  • If AI stopped being a buzzword tomorrow, would this company still have a good business?
  • If models get 10x cheaper, does this help them (scaling usage) or hurt them (their moat disappears)?
  • Are they using AI to reduce costs and improve margins, or only as a story to sell shares?

If the only answer is “they keep calling themselves an AI leader in every press release,” I’d personally pass, no matter how many YouTube guys say it is “the next Nvidia.”

TL;DR: use AI as a lens, not a shopping category. Focus on where durable power sits: chips and tools, data moats, and entrenched enterprise workflows. Let everyone else chase whichever company just said “AI” 14 times on the last earnings call.