If you spent the first half of 2026 glued to Nvidia’s $NVDA ( ▲ 2.93% ) ticker, you might have walked away feeling a bit underwhelmed. In a year when the undisputed poster child of the generative AI boom barely budged — logging modest low-teens gains — the real market fireworks were happening elsewhere.

Wall Street didn't abandon the AI trade. Instead, investors broadened their scope, sparking an aggressive rotation into physical infrastructure, memory, and foundry turnarounds. If you were searching for the best AI stocks to buy now in 2026, the winners weren't always the usual mega-cap suspects.

Let's look at the scoreboard and unpack why memory and CPUs stole the show in H1.

📉 The H1 2026 AI Stock Scorecard

The narrative that "Nvidia is the entire AI trade" officially fractured over the last six months. While Nvidia consolidated its multi-trillion-dollar gains, suppliers addressing critical physical bottlenecks surged hundreds of percent.

🔍 Memory Scarcity Takes Center Stage

Memory chip stocks stole the show

I’m watching memory markets closely right now, and for good reason. As AI models scale past parameters that fit neatly into standard GPU memory pools, high-bandwidth memory (HBM) has become the ultimate chokepoint in the datacenter.

Micron $MU ( ▼ 5.9% ) didn’t just participate in this cycle; it dominated it. With HBM supply completely booked out through late 2027 and well into 2028, pricing power shifted decisively toward manufacturers.

  • Pricing power is sticky. Unlike commoditized DRAM cycles of the past, AI-grade memory commands structural margin expansion.

  • The beta trade. Micron acted as a high-beta play on infrastructure scarcity, rewarding patient investors who looked past software hype.

As I noted in our recent deep dive on Lam Research $LRCX ( ▼ 1.58% ), wafer fabrication equipment and advanced packaging are where the rubber meets the road. Memory isn't just an auxiliary component anymore: it's the core limiter of inference speed.

⚡ The CPU Resurgence and Foundry Turnaround

For the past three years, the market treated traditional server CPUs as legacy tech destined to stagnate. Then came 2026.

Intel surprised skeptics by proving that data center AI isn't only about GPUs: it's about orchestrating massive heterogeneous workloads. Intel’s Data Center and AI segment jumped over 20% year-over-year, providing the foundational compute backbone that complementary accelerators plug into.

  • The re-rating catalyst. When a legacy heavyweight trades at depressed valuations and delivers triple-digit gains, it signals a massive sentiment shift.

  • Supply chain rebalancing. Buyers are actively diversifying supply chains away from single-vendor bottlenecks, creating breathing room for turnaround plays.

It’s a classic reminder: in cyclical growth investing, you never want to write off an asset when consensus sentiment reaches peak pessimism.

⚠️ The Great Rotation: Hardware vs. Software

Surging CPU demand

While hardware and memory suppliers enjoyed a banner first half, high-multiple software names faced a reality check. Palantir $PLTR ( ▲ 0.65% ) and other pure-play AI software darlings experienced meaningful pullbacks — some dropping over 30% — as Wall Street demanded proof of monetization matching the lofty valuations.

This wasn't a rejection of enterprise AI adoption. It was a healthy digestion phase.

  • Infrastructure comes first. You cannot deploy autonomous agentic workflows or heavy enterprise analytics without reliable power, cooling, and high-speed memory.

  • Valuation discipline matters. Markets are currently rewarding cash-flowing physical assets over speculative future software multiples.

This echoes what we saw across the power grid landscape. As highlighted in our analysis of Bloom Energy $BE ( ▼ 0.63% ), physical constraints dictate the speed of digital progress.

🔭 Looking Ahead to H2 2026

Investment roadmap

So, how should we position our portfolios as we head into the second half of the year?

First, keep risk controls tight. Multi-hundred-percent runs in names like Micron and Intel naturally invite near-term volatility and profit-taking. Second, don't mistake a quiet stock price for a broken business: Nvidia’s underlying cash flows and GPU dominance remain entirely intact, even if its stock took a breather to digest prior gains.

The overarching takeaway from H1 2026 is simple: the AI trade has matured. The era of blindly buying any stock with "AI" in the press release is over. Today, sustainable outperformance requires identifying where the next physical bottleneck lies: whether in memory, advanced packaging, or power generation.

Want to stay ahead of these sector rotations before they hit mainstream headlines? Make sure you're subscribed to The Latte to get our institutional-grade research delivered straight to your inbox. If you're ready to unlock our full high-conviction growth portfolio and early-access deep dives, consider upgrading to The Latte Premium.

Stay diversified, keep your research rigorous, and I'll see you in the next issue.

George

Disclaimer: This post is for informational purposes only and does not constitute financial advice. Always conduct your own stock market research or consult with a financial advisor before making any investment.

Keep Reading