2 AI Infrastructure Stocks That Could Rise 25% and 80% Despite Overdone Spending Fears

AMD and Micron are highlighted as major AI beneficiaries, with Wall Street price targets raised sharply to $625 for AMD and $1,625 for Micron. The article argues AMD is early in a large inference and agentic AI opportunity, while Micron is benefiting from a memory supercycle driven by HBM demand and tight DRAM supply. Both stocks are framed as having meaningful upside, though the piece is largely opinionated analyst commentary rather than new company-specific news.

Analysis

The incremental winner here is not just AMD or MU in isolation, but the whole non-Nvidia AI supply chain that benefits from a broadening of workloads from training into inference and agentic orchestration. That shift changes the bottleneck from peak FLOPS to memory bandwidth, package integration, and CPU coordination, which is why HBM vendors and advanced packaging/tooling providers can sustain margin expansion even if AI accelerator ASPs flatten. The second-order implication is that Nvidia’s moat becomes more price-sensitive at the margin while AMD gains share by being the “good enough” alternative for hyperscalers optimizing total cost per token.

For Micron, the market is likely underappreciating how capacity discipline can extend the cycle beyond a normal memory upturn. Multi-year supply commitments reduce spot volatility and should support a higher valuation floor, but the real risk is not demand collapsing—it is supply response slipping in faster than expected if HBM yields improve or if one of the big three decides to chase share. The constraint from lithography equipment helps, but it is not permanent; once capex translates into bits, memory equities can rerate down very quickly over a 2-3 quarter window.

The cleanest way to express the view is long AMD versus a short basket of beneficiaries of a slower AI spend cycle, or long MU versus semiconductor cyclicals that do not have direct HBM exposure. Near-term catalysts are earnings revisions and hyperscaler commentary over the next 1-2 quarters; the medium-term catalyst is evidence that inference deployments are driving a second wave of GPU/CPU orders rather than simply shifting spend away from training. The main contrarian risk is that the market has already priced in a lot of the “AI second leg” narrative, so upside likely requires visible gross margin and backlog conversion, not just TAM expansion language.

Key Decisions for Investors

• Stay long AMD into the next 1-2 earnings cycles, but size as a barbell: core equity plus upside call spreads to capture a rerating if hyperscaler commentary confirms inference demand acceleration; cut if design-win conversion slows or gross margin expansion disappoints.
• Long MU on pullbacks with a 3-6 month horizon; the setup is strongest while supply remains constrained and pricing power is still being revised upward. Use stop-loss discipline if memory capex announcements start to imply earlier-than-expected supply normalization.
• Pair trade: long MU / short a basket of lower-quality semiconductor cyclicals with weak pricing power and no HBM exposure. This isolates the memory supercycle while reducing beta to a broad AI hardware selloff.
• Consider long AMD / short NVDA small-to-moderate notional as a relative-value trade if hyperscalers continue diversifying suppliers; the trade works best if AI inference spending broadens faster than model-training spending.
• Avoid chasing after multi-day spikes; both names are now momentum-sensitive. Best entry is on post-earnings or post-guidance consolidations where revisions can be validated by backlog and supply-chain commentary.