Here are Thursday's biggest analyst calls: Nvidia, SpaceX, Tesla, AMD, Toast, Meta, Five Below, MP Materials & more

Wall Street delivered a largely mixed slate of calls: Citi reiterated Nvidia as a top data-center semi pick citing strong DRAM supply constraints, while Mizuho raised Applied Materials’ estimates/PT to $650 (prior $540) on AI-driven demand. Offsetting this optimism, KeyBanc downgraded Salesforce to sector weight after survey checks and feedback on Agentforce failed to show momentum, and JPMorgan downgraded Stellantis to Neutral citing roughly ~14 months before lower component costs flow through to launches in FY27/28. Overall, most single-name changes are positioning-driven rather than a clear market-wide catalyst.

Analysis

This call stack is less about “AI winners” and more about second-order bottlenecks: power, cooling, test, and deployment capacity. That favors NVDA/AMAT/LRCX/MKSI/COHU on the obvious side, but also turns FIX and CECO into higher-beta ways to play capex translation because backlog can convert faster than chip-end demand if hyperscaler spending stays elevated. The market usually underprices these adjacency beneficiaries early in an infrastructure cycle, then rerates them when order books stop looking like one-quarter noise and start looking like a multi-quarter capacity constraint.

The cleaner short is AMD: if investors keep rewarding only the names with clear leading-edge share and memory access, AMD becomes a “good company, crowded multiple” risk rather than a growth compounder. Any stumble in inference adoption or GPU share math would likely hit both estimate revisions and multiple, so the downside is more about narrative loss than one quarter of earnings. STLA is the opposite problem—fundamental improvement is too far out to matter for current valuation, so the stock can drift while the market waits for FY27/28 proof.

Contrarian read: the consensus is still treating AI spend as a chip-only trade, but the more durable revenue bridge may be in infrastructure plumbing and contamination-control adjacent names. On the other side, TSLA’s valuation is carrying future products as if timelines are near-dated; that leaves the stock vulnerable to even modest delays, especially if rates stay high and long-duration duration stocks keep de-rating. The immediate price reaction can be noisy, but the 1-3 month catalyst path is earnings revisions and capex commentary; 6-18 months is whether AI infrastructure becomes a sustained buildout versus a temporary digestion phase.

Key Decisions for Investors

• Long NVDA vs short AMD for 1-3 months: express the view that scarce memory access and clearer share leadership should outperform a more contested second-best beneficiary; stop if AMD prints a meaningful data-center share gain or NVDA guides below the market’s capex assumptions.
• Buy a basket of CECO/FIX on pullbacks for 6-12 months: these are higher operating-leverage ways to own the AI buildout beyond semis; risk/reward improves if hyperscaler capex and electrical backlog remain firm into the next earnings cycle.
• Add AMT modestly on any rate-driven weakness for a 3-6 month trade: the thesis is post-consolidation organic growth and multiple support, but it should be sized smaller because higher real yields can still cap REIT multiple expansion.
• Initiate a tactical TSLA put spread or underweight for 1-2 months: the market is pricing execution on products that are still far from contributing; the trade works if delivery/margin prints fail to validate the current growth narrative.
• No strong urge to chase STLA or SPG here; both look more like capital-allocation or patience trades than near-term catalysts unless there is a clear earnings revision inflection.