5 AI Infrastructure Stocks to Buy and Hold Through 2030

The article presents five ways to invest across AI infrastructure, from Marvell’s networking chips and AMD’s data-center systems to Forgent’s power equipment, Oracle’s cloud capacity, and Brookfield’s physical assets. Reported growth includes Marvell data-center revenue up 46%, AMD revenue up 50%, and Forgent’s projected revenue growth of about 69%, backed by a $3 billion backlog. Risks include premium valuations, Oracle’s data-center financing needs, and dependence on continued hyperscaler spending; the article also cites planned acquisitions by Marvell and AMD and Brookfield’s proposed $100 billion Kentucky campus.

Analysis

The proposed “diversified” basket is still one crowded trade: hyperscaler capex. Different layers change who captures the spend, but not the common downside if projects are delayed, power connections slip, or AI returns fail to justify build-outs. The more useful distinction is conversion risk: equipment orders can become revenue sooner than campuses become utilized assets, while contracted commitments do not guarantee profitable delivery.

Near term, treat the named deal and project claims as diligence items, not catalysts to price in: verify the status, terms, and closing conditions of Marvell–Lumentum and AMD–World Labs, and the financing, permits, and power/interconnection path for Brookfield’s Kentucky plan. A deal can improve optical access or talent, but integration and customer concentration may absorb the benefit. For Forgent, backlog visibility is not a substitute for cash conversion; the CFO departure raises the value of checking controls, working capital, and guidance before buying the drawdown.

Over 1–3 months, watch hyperscaler capex commentary, order cancellations, and supplier lead times. Over 6–18 months, the key test is whether installed capacity earns returns: weak utilization would pressure Oracle’s financing economics and Brookfield’s project pipeline, as well as upstream equipment demand. Oracle’s contracted revenue can support visibility but does not remove the risk of capex arriving ahead of cash receipts. Power scarcity may benefit equipment suppliers and generators, but interconnection and permitting can defer revenue rather than eliminate the bottleneck.

Contrarian point: owning five AI-linked names is not meaningful factor diversification. The article’s valuation and growth framing also underweights execution and funding risk. The clearest relative-value expression is a small, hedged Marvell-over-NVIDIA position only if custom-chip and optical wins are verified; otherwise, no immediate basket trade.

Key Decisions for Investors

• Relative-value watch: consider a dollar-neutral long MRVL / short NVDA only after verifying Marvell’s custom-chip customer wins and the Lumentum transaction terms. Thesis is share gains in networking/custom silicon, not a broad AI-spending acceleration. Reassess if Marvell data-center growth or order commentary weakens, or NVIDIA’s competitive position proves more durable than expected.
• Do not buy FPS solely on backlog or the share-price decline. First verify the CFO transition, backlog cancellation terms, cash conversion, and updated delivery guidance. A clean transition plus maintained execution could make the pullback attractive; adverse guidance, working-capital deterioration, or unexplained control issues falsify that setup.
• Keep ORCL on a financing watchlist rather than treating RPO as de-risked earnings. Track capex versus operating cash flow, debt issuance and funding costs, and evidence that contracted capacity is being delivered and monetized. Slower RPO conversion or rising financing pressure would argue against adding exposure.
• Before underwriting BN or NEE from the Kentucky campus narrative, confirm project-level financing, permits, customer commitments, and grid/interconnection milestones. Delays would push cash flows out and weaken the near-term thesis even if long-run power demand remains strong.