Better Artificial Intelligence Stock: Applied Digital vs. Super Micro Computer

Super Micro Computer is presented as the preferable long-term AI infrastructure investment, combining FY2026 revenue growth of 77.8% to $39.1 billion with $2.2 billion of net income and a substantially lower valuation of 0.6x sales and 8.6x forward earnings. Management forecast FY2027 sales of $65 billion-$72 billion, implying 66%-84% growth, although free cash flow was negative $7.0 billion and legal, reporting-control, supply-chain, and export-control risks remain. Applied Digital grew FY2026 revenue 167% to $611.3 million and has $36 billion of lease commitments, but posted a $249.2 million loss, negative $2.8 billion free cash flow, more than $5 billion in debt, and trades at a 12.7x price-to-sales ratio.

Analysis

The economically cleaner AI-infrastructure exposure is SMCI, but the apparent valuation discount is compensation for governance and working-capital risk rather than a free multiple expansion. Its incremental revenue is highly sensitive to GPU availability and rack-level mix: a sustained shift toward liquid-cooled, fully integrated systems can preserve earnings leverage, while a component-allocation shortfall or customer insourcing would turn the large inventory build into a cash-conversion problem within 1-2 quarters. NVDA retains the stronger bargaining position in this chain; SMCI captures assembly throughput, whereas NVDA captures scarcity rents.

APLD is a levered construction-and-financing vehicle, not yet a conventional recurring-revenue landlord. The key variable is not announced lease value but conversion of contracted capacity into energized, commissioned facilities without further equity or expensive debt issuance. Any delay in power interconnection, customer financing stress at CRWV, or higher project borrowing costs can create a double hit: deferred rent commencement plus dilution/refinancing risk. Over 6-18 months, successful commissioning could sharply reduce perceived risk and rerate the equity; before then, equity holders are effectively subordinated to project lenders.

Consensus likely overweights the headline discount in SMCI and the headline backlog at APLD. SMCI's low sales multiple can remain structurally low if server margins normalize as OEM competition intensifies, while APLD's premium can compress materially even with backlog intact if its cost of capital remains above project return thresholds. The near-term catalyst set favors SMCI only if its next earnings release demonstrates inventory conversion and reiterates gross-margin discipline; APLD needs independently verifiable construction milestones, power availability, and fully funded capex—not additional contract announcements.

Key Decisions for Investors

• Initiate a 3-6 month relative-value position: long SMCI / short APLD, sized beta-neutral. Target a 15-25% relative return if SMCI converts inventory to revenue while APLD faces financing or commissioning scrutiny; exit if SMCI gross margin declines by more than 200 bps sequentially or APLD demonstrates fully funded, on-time facility energization.
• For SMCI, wait for the next reported quarter to confirm operating-cash-flow recovery and no renewed filing/control issue before increasing to a core long. A revenue beat without inventory-days improvement is not confirmation; governance or export-control developments are thesis-invalidating event risks.
• Maintain a credit-and-counterparty alert on CRWV: deterioration in its funding access, lease-payment disclosures, or GPU-capex guidance should trigger a reduction in APLD exposure immediately, since APLD's contracted economics are concentrated and long-dated.
• Prefer NVDA over both names for unhedged AI-capex exposure over 6-12 months: it is less exposed to data-center construction execution and server-assembler margin compression. Reassess if hyperscaler GPU procurement guidance weakens or supply broadens enough to reduce accelerator pricing power.