Jensen Huang Says AI Needs 1,000 Times More Power. Here's Why Oklo and NuScale Power May Still Suffer.

Goldman Sachs forecasts AI spending will exceed $1 trillion in 2026, but investors are increasingly concerned about data-center funding, community support and power availability—not AI demand. Bank of America estimates meaningful real-world adoption of small modular reactors may not scale until the 2030s or 2040s; Oklo and NuScale remain unprofitable and have raised cash through share sales that dilute shareholders. Both stocks are down more than 60% this year, and delayed deployment could require further issuance at lower prices.

Analysis

The key distinction is between AI demand and the ability to finance and energize capacity. That weakens the near-term equity case for SMR developers more than it weakens AI infrastructure demand: projects can be deferred, resized, or served first by gas, renewables, storage, and grid upgrades while nuclear licensing and construction timelines play out. The second-order beneficiaries are therefore dispatchable generation and transmission/grid-equipment providers; SMR developers risk losing the bridge-period economics even if long-run nuclear demand proves real.

For Oklo and NuScale, the critical variable is not headline data-center power demand but whether they can reach funded, permitted, binding projects before cash needs force further dilution. Their recent share declines may already price in meaningful execution risk, so a straight short has asymmetric squeeze risk from policy or project announcements. Conversely, treating AI power demand as a near-term SMR earnings catalyst conflates long-run need with deployable supply.

Days to weeks: sentiment and financing headlines can dominate. Over 1–3 months, track cash burn/runway, equity issuance, customer commitments with financing attached, and licensing milestones. Over 6–18 months, evidence of delayed data-center builds or power procurement shifting to nearer-term alternatives would further defer SMR revenue. The thesis is falsified by credible, financed projects advancing through regulatory milestones on schedule, without material dilution. NVDA demand is not directly disproved; the risk is that constrained power and capex funding delay customer deployments and eventually equipment orders.

Key Decisions for Investors

• Avoid adding to OKLO or SMR solely on the AI-power narrative. Before taking exposure, verify each company’s cash runway, expected burn, remaining issuance capacity, and whether announced customer projects have committed financing and power-offtake terms.
• For a bearish expression, prefer a defined-risk bearish call spread in OKLO or SMR rather than an outright short, and only after checking option liquidity and implied volatility; the sharp prior drawdown leaves substantial event-driven squeeze risk. Keep sizing small and exit the thesis if funded projects clear key licensing milestones without further equity issuance.
• Do not short NVDA on this evidence alone. Monitor hyperscaler capex guidance and deployment schedules; a sustained capex slowdown or order deferrals would be the cleaner signal that power and funding constraints are reaching AI hardware demand.
• Watch for substitution toward gas generation, renewables plus storage, and grid investment as near-term beneficiaries, but require evidence in utility procurement, interconnection approvals, or project awards before making a sector trade.