Meta eyes equity raise while oil demand slides and Iran tensions spike

Meta is weighing a major equity raise to fund AI infrastructure, according to the Financial Times. The move signals that even the best-capitalized hyperscalers are re-rating the cost of staying competitive in the foundation-model race. When a company with Meta's free-cash-flow profile considers dilution, the message is clear: AI capex is not a one-year budget line, it is a multi-year balance-sheet commitment.

Goldman Sachs cut its global oil demand forecast and flagged downside risks to its price deck, citing slowing consumption across major economies. The revision arrives as the WTO reports early signs that global goods trade is decelerating. The demand picture is deteriorating faster than supply adjustments can absorb, and the usual lag between macro slowdown and energy repricing is compressing.

At the same time, Iran launched multiple drones toward the Strait of Hormuz, prompting US strikes on Iranian radar sites. The US energy secretary said lower gasoline prices will ultimately require a resolution with Iran, acknowledging the geopolitical premium embedded in current pricing. The tape now reflects two opposing forces: weakening demand fundamentals and rising supply-disruption risk.

Elsewhere, India reported 7.8% GDP growth in the January-March quarter, outpacing most large economies. Indonesia's central bank and finance ministry agreed to boost asset yields to support the rupiah, a rare coordinated intervention that underscores currency pressure across emerging Asia. SpaceX signed a cloud deal with Google, adding another hyperscale customer to its Starlink enterprise pipeline.

The through-line across these stories is re-rating. Meta is re-rating the capital cost of AI leadership. Goldman is re-rating oil demand. Policymakers are re-rating the geopolitical risk premium. And central banks in the developing world are re-rating the tools they need to defend their currencies. Markets price forward, but the forward is moving faster than the consensus models assumed.