Meta Signs Multi-Gigawatt Nuclear Deals for AI Data Centers | Bloomberg Intelligence

Meta’s new multi-gigawatt nuclear power agreements were framed as a sign that AI infrastructure spending is still accelerating and that power availability is becoming a core constraint. The analyst said roughly 1 gigawatt of AI data center capacity can cost about $50 billion, and argued that Meta’s deal suggests the company may keep spending more than $100 billion in capex in 2026 and remain on that path for the next 3 to 4 years. Nuclear was presented as an important alternative because many hyperscalers have preferred natural gas, but turbine backlogs make those projects slower to secure.

The restaurant segment focused on a weak December followed by a likely rebound in 2026. The analyst said cold weather, snow, and the worst flu season in 25 years hit traffic, with BI’s U.S. restaurant same-store sales down 1.1% in December. Quick service and fast casual held up better than casual dining, but the view was that January and the first half of 2026 should improve as comparisons ease and the industry benefits from tax reform and lower oil prices.

General Motors’ latest $6 billion charge on EV and battery cutbacks was cast as part of a broader reset in the U.S. auto market. The analyst said GM’s pivot reflects weaker EV demand and battery supply realities, but also that the market still values the company’s profitable full-size pickups and SUVs. He said hybrid technology is increasingly the bridge, especially as excess battery capacity gets absorbed by more efficient gasoline-electric vehicles.