All research

Equity research

Gold Above $4,000 and the Central Banks Behind It

Central banks have bought about 1,019t of gold a year since 2022, more than twice the 2010-21 pace, and spent more even as tonnes fell: about $94bn in 2025 against $63bn in 2022. Our flow model puts gold at $4,850 by December 2026 (+12%). We initiate Overweight on Newmont (PT $132), Neutral on Barrick (PT $54) and Wheaton (PT $146), and Underweight on Agnico Eagle (PT $203).

By K2 Research TeamDec 22, 2025PDF · 9 pages
Gold Above $4,000 and the Central Banks Behind It

Central banks have bought about 1,019t of gold a year since 2022, more than twice the 2010-21 pace, and spent more even as tonnes fell: about $94bn in 2025 against $63bn in 2022. Our flow model puts gold at $4,850 by December 2026 (+12%). We initiate Overweight on Newmont (PT $132), Neutral on Barrick (PT $54) and Wheaton (PT $146), and Underweight on Agnico Eagle (PT $203).

What’s inside

  1. Gold above $4,000 and the central banks behind it: initiating at $4,850 gold; Overweight Newmont
  2. Central banks set the floor; private buyers set the pace
  3. Real yields still matter, but only at the margin
  4. The miners do not need higher gold, but they gear to it
  5. From 473t to 1,000t a year: who is buying
  6. Why the buying is sticky
  7. Tonnes fall when the price rises; dollars do not
  8. The real-yield anchor broke in 2022
  9. How our gold model works
  10. What it takes to break $4,000
  11. Where we differ
  12. The miners: more leverage than the market is paying for
  13. Ratings, price targets and key risks
  14. Newmont (NEM)
  15. Barrick (B)
  16. Agnico Eagle (AEM)
  17. Wheaton Precious Metals (WPM)
  18. What would make us wrong?
  19. Signposts

More research

View all

Neoclouds Fell Up to 17% This Week, and Beta Explains Nearly All of It
Equity researchPDF · 9 pp

Neoclouds Fell Up to 17% This Week, and Beta Explains Nearly All of It

All four neoclouds fell more than the Nasdaq-100’s 4.8% after Microsoft’s and Alphabet’s capex misses, from SpaceX (-7.2%) to IREN (-17.0%), and beta explains each move to within 3pp. We reiterate Overweight on Nebius (PT $207, now +34% upside) and IREN (PT $37), and Underweight on SpaceX (PT $90) and CoreWeave (PT $44).

K2 Research TeamOct 31, 2026
Alphabet’s $500bn One-Day Fall Came From Its Own Results
Equity researchPDF · 9 pp

Alphabet’s $500bn One-Day Fall Came From Its Own Results

Alphabet lost about $500bn of value on 29 October, falling 12.6% on its own 3Q26 results before Reddit had even reported: Search grew 12% against about 15% expected and paid clicks slowed to 8%. We initiate Neutral on Alphabet (PT $305) and Overweight on Reddit (PT $130).

K2 Research TeamOct 31, 2026
Anthropic’s $120bn Deal Is 6% of SpaceX and 2.2x Nebius, So We’d Rather Own Nebius
Equity researchPDF · 9 pp

Anthropic’s $120bn Deal Is 6% of SpaceX and 2.2x Nebius, So We’d Rather Own Nebius

SpaceX is the best-capitalised neocloud we cover, and the dearest: at $1.97tn the market pays about $1.4tn, or 46x annualised revenue, for its AI segment. We initiate on four neoclouds: Underweight SpaceX (PT $90), Overweight Nebius (PT $207) and IREN (PT $37), and reiterate Underweight CoreWeave (PT $44).

K2 Research TeamOct 28, 2026