What happens to the extra buy orders in London’s gold auction?

Read the figure at the end of the round

ICE Benchmark Administration runs the auctions behind the London AM and PM gold benchmarks. The instrument is unallocated gold for spot settlement in London. Participants see a proposed price and have a fixed period to enter, amend or cancel orders. At the end of the round, order entry pauses while the system calculates the difference between quantities to buy and sell.

The sign convention matters in a note. I would explicitly label positive as excess buying in the example below. The measured quantity describes orders in that round at that proposed price. An estimate of how many bars are available in Dubai, or how much gold was mined this week, would answer an entirely different question.

Source notes: [1] ICE Benchmark Administration — LBMA Precious Metals · [2] ICE Benchmark Administration — IBA Gold Auction Specification

Work through an auction with two rounds

Consider invented orders for 66,000 ounces to buy and 42,000 ounces to sell. Subtraction gives 24,000 ounces of excess buying. The March 2025 IBA specification lists a standard imbalance threshold of 10,000 ounces for gold and also provides for a non-standard threshold. Assume the standard threshold applies throughout this example. With a 24,000-ounce difference the auction goes to another round at a revised price.

Suppose the next round ends with 53,000 ounces to buy and 47,000 to sell. The difference is now 6,000 ounces. That falls within the assumed tolerance. The final round can establish the benchmark even though buy and sell quantities before allocation differ. IBA describes the residual as shared equally among direct participants. The example supplies no participant count, so no allocation per participant can be calculated.

Source notes: [1] ICE Benchmark Administration — LBMA Precious Metals · [2] ICE Benchmark Administration — IBA Gold Auction Specification

Avoid counting the same interest twice

It is tempting to add 66,000 and 53,000 and describe 119,000 ounces of buying. Some or all of the second-round interest could come from participants revising orders they submitted in the first round. The first round failed to set the price. Its order totals cannot be appended to final-round trades as additional executed volume.

For the same reason, I would resist interpreting a shrinking imbalance as evidence of a particular seller entering. Buyers could have reduced quantities, sellers could have increased theirs, or both could have adjusted. Aggregate figures alone leave the underlying identities and decisions unresolved.

Record when the price became available

The gold auctions are scheduled to start at 10:30 and 15:00 London time. The benchmark follows the final round, whose time depends on how the auction proceeds. A research note should retain the AM or PM label, auction date, applicable threshold, final round and publication or receipt time when available. A chart stamped exactly at the scheduled start cannot by itself show what traders knew about the final benchmark.

Source notes: [1] ICE Benchmark Administration — LBMA Precious Metals · [2] ICE Benchmark Administration — IBA Gold Auction Specification