We have two broker swap sheets open on the desk. One is dated 14 October 2026, from a DFSA-licensed operator on the ADGM-adjacent side of the market. The other is timestamped 03 November 2026, from a Seychelles-regulated book routing Gulf retail flow. The same instrument — XAU/USD — shows a short-side overnight swap of -4.82 on one sheet and -6.17 on the other. Neither broker publishes the arithmetic. Neither names the counterparty bank quoting the tom-next rate that sits underneath. Every explainer we have read on the question of who sets these numbers resolves it in two or three paragraphs and leaves the chain unexamined. The chain has four parties in it.
What They All Get Wrong
The repeated mistake is a single-authority fallacy. Half the explainers tell the retail reader that the broker sets the swap rate. The other half upgrade the authority and say the central bank sets it, through the overnight policy rate differential between the two currencies in the pair. Both answers are wrong in the same way. They collapse a layered price into a single decision-maker.
Run the test from our opening against either frame and the collapse is visible. If the broker sets the number unilaterally, there is no explanation for why both of those XAU/USD short-side swaps moved in the same direction after the Federal Open Market Committee statement on 18 September 2026. Retail books do not coordinate. They respond to the same upstream signal. The signal is not sitting inside either broker's risk desk. If, conversely, the central bank sets it through pure rate differential, the two sheets should quote nearly identical numbers for the same instrument on the same date — bar a thin pass-through margin. They do not. The spread between the two short-side numbers on XAU/USD is 1.35 in swap units. That is not a rounding artifact. It is a different funding stack on each sheet.
There is a second, quieter error that shows up in the better-researched pieces. They describe the mechanic for FX pairs — long the higher-yielding currency, receive the differential; short it, pay the differential — and then silently extend that logic to gold and silver and oil. The extension does not hold. XAU/USD is not a currency pair in the funding sense. A short position in XAU/USD is a long-dollar, short-gold position for carry purposes. The relevant inputs are the US dollar overnight rate, the gold lease rate (the cost or yield on borrowing physical gold for a term), and the storage and insurance embedded in loco London quoting. Nothing in the standard FX swap explainer prepares the reader for a commodity carry structure.
The third thing conventional coverage flattens is the Islamic account question. Swap-free is presented as a flat trade-off — the broker removes the overnight charge, the client accepts an administration fee on positions held past some window. The mechanism underneath is less tidy. The broker is still rolling the position through an interbank market that charges funding. Someone is absorbing the carry. Either the broker is wearing it as a cost of acquisition for Gulf retail, or it is rebuilt into a wider spread or into the administration fee itself. The explainer that treats swap-free as a simple on-off switch omits the fact that the chain under the position has not changed — only the invoicing has.
What Is Almost Always Missing
Four layers go unnamed in nearly every public explainer, and all four of them are where the number actually gets made.
The first missing layer is the tom-next funding leg. Spot FX and spot gold both settle T+2. A position held past the New York close at 17:00 ET has to be rolled from today's value date to tomorrow's. That roll is quoted as a tom-next rate in the interbank market by a small group of tier-one banks — the same desks that print the LBMA AM and PM fix for gold. The retail broker does not call the central bank. It takes the tom-next quote off its prime broker's pricing feed, which is itself a dressed version of what the top-of-book banks are showing. This is where the overnight number is actually born as a market price. The policy rate differential sits upstream of it, as an input, but the tom-next is the price that gets passed down.
The second missing layer is the prime broker markup. A retail broker clearing Gulf flow does not have a direct seat at the LBMA bullion fix. It has a credit line with a prime broker — typically a tier-one European or US bank, or a non-bank prime. The prime broker shows the retail broker a funding rate that is already marked up from the interbank tom-next by two to five basis points per side, depending on the credit relationship and the notional cleared. This markup is real money and it is invisible to the client.
The third missing layer is the retail broker's own markup on top of the prime broker feed. This is where the dispersion between our two sheets comes from. The DFSA-licensed operator and the Seychelles-regulated book are almost certainly pricing from similar prime broker feeds — the pool of major banks willing to warehouse gold funding risk for retail brokers is small. What differs is the retail markup each book applies to that feed. Some books mark both sides symmetrically. Some mark the direction their client base typically holds — overweight the long-side markup for a client base that holds trend-following positions in gold. This decision is made internally, is not disclosed, and is not standard across the industry.
The fourth missing layer is regulator silence. Read a conduct-of-business sourcebook from the Dubai Financial Services Authority and you will find general obligations to disclose all costs and charges to retail clients. Read the Financial Services Authority of Seychelles equivalent and you will find a thinner but still existent cost-disclosure principle. Then read the actual broker terms-of-service documents those regulators have approved. The swap rate is almost always framed as "variable, subject to change by the broker at its discretion, reflecting interbank conditions." That phrasing is operative. It is also the opposite of a real disclosure. Two primary documents — the regulator's conduct rule and the broker's approved TOS — point in different directions, and both are live.
What I Would Say Instead
Here is the chain, named end to end, with the math shown.
The first party is the central bank — plural, actually, because every pair has two sides. For XAU/USD the dollar side is the Federal Reserve, which in late 2026 is quoting an effective federal funds rate in a band that puts overnight dollar funding at roughly 4.50 percent annualized. The gold side has no central bank. It has a lease rate quoted by the bullion banks against the GOFO reference — the gold forward offered rate convention revived in internal bullion-bank books after the LBMA stopped publishing GOFO in 2015. This rate has been hovering in a band around 0.25 to 0.40 percent for one-month tenors in recent months. The policy-rate differential for a short XAU/USD position is therefore roughly 4.50 minus 0.30, or 4.20 percent annualized, in favor of the short side. Policy differential wants to pay the short holder.
The second party is the interbank tom-next market. The tier-one banks that quote tom-next add a funding adjustment that reflects balance-sheet constraints on month-end, quarter-end, and year-end rolls. In quiet mid-month conditions this adjustment is thin — one to three basis points off the pure differential. On the last business day of a quarter it can widen by a factor of five. Our 14 October and 03 November sheets are both away from quarter-end, so the interbank layer is thin. Call it 3 basis points off the differential, taking the gross funding closer to 4.17 percent.
The third party is the prime broker. The retail-facing book is clearing through a credit line with a bank prime or non-bank prime, and that prime is marking the tom-next by somewhere between 2 and 5 basis points per side. For a short XAU/USD position the prime takes the markup out of the funding the retail broker receives. Call it 4 basis points of markup against the retail book. The funding received by the retail broker is now roughly 4.13 percent annualized.
The fourth party is the retail broker itself. On a $265,000 notional short position — 1 standard lot of XAU/USD at $2,650 per ounce, 100 ounces — the theoretical funding received at 4.13 percent is $265,000 × 0.0413 ÷ 360, or roughly $30.40 per day. The sheets we are holding show the short-side swap as -4.82 and -6.17 in swap units. The direction of the number is the forensic tell. A short XAU/USD position on the pure chain should pay the holder, not charge the holder. The negative sign is the retail broker's markup absorbing the funding income and converting it into a cost. The size of that conversion — approximately $35 of markup per lot per day on the first sheet, approximately $37 on the second — is the entire reason retail traders carrying overnight gold positions lose money to the roll even when the rate math is in their favor.
On swap-free accounts, this same $35 to $37 does not disappear. It is rebuilt into an administration fee, usually charged flat per lot after a tolerance window of three to seven days. The chain is identical. The invoice is cosmetic. Two broker swap sheets, same instrument, same currency pair, same month: -4.82 and -6.17.
FAQ
Why do two brokers quote different swap rates on the same pair on the same day?
Because the retail markup layer is set internally by each broker and is not standardized. Both books likely price off similar prime broker feeds — the pool of banks willing to warehouse gold funding risk for retail is small — but each retail broker decides unilaterally how much markup to add. On the two sheets we examined the markup gap was roughly $2 per lot per day, which is $730 per lot per year held continuously. That gap is unexplained in either broker's public documentation.
Does the central bank actually set swap rates?
No. The central bank sets the policy rate that becomes the first input into the chain. For a pair like EUR/USD that policy-rate differential is passed through the interbank tom-next market, marked by the prime broker, marked again by the retail broker, and arrives at the client terminal several layers downstream. For XAU/USD the dollar side comes from the Federal Reserve policy band, but the gold side comes from a bullion-bank lease rate convention, not a central bank at all.
What is tom-next and why does it matter here?
Tom-next is the one-day forward swap used by tier-one banks to roll a spot position from today's value date to tomorrow's. It is the price at which a spot FX or spot gold position is actually funded overnight in the interbank market. Every retail swap number is derived from a tom-next quote the broker sees through its prime broker pricing feed. The retail swap is not the tom-next — it is the tom-next plus two layers of markup.
Are Islamic swap-free accounts actually free of the swap cost?
Not in the economic sense. The chain underneath the position still charges funding — the prime broker is still rolling the exposure through tom-next. What changes on a swap-free account is the invoicing. The broker typically removes the overnight charge for a tolerance window of three to seven days and then substitutes a flat administration fee per lot or per position. Over a long hold, the administration fee usually recovers most or all of the original swap cost.
Does any regulator require brokers to disclose the markup chain?
No regulator relevant to Gulf retail flow — DFSA, ADGM FSRA, SCA UAE, or the Financial Services Authority of Seychelles — requires a retail broker to disclose the prime broker it clears through or the markup applied on top of the prime's funding feed. The conduct rules require disclosure of costs in general terms, and the broker TOS typically satisfies this with language calling swap rates "variable, reflecting interbank conditions." That language is approved by the regulator and is not a disclosure of the arithmetic.
How does the DGCX 995 contract relate to spot broker swaps?
The DGCX 995 gold futures contract sets its carry structure through the futures curve, which embeds funding and storage in the contract term. It does not directly feed the spot broker swap number, but it is a reliable sanity check. If the DGCX curve is pricing a given monthly carry and the retail broker's cumulative spot swap over the same month is materially wider, the gap is the retail broker's markup made visible against a listed, transparent reference.
Why is the short-side swap on XAU/USD often negative when the math says it should be positive?
Because the retail broker's markup is wider than the funding income the short-side position theoretically earns. On a 1-lot short XAU/USD at $2,650 per ounce with roughly 4.13 percent annualized funding after prime broker markup, the daily funding received is near $30. If the retail broker adds a markup of $35 against that position on the roll, the net quote to the client flips sign. The client sees a negative short-side swap and is told "this reflects interbank conditions." The interbank layer is contributing a positive number. The retail layer is where the negative comes from.
Where does the number actually get made?
At the retail broker's risk desk, on top of a prime broker feed that is already a marked version of the interbank tom-next, which is itself a funding adjustment over the policy-rate differential between two central banks or between a central bank and a bullion lease rate. The chain has four parties in it. The number on the client terminal is the sum of four separate decisions, only one of which is visible in any broker's public documentation.