Why Is So Little Tokenized Gold Used as Collateral?

In brief: Programmable gold, meaning gold-backed digital tokens redeemable for allocated bullion, passed a live market stress test with its peg intact, yet less than 2% of the supply is pledged as collateral. The gap between a proven store of value and a working collateral asset is a plumbing problem, not a trust problem, and it is the part institutions should watch.

Programmable gold has cleared the bar that matters most to a reserve asset. According to a March 2026 tokenization report from RedStone, Credora, Gauntlet and Dune, gold-backed tokens tracked their underlying metal through a period of sharp market movement without breaking, while allocators rotated between commodity and Treasury exposure in a pattern that looked more like portfolio management than speculation. The catch, reported by Cointelegraph, is that under 2% of outstanding programmable gold is actually pledged against loans. The asset works. The market that should be using it has barely started.

That distinction is worth sitting with, because it inverts the usual objection. The standard institutional worry about a novel reserve asset is that it will fail under pressure: lose its peg, gate redemptions, or reveal a reserve shortfall at the worst moment. Programmable gold did the opposite. It behaved. What it has not yet done is become useful in the one role that separates a passive holding from a piece of financial infrastructure, serving as collateral you can borrow against, post to a counterparty, or compose into a structured position.

What is programmable gold, and how did it pass the stress test?

Programmable gold is a digital claim on physical bullion, one token to one fine troy ounce of a London Good Delivery bar held in an audited vault and redeemable by holders of size. The market is concentrated in two products. Pax Gold (PAXG), issued by the New York regulated trust company Paxos, and Tether Gold (XAUT) together account for roughly 89% of supply, per a December 2025 market survey. The sector as a whole grew about 177% in 2025, from roughly $1.6 billion to $4.4 billion, Cointelegraph reported, a stretch in which the spot gold price itself rose nearly 65% to records above $4,000 an ounce, per BullionVault.

The stress test was not a laboratory exercise. It was the ordinary business of a rallying, volatile market, and the relevant question for a collateral asset is narrow: did the token stay pinned to the metal when it mattered, and did the pricing that governs any loan against it stay honest around the clock? On both counts the answer held. Redemption remained open, the peg tracked, and because the underlying trades globally, programmable gold produced a continuous price signal even across weekends when traditional gold markets sit closed. For an asset whose entire purpose is to be trustworthy under duress, passing that test quietly is the strongest possible result.

So the reserve-asset case is largely settled. Programmable gold is auditable down to the bar, redeemable for the real thing, and durable through a real move in the market it references. The interesting problem is what happens next.

Why does less than 2% end up as collateral?

Because holding an asset and financing against it are two different acts, and the second one needs machinery the first does not. Most programmable gold today sits in wallets and on exchanges as a directional bet on the metal, the digital equivalent of a bar in a drawer. Using it as collateral means a lending venue has to accept it, price it continuously, and be able to seize and sell it cleanly if a borrower defaults. That last step is where the friction concentrates.

Gold is not a natural fit for the automatic liquidation logic that governs most programmable lending. Its price gaps when reference markets are shut, its depth is thinner than a large stablecoin's, and a forced sale into a stressed book can move the price against the very position being unwound. Lenders know this, so they either refuse the asset or apply haircuts steep enough to make borrowing against it unattractive. The 2% figure is the visible residue of that caution. It is not a verdict on gold's quality as backing. It is a verdict on the settlement and liquidation rails that would have to sit underneath it.

The rest of the real-world asset market makes the contrast sharp. RedStone's data show tokenized Treasury-bill deposits on one major lending venue falling 92% over a window in which programmable gold on the same venue grew sevenfold, a rotation that says allocators will move between programmable collateral types as the macro backdrop shifts, as the RedStone report describes. That behavior is exactly what you would expect from a functioning collateral market. It simply has not scaled into gold yet, because the plumbing that makes an asset safe to lend against lags the asset itself.

What would close the gap between holding and financing?

A settlement layer built for assets that cannot be dumped instantly. The clearest signal that the industry has diagnosed the problem correctly is that the oracle provider behind much of this data has responded not with more price feeds but with a liquidation product. RedStone launched a settlement layer, Settle, aimed squarely at the real-world asset collateral gap, Cointelegraph reported, and a subsequent account framed the target as roughly $30 billion in tokenized assets that could serve as collateral once liquidation is handled with more grace than a market order into a thin book.

The design point is that a redeemable, less-liquid asset needs a redemption-aware unwind, a process that can convert collateral in an orderly way rather than firing it into whatever bid happens to exist at the moment of stress. Get that right and the haircuts compress, the venues open up, and the same programmable gold that passed the peg test becomes something an institution can actually finance against, post to a counterparty, or embed in a structured trade. This is the composable half of the promise, and it is the half still under construction.

Regulators are moving in parallel. Through late 2025 the CFTC issued guidance welcoming tokenized assets as posted collateral, as several firms analyzing the advisories noted, which begins to answer the separate question of whether a supervised institution may treat programmable gold as eligible margin at all. Rails and rules are converging on the same conclusion from opposite directions.

For an institution reading the data, the takeaway is not that programmable gold is unproven. It is that the asset has already done the hard part, holding its value and its peg through a live test, while the market for using it as working collateral is still being assembled. The reserve question is answered. The collateral question is the one worth underwriting now, and the winners will be the issuers and infrastructure providers who make programmable gold not just auditable and redeemable but genuinely composable, safe to lend against and clean to unwind. That is the register Issuant works in, and it is where the next several billion dollars of this market will be decided.