Ask the internet how to split your gold across jewellery, coins, sovereign bonds, ETFs and the newer tokenised options, and you will get a pie chart. Some confident percentage in each, presented as though there were a single correct answer that applies to everyone from a salaried thirty-year-old in Pune to a retiree in Toronto. There isn't. The pie chart is the wrong artefact, because it answers a question — "in what proportion?" — that cannot be answered without first asking the one that actually matters: what are you holding gold for, and whom are you willing to trust to hold it?
That second question is the key, and it is the thing almost no one says out loud. The five forms of gold look like substitutes — they are all "gold," they all roughly track the same price — so people treat the choice between them as a matter of convenience or cost. It isn't. They are not substitutes. They are complements that fail in completely different situations, because each one rests on a different assumption about who can be trusted to make good on it.
A gold allocation is a portfolio of trust
Here is the reframe that makes the whole thing legible. Every form of gold is a bet on a different party staying solvent, honest, and reachable. Sort them not by cost or convenience but by whom you are trusting, and the picture organises itself:
- Physical coins and bars — you trust no one. No counterparty, no intermediary, no system required for it to retain its value. It is the only form that survives an institutional or systemic failure intact, and the only one that passes to the next generation as a thing rather than a claim. The job it does is sovereignty and endurance. What it costs you: secure storage, insurance, a buy-sell spread, and no income. Its failure mode is mundane and personal — theft, loss, or an heir who never knew it existed.
- Jewellery — you trust no one, but this isn't really investment. It sits in your hand like bullion, but making charges, GST and weak buyback quietly strip away a large slice of its value the moment it becomes ornament. Its real job is use, culture and sentiment, with a gold core attached. Hold what you would wear or pass on for love; just don't count it as your gold allocation, and don't buy it as an investment, because as an investment it is the worst of the five.
- Sovereign Gold Bonds — you trust the state. Where available, they are the most efficient domestic form: they track the gold price, pay interest on top (2.5% in India), carry no storage cost, and offer tax-free capital gains if held to maturity. The job is efficiency for someone who trusts their government and doesn't need the metal in hand. The catches are real: issuance has been heavily curtailed in India, so availability is uncertain and must be checked; you are trusting the same sovereign whose currency gold partly exists to hedge against; and it passes on as a dematerialised claim requiring process, not as something you can simply hand over.
- Gold ETFs — you trust the market's plumbing. A fund, a custodian, a depository. In exchange you get the most liquid, lowest-friction, easiest-to-rebalance exposure there is — buy and sell in seconds, hold it in your demat account, pay a small annual fee. The job is liquidity and price exposure for the investor who is managing a portfolio rather than holding insurance. The failure mode is precisely the one that matters most: in a genuine system breakdown, an ETF is a claim on the very infrastructure that is breaking. It is excellent for everything except the scenario physical gold is for. (Taxation of gold funds and ETFs has changed in recent years; check the prevailing rules before relying on old assumptions.)
- Tokenised gold such as PAXG — you trust code and an issuer. Each token represents allocated physical gold held by a custodian; you can hold it in your own wallet, move it across borders, settle at any hour, and in principle redeem it for metal. The job is mobility — genuinely useful for someone global or cross-border. But you are now trusting the issuer's solvency and conduct, the custodian, the smart contract, and your own ability to keep a private key safe. For an Indian resident the tax treatment is punishing and the regulatory ground keeps shifting. And self-custody cuts both ways: the same key that makes it sovereign makes it disappear if the key is lost or dies with you — the digital version of an estate no one can find.
Read that list again and notice what it actually is. It is not five qualities of the same asset. It is five different answers to the question "who has to not fail for this gold to still be mine when I need it?" Trust no one. Trust no one, for love. Trust the state. Trust the plumbing. Trust the code. Those are not interchangeable. They are insurance against different catastrophes.
Which means the mix is a use case, not a number
Once you see the forms as trust assumptions, the proportion stops being a universal figure and becomes a function of who you are and what you are guarding against. A few archetypes show the framework doing its work — not as templates to copy, but as illustrations of how the reasoning produces different, equally sensible answers.
The crisis-hedger holds gold for the tail — the scenario where currencies wobble, institutions seize, or the system itself is in question. For this person, the paper and digital forms are nearly pointless, because they all depend on the very things the hedge exists to survive. The mix leans heavily, even overwhelmingly, toward physical bullion held under their own control. Liquidity and yield are luxuries they are happy to forgo, because the whole point is the asset that needs no one.
The efficiency-maximiser trusts the system they live in and treats gold as one more portfolio allocation to be held cheaply, rebalanced easily, and taxed favourably. They will weight toward sovereign bonds (where available) for the yield and tax treatment, and ETFs for the liquidity, and feel no need to hold metal at all. Their failure mode — total dependence on institutions — is one they have judged, reasonably, to be remote. They are optimising for the ninety-nine percent of the time the system works.
The global or mobile holder — the diaspora professional, the family split across countries — values the ability to move value across borders and hold it outside any single jurisdiction. For them a tokenised sleeve earns its place despite the risks, because mobility is the job that matters most. But the wise version of this person sizes it with eyes open: small relative to the whole, never the core, and with the key-custody and succession problem solved deliberately rather than discovered too late.
The legacy-holder is thinking past their own lifetime, and for them a hidden criterion dominates everything else: how cleanly does each form pass on? Physical gold, documented, transfers simplest. Bonds and ETFs transfer as claims that require a nominee, a process, and an heir who knows the account exists. Tokenised gold transfers only if the key does — and keys die with people. For this holder, the mix is shaped less by returns than by what their family will actually be able to find, reach, and identify when they are no longer there to explain it.
The real diversification is across failure modes
The deeper point, and the one the pie charts entirely miss, is this. People think of gold as diversification against their other assets. But within a gold holding there is a second, subtler diversification available, and it is the one that actually protects you: spreading across forms of trust, so that no single failure can take all of your gold at once.
If every gram you own depends on the same party — all physical and vulnerable to one loss, or all in one custodian, or all behind one key — then you have concentration risk dressed up as a gold allocation. Holding across forms means a system failure, a state default, a custodian collapse, and a lost key are four separate problems, not one.
That is the honest case for owning more than one form: not because some expert decreed a ratio, but because each form is strong exactly where another is weak, and the catastrophes they protect against do not arrive together.
So there is no number to give you, and anyone who hands you one is selling the comfort of false precision. The work is to decide, deliberately, which jobs you actually need gold to do — survive a crisis, sit efficiently in a portfolio, travel across borders, pass cleanly to your children — and to hold the form that does each job you care about, in whatever weight reflects how much you care about it. The right mix is simply the visible shape of your own answer to "whom do I trust, and what am I holding this against." Get that reasoning right and the proportions follow. Copy someone else's proportions and you have inherited their fears instead of examining your own — which, with the one asset people buy precisely to feel secure, rather defeats the point.