I keep running into people who treat gold exposure as a single decision. They want gold, they buy whatever their broker or wallet puts in front of them, and they assume the only difference between the wrappers is branding. Then a year later they are annoyed their position lagged spot, or they get a tax bill they did not expect, or the collateral they thought was liquid is stuck behind a redemption process. The metal is the same in all three. Almost nothing else is.
So here is how I think about PAXG, GLD, and COMEX gold futures. They are optimized for different jobs, and the cost that dominates changes depending on how long you hold and where your money already lives.
What each one actually is under the hood
GLD is the boring, well understood answer. It is a physically backed gold ETF. You buy shares in your brokerage account, a custodian holds allocated bars in a vault, and the share price tracks a fraction of an ounce of gold minus the fund's ongoing fee. You are trusting a chain of institutions, the sponsor, the trustee, the custodian, but one that has been stress tested for a long time and trades on a normal exchange during normal hours.
PAXG is tokenized gold. Each token represents ownership of a specific quantity of physical gold held by the issuer, and it lives on a public blockchain, so it moves the way any ERC-20 token moves. You can hold it in your own wallet, send it peer to peer, post it as collateral in DeFi, and trade it around the clock. The counterparty is the issuer and their vaulting arrangement rather than an ETF sponsor. It behaves like crypto while pointing at metal.
Gold futures are a different animal. A COMEX contract is a leveraged, expiring agreement to take or make delivery of a set amount of gold at a future date. You post margin, not the full notional, so a small amount of cash controls a large position. That leverage is the whole point for a lot of traders, and it is also what quietly punishes people who wander in unprepared.
Where the cost actually comes from
People get this part backwards, because the sticker fee is not the real fee for any of them. With GLD you pay an expense ratio, a small annual percentage skimmed continuously from the fund. It compounds against you over a decade, but for most holders it is a rounding error relative to gold's moves. There is no roll to manage and no gas to pay. The hidden cost is mostly the bid-ask spread when you trade and being exposed during exchange hours only.
With PAXG the ongoing custody fee is typically low, sometimes zero at the token level, but the real costs are transactional. You pay network gas to move it, a spread on whatever venue you trade it, and depending on the issuer there can be creation or redemption fees to convert between token and physical or cash. Trade small size frequently on a high-fee chain and those costs can dwarf what GLD would have charged over the same period. Buy and hold in your own custody and PAXG can be cheaper than GLD on a pure carrying basis.
Futures are where the cost story gets genuinely interesting, because gold usually trades in contango, meaning the further-out contract is priced above the near one. To stay long past your contract's expiry you have to roll, selling the expiring contract and buying the next one, and in contango that roll typically costs you a little every time. Hold through futures for a year and you pay that roll drag repeatedly, which can add up to more than an ETF expense ratio, on top of the operational risk of letting a contract run to delivery by accident.
Custody, counterparty, and how they track
The risk profiles do not line up either. GLD hands custody to institutions and gives you a very liquid, tightly tracking instrument in exchange for trusting that chain and accepting market-hours-only trading. PAXG puts the keys in your hands, either the best or worst feature depending on your operational discipline, and shifts the trust to the token issuer's attestation and vaulting. Lose your keys and that is on you. If the issuer's backing is ever in question, that is a real counterparty exposure, so I care about who is attesting to the reserves.
Tracking behavior differs more than people expect. GLD hugs spot gold closely with tight spreads. PAXG usually tracks well too, but on thin venues or during crypto-specific stress it can wobble away from spot more than GLD would, because its liquidity is tied to crypto markets rather than the deep institutional gold market. Futures track spot through the basis, which is generally clean, but the leverage means a modest gold move becomes a large P&L swing, and a sharp move against you can trigger a margin call at the worst possible moment.
Taxes are the quietly decisive factor for a lot of people, and they are jurisdiction-specific, so this is a talk-to-your-accountant zone. In the United States, physically backed gold ETFs are generally treated as collectibles for long-term gains rather than getting the friendlier equity rate, and regulated futures get their own blended treatment regardless of holding period. Tokenized gold depends on how your jurisdiction classifies the token. Two positions with identical price performance can leave you with meaningfully different after-tax outcomes, so do not skip this step.
A decision table you can actually use
Here is how I match the wrapper to the person. Read it as a starting bias, not a rule, and adjust for your own size and jurisdiction.
- Long-term buy-and-hold in a brokerage account. GLD. You want tight tracking, deep liquidity, no roll to babysit, and you already live in that account. The expense ratio is the price of not thinking about it.
- Active leveraged trader with a real risk framework. Gold futures. If you are sizing positions deliberately, respecting margin, and rolling on purpose, the capital efficiency is worth the overhead. If you have no plan for the roll and for margin, this is not your wrapper.
- Crypto native who wants gold as portfolio collateral. PAXG. If your capital already lives on-chain and you want gold you can self-custody, move instantly, or post as collateral in DeFi, the token does something the other two structurally cannot. Just price in the gas and know your issuer.
- Gold parked alongside a crypto-heavy book without leaving that world. PAXG for the on-chain leg, plus a small GLD sleeve in a brokerage account if you want deeper market-hours liquidity as a counterweight.
The failure mode I see most often is a crypto trader who buys gold futures for leverage they did not need, gets margin-called into a loss on a metal that barely moved, and concludes gold is a bad hedge. Gold was fine. The wrapper was wrong for the job.
One practical habit before you commit. Write down your holding period, your account location, and whether you need the position to be collateral or just exposure. Those three answers usually pick the wrapper for you before fees even enter the conversation. If you already track your crypto and traditional positions together, seeing gold and your other legs on one scorecard, the kind of cross-market view I built into Blockcircle, makes it easier to catch when a wrapper is quietly costing you more than the metal is earning.