Gold Rush Drives $29 Billion Monthly Influx to US

In early 2025, a significant event unfolded in the global gold market, driven by a peculiar arbitrage opportunity that saw a massive influx of gold bullion into the United States. The story began with the transportation of two-ton pallets of 99.5% pure gold from London to Zurich, and subsequently to New York. The gold, initially cast in 400-ounce formats, had to be recast into 100-ounce or kilobar form to meet the specifications of the COMEX vault in New York.

This process involved melting down the gold in Swiss furnaces and reshaping it, triggering new customs declarations each time. The full market value of the gold was recorded at every checkpoint, accumulating significant dollar signs. Traders capitalized on the price wedge between COMEX futures and London spot, which stood at $40 to $50, enough to cover refinery costs and freight while still locking in profits.

Within weeks, these shipments swelled to an astonishing $29 billion a month, a scale that economists at the Atlanta Fed admitted they had never seen in three and a half decades of trade data. The arbitrage opportunity was fueled by traders front-running President Trump’s mooted tariff barrage, which created a juicy futures-versus-spot arbitrage. Traders could buy cheaper London metal, pay Swiss refiners to recast it, and still pocket profits once the bars were eligible for COMEX delivery.

However, once the White House formally exempted precious metals on 3 April, the COMEX–London premium collapsed to $20/oz, and the incentive to keep air-freighting bullion vanished. Meanwhile, the Atlanta Fed’s GDPNow model, which is updated hours after every data release, suddenly skidded from modest-growth territory to a recession-screaming -3.1% in late February. The model was duped by the bullion bonanza, as gold bars are classified by the Bureau of Economic Analysis as “non-monetary gold.” Purchases count as imports, which are subtracted from GDP, even though the metal often sits inert in vaults rather than coursing through factories.

The January–February spike left gross imports $22 billion above the Q4 average. Annualised, that gap tops $265 billion. The Fed’s Pat Higgins wrote that this was enough to hit the GDPNow print by 3.6 percentage points. On 6 March, the Atlanta team bolted a “gold-adjustment” onto the codebase, literally yanking bullion flows out of the net-exports equation. “The model is forecasting smaller, but still slightly negative, first-quarter real GDP growth,” Higgins explained in an internal blog post as he promised to replace the old version on 30 April.

In one stroke, GDPNow lurched from doom-laden 2-ish prints to a far tamer 0.1 percent, a 250-basis-point facelift with the click of a Git commit. The first estimate for Q1 GDP eventually came out at 0.3% and was later revised to 0.2%. GDPNow’s forecast for Q2 now sits at a much healthier 2% using the new gold-adjusted model.

The BEA itself was not fooled, as the official advance estimate showed that Q1 GDP fell only 0.3%, which is hardly catastrophic because statisticians have already stripped “valuables” like gold and silver from domestic investment. Imports still clobbered growth, subtracting almost five full percentage points, but that drag was partly optical, a ledger quirk rather than a real-economy crash. Higgins conceded that inventory data is patchy for the farm and utilities sectors, so the first print could be revised once those beans are counted.

Looking forward, the same trade-war jitters that drove bullion stateside remain unresolved, and Higgins warns the absence of another gold wave could whipsaw Q2 nowcasts in the opposite direction. Should bullion flows normalise, GDPNow might overstate growth as imports retreat. Conversely, a fresh premium could again punch the model below the waterline. Either way, the Atlanta Fed’s willingness to hot-patch its algorithm highlights a larger lesson: data science is only as good as the metadata you feed it.