Five buyers now owe Nvidia $44bn of the demand its guidance promised

A receivable is revenue the auditors have blessed and the treasury has not banked, so the $24.6bn Nvidia added to that line between January and July ^[10] is a different animal from the outlook the stock rallied on. The filing rounds it to 63% ^[1]; the ratio is 63.9% ^[11]. What makes the six-month clip awkward is that the sell-side models tracking the line assume it does not repeat.

Bank of America's $71bn for January 2027 ^[4] sits only $7.9bn above July's $63.1bn, about a third of the $24.6bn added in the half before it ^[14]. Morgan Stanley is more generous at $78.6bn, which is $15.5bn of fresh unpaid invoices over the same six months ^[15] and $7.6bn above BofA's figure ^[16]. Out at January 2029 the two houses are $24bn apart, $171bn against $147bn ^[17], which is, near enough, this half's entire receivable increase being argued about as a modelling preference three years out ^[10]. Both still roughly double the balance from 2027, BofA by around 107% and Morgan Stanley by around 117% ^[4][5].

Seventy per cent of $63.1bn is $44.2bn ^[12], an average of $8.8bn owed by each of the five direct customers Nvidia disclosed, most of it large cloud providers ^[13][2]. A year earlier the same disclosure read three customers at 56% ^[3], and the honest caveat is that the source gives no year-ago receivable base, so that 56% cannot be converted into dollars here. More names, larger share.

This is probably wrong, but the version of the thesis I would defend rests on slow cash, not bad credit: the demand signal buyers paid 9% up for ^[18] is increasingly carried on the seller's balance sheet. There is a straightforward counterargument to this. Receivables scale with revenue, and five hyperscale counterparties are close to the best corporate credit available. If revenue also grew 63.9% across the half, days sales outstanding did not move, and there is nothing here ^[11]. The source does not print the revenue line, so that ratio is not derivable from this material, and a DSO quoted off these two numbers alone is a guess.

The thesis breaks if the next quarter shows receivables growing slower than revenue. It breaks more quietly if the concentration disclosure drifts back toward three names ^[3].

Gil Luria, tech research head at D.A. Davidson, told CNBC the line is "worth keeping an eye on", adding that "the numbers are so big and they're making really big commitments way out into the future" ^[6]. Note what the cash is doing meanwhile, or rather what management is choosing to fund while $63.1bn sits uncollected: a reported $12.9bn for Hugging Face, sourced to one person with direct knowledge, with Business Insider putting it only at "in talks" ^[20], plus engineering time spent optimising for DeepSeek V4 Flash, Alibaba's Qwen 3.8 and Google's models ^[7] even as the company tells the SEC that White House rules aimed at Chinese-origin AI could hurt parts of the business ^[8]. Money is flowing toward developer distribution while invoices pile up with five clouds.