Analysis
Four years into the generative AI boom, AMD and Google have closed enough of the gap that Nvidia's technology lead is no longer the whole story, CNBC reported. What Nvidia has that neither competitor can match is a balance sheet, and the company has started using it as the primary competitive instrument. A separate Bloomberg analysis reached a similar conclusion about the shift from silicon to capital.
The evidence is a week old. Nvidia struck a pact with Wall Street firms to pursue $500 billion of financing for its GPUs, then announced Monday it would provide up to $105 billion in support for OpenAI's Ohio data center -- a backstop if the ChatGPT maker's fortunes turn. Pulse covered the guarantee's shrinking size separately; the strategic point here is different. Demand for AI infrastructure appears limitless, but a handful of hyperscalers do the buying, and Nvidia is removing financing friction so those buyers never pause.
Where the Cash Is Going
The capacity to do this is extraordinary. Quarterly free cash flow reached $48.5 billion in the latest period, up 18-fold in three years, after twelve consecutive quarters of revenue growth above 55%. The balance-sheet moves stacking on top of that cash generation:
- Dividend -- raised from a penny to 25 cents quarterly in May
- Buyback -- $80 billion authorized, with roughly half of free cash flow pledged to shareholders this year
- Marketable equity securities -- grew to $30.2 billion, up from $12.9 billion a year prior
- OpenAI investment -- $30 billion committed in February
- SB Energy investment -- $1.5 billion committed Monday
"They remain dominant, but they're very paranoid about making sure they don't lose ground," said Ram Bala of Santa Clara University's Leavey School of Business. Cantor analysts pushed back on the circularity framing in a Monday client note, reiterating a buy and calling the OpenAI agreement a "clear signal that the current AI investment cycle will be elongated and durable" -- "less circular and more facilitating the coming AI buildout while at the same time creating additional competitive moats."
The bear case is the mirror image of the bull case, using identical facts. A supplier that finances its customers converts revenue quality into credit exposure. Every dollar of demand created by a Nvidia guarantee is a dollar that would not exist at a market clearing price for capital, and the accounting for it lives in contingent obligations rather than the income statement. Investors modeling Nvidia now have to underwrite a lender's book alongside a chip business, and the two carry different risk profiles and deserve different multiples.
AMD and Google are the specific pressure that prompted this. AMD's MI-series accelerators have won real hyperscaler commitments and the company has done 19 private financings this year alongside its own share strength, while Google's TPU line gives it an internal alternative that removes a large buyer from Nvidia's addressable market entirely. Neither has caught Nvidia on raw performance across the full training and inference stack, but the gap is narrow enough that price and availability now decide some deals -- which is exactly the condition under which financing terms become the tiebreaker.
Historical analogues for vendor financing are not encouraging, and investors know them. Lucent and Nortel financed customer purchases into the telecom buildout of the late 1990s, booked the revenue, and absorbed the losses when those customers failed. The structural difference is solvency: Nvidia is generating $48.5 billion of quarterly free cash flow and financing hyperscalers and well-capitalized labs, not speculative competitive local exchange carriers. The similarity is the mechanism, and the honest position is that the mechanism is identical while the counterparty quality is far better.
For competitors, the implication is bleaker than any benchmark result. AMD can match Nvidia on a chip. It cannot guarantee $105 billion of someone else's leases.
