Nvidia’s Huang defends AI cloud firms as ‘good investment’

Nvidia ↘ 0.78% CEO Jensen Huang used the company's GTC Berlin conference to defend specialized AI cloud providers, often called neoclouds. He told Bloomberg's Tom Mackenzie that they are a "good investment" because "the demand of their compute is incredible," and said he is "quite confident" they will solve their current problems. A day earlier, he argued that Nvidia GPUs stay useful for eight years or more, long enough for lenders to treat them as loan collateral.cnbc.com+1

The comments come as shares and bonds of Nvidia-backed CoreWeave ↘ 2.00% and other neoclouds have dropped sharply. That has raised fears that companies borrowing billions to buy Nvidia chips may have trouble repaying their debts.tipranks.com

Nscale told to work around the clock

At a press briefing in Berlin, Huang said he had told UK provider Nscale, which Nvidia backs, to "work around the clock" to get its data centers running. Nscale has faced site delays, equipment breakdowns and power shortages, and investors have pushed back on the $30 billion valuation it is seeking in a planned IPO. "Their solution starts with the fact that they have excellent customers, demand is really high, and they just need working capital," Huang said. He noted that data centers once took three years to build, but "now, if you don't get it up and running in six months, everybody's up in arms".tipranks.com

The sector has had a rough few weeks. Australia's Firmus Grid dropped plans for a stock listing, and the PHLX Semiconductor Index ↘ 1.44% recently posted its longest losing streak since 2008. CNBC reported Huang as saying neocloud demand is "off the charts" but capital is scarce.cnbc.com+1

Chips as collateral

In an interview with Deirdre Bosa, Huang said that "what people thought used to be a four-year life cycle really is probably eight to maybe more." He said Nvidia keeps updating the software over the life of each system. "It's productive, it's fungible, and it's durable. Now, when that happens, you might be able to underwrite a loan against the asset itself," he said. Huang also said each gigawatt of AI computing costs about $50 billion and brings in about $40 billion a year in revenue.wccftech.com+1

His argument underpins a $500 billion financing pool that Nvidia set up in August with Apollo ↗ 3.38%, BlackRock ↗ 0.91%, Blackstone ↗ 1.93%, Brookfield ↗ 2.31%, Goldman Sachs The Goldman Sachs Group, Inc. ↗ 0.59% and KKR ↗ 2.50%.finance.yahoo.com

Lenders remain cautious

Wall Street has not fully accepted that view. Reuters reported that bankers want stronger guarantees than the residual-value support of up to 25% that Nvidia first offered. Banks typically value GPUs over three to four years, said Tony Trzcinka of Impax Asset Management. Even CoreWeave's $8.5 billion investment-grade GPU-backed loan got its rating largely because lenders count on contracted payments from Meta ↗ 2.72%.money.usnews.com+1