獅山巡禮
投資雜談
One Earnings Report Cant Erase the Stock Markets AI Bubble Concerns Dont fight the Fed is a financial market shibboleth most investors learn at a very early stage in their career.
The central bank might be wrong in its assessment of inflation pressures and the job market, but it makes the rules nonetheless. Push against it at your peril. Nvidia
NVDA
-3.15%
seems to occupy a similar space in the financial markets these days. As the worlds most-valuable company, selling a product that powers the global economys biggest investment boom, its become as definitive as the Fed in determining broader risk appetite.
So when CEO Jensen Huang says hes seeing something very different from an AI bubble, it doesnt pay to question his assessment.
After all, his thoughts about demand, and Nvidias better-than-expected third quarter earnings, are likely to boost the chip makers market value by $220 billion on Thursday alone. Thats more than that of McDonalds, and nearly as much as Goldman Sachs.
But is he right? Thats hard to say.
What Huang sees in terms of overall sales for his companys chips and processors says a lot more about the health of Nvidias income statement than it does about the future of the artificial-intelligence investment boom.
Investors arent worried about AI demand slowing down, as the companies at the heart of the infrastructure buildout have said they are going to spend billions over the coming years. And those company forecasts were nearly all firmly ahead of Wall Streets forecasts. What they do lose sleep over, however, is whether those massive commitments, some of which are now being funded by debt as opposed to cash flow, will generate an actual profit.
Nvidias success doesnt mean that valuations are appropriate across the board and that companies arent overspending on a technology that will take years (or decades) to prove its worth in terms of return on investment, said Chris Zaccarelli, CIO at Northlight Asset Management.
A secondary issue is whether companies like Microsoft
MSFT
-1.60%
, Amazon
AMZN
-2.49%
, and Google parent Alphabet are buying the right chips.
Nvidia plans to introduce a newer, more-efficient version of its AI-powering processors roughly every two years. Companies are projecting the useful (i.e. profitable) life of the chips they buy to be as long as six years.
Nvidia said software improvements can help keep older chips relevant, with Huang noting that the A100 GPUs we shipped six years ago are still running at full utilization today.
But the truth is that investors simply dont know if a data center humming with 2025s Blackwell chips in 2030 will be attractive to an AI customer when Nvidias 2030 processors are faster and possibly cheaper.
That changes the outlook for longer-term profits from the spending were seeing today, and, by extension, the value of the biggest stocks themselves.
And while we know that the biggest tech companies are committed to building hundreds of new data centers, we dont know how theyre going to connect to the real economy customers that are meant to use their massive computing power. MITs now infamous study of AI deployment, published over the summer, found that 95% of the U.S. companies it looked at were extracting no value at all from rolling out tests of the new technology.If that doesnt change, hyperscalers may find themselves sitting on hundreds of gigawatts of computing power with no useful purposeechoing the dark fiber era of the early 2000s that followed the dot-com collapse.
Nvidia hasnt proposed a way out of that conundrum, either.
The issue of circular financing in the AI space, highlighted by Nvidias investments in unprofitable companiessuch as OpenAI and Anthropicas well as start-ups such as Mistral and Reflectionalso remains unresolved. Nvidia itself said there is no assurance that current deals will be completed on expected terms.
None of this will matter to Nvidia, which will continue to lead markets higher into the end of the year and beyond. Its an incredible business with a iconic leader who gets a great deal of support from the White House.
But there wasnt much benefit in tracking sales of shovels, pick axes, and sifting pans in order to validate the bull case for the California Gold Rush. And assuming that Nvidia sales will cement the case for the the AI revolution might be similarly flawed.
