• Markets initially rallied following Nvidia’s earning report, but has since lost momentum
• We think that AI bubble conditions are forming, but a peak is not yet in sight
• Investors should regard AI rallies as opportunities to diversify further

Better than expected results

In an earnings report that has rarely been as eagerly or widely awaited, Nvidia announced on Wednesday night that its earnings were significantly up compared to the same period last year. The company’s 3Q revenues and net profits increased by 62 percent and 65 percent respectively, and 4Q sales are forecast to reach around US$65 billion.

These results beat analyst forecasts and having faltered in previous days, markets initially responded positively. However, the relief rally soon lost momentum. Japan’s Nikkei 225 and South Korea’s KOSPI both opened lower on Friday, mirroring overnight declines in US markets amid renewed concerns over valuations of AI-related stocks.

An indicator of AI health

Nvidia is the world’s largest company. Its share price has grown by over 1,200 percent over the past five years, bringing the company’s market cap to an unprecedented US$5.76 trillion. This is higher than the GDP of the world’s third largest economy, Germany, so it is no exaggeration to say that Nvidia is a driver of global growth.

But it is not just Nvidia’s size that is generating so much interest. The company is also seen as a gauge of the AI industry’s health at a time of mounting anxiety about AI overhype. To justify the high valuations, analysts are looking to Nvidia to help provide clues on how successfully the top three industry challenges are being addressed:

1. Overconcentration
   The bulk of Nvidia’s revenues currently comes from US hyperscalers i.e. large cloud providers like Amazon, Microsoft and Google. In fact, about half of their total AI infrastructure spending in recent years has gone to Nvidia. The company’s results point to a continuation of this trend, with data centre demand driving much of 3Q’s revenue growth.
   
   While Nvidia has said that it expects as much as US$3 – 4 trillion in capex spending through to 2030, there appears to be a high level of interdependence amongst a small number of players. Going forward, “AI capex fatigue” within this very limited group could severely impact the industry as a whole.
   
   These same tech giants are also behind a web of deeply interconnected and circular deals. Such investments involve joint holdings of new ventures which in turn become their customers. Their rapid rise has fuelled concerns that AI revenues and profits are being artificially inflated. High profile analyst, Michael Burry, is not the only one to suggest that data center profitability could be due to creative accounting and that debts arising from capex spending are not fully transparent.
   
   
2. Physical constraints
   There are ongoing fears that data center capacity, power and water supplies will not be able to scale up as much or as quickly as needed. Consultants Bain and Co estimate that the demand for data centre capacity and electricity will double over the next five years as AI usage becomes more widespread.
   
   Bain warns that this dramatic increase in power and water supply will require unconventional approaches and coordinated actions from utility companies and regulators. The construction of data centers also comes with multiple hurdles. At the present time, it remains uncertain whether the current physical infrastructure is able to keep up with AI demand and innovation. Despite Nvidia’s strong earnings report, power and capacity constraints will continue to threaten the AI industry’s growth trajectory.
   
   
3. Enterprise application
   Perhaps most importantly, analysts are searching for indications of broadening enterprise adoption of AI. At current valuations, analysts expect enterprises to view AI as a “survival imperative”.
   
   This is not yet the case, with nearly two thirds of respondents in a recent McKinsey survey saying that they have not yet scaled AI across their entire enterprise. Nevertheless, Nvidia’s 3Q sales of its new Blackwell chips – described by the company as “off the charts” – will give hope that this is starting to turn. Increased demand for Blackwell chips’ ability to deliver high performance computing is a sign that enterprises are starting to shift from pilot AI projects to large scale AI deployments.
   
   Similarly, Nvidia saw good earnings growth for its full stack AI solutions across a variety of industries and regions, prompting the company to comment that, “AI is simultaneously reaching every corner of the globe, empowering everything.” While this will help to allay some fears, we remain on the lookout for signs that AI is permeating into everyday processes, especially within smaller, non-tech companies.

In general, the latest Nvidia results reinforces our view that despite signs of frothy valuations and increased investor nervousness, it is still too early to jump the AI ship. As pointed out in an earlier article, the 1990s proved that technology bubbles can last several years before panic sets in.

That said, investors should seek to diversify their equity portfolio across sectors and regions. The US economy is disproportionately dependent on just one sector and there is a real need for US (and global) growth to broaden out beyond the tech giants.

On a more micro level, Nvidia’s impressive results raises the risk of more AI-related companies enjoying a rapid but unsustainable rise in investor interest. Careful stockpicking and close monitoring of the AI industry will be more important than ever.