Can anything stop NVIDIA?

Our company is 30 days from going out of business – The motto of Nvidia’s CEO.

The warning sticks with Jensen Huang from more precarious times. He recalls several instances in the technology company’s 30-year history when it could have gone under. As its shares go gangbusters, Huang is wary of complacency in a market his firm now dominates.     

Last week’s earnings reports defied already sky-high expectations. Nvidia’s revenue of $22.10bn was almost 10 percent higher than forecast. Its market value touched $2tn, consolidating its position as one of the world’s 5 most valuable companies. A constituent of the so-called Magnificent 7 (alongside Apple, Amazon, Alphabet, Meta, Microsoft and Tesla), these stocks now comprise about 30 percent of the S & P 500. The index’s most top-heavy weighting since 1929 may concern historically-minded investors.

Why is Nvidia so successful?

Nvidia is the biggest stock winner of the generative AI boom to date. It’s the chief supplier of the computer chips essential for training models like OpenAI’s ChatGPT. Nvidia currently controls about 80 percent of the market in AI data centres operated by Amazon, Alphabet and Microsoft.

So why the caution in the midst of such triumph? Because Huang is well-aware that big risks remain:

  1. Competitors: First movers don’t always guarantee enduring advantage. Think MySpace, Nokia or Kodak. Nvidia has a target on its back and deep-pocketed competitors are coming.
  2. Regulation: New industries get new rules. Add in geopolitical tensions and the future of the chip market looks less laissez-faire.
  3. AI hype cycle: Are Large Language Models (LLMs) all they’re cracked up to be? If the buzz dissipates, so will Nvidia’s sales.

we have no moat

Said a Google executive in a leaked memo. He or she argued that the tech giant could not build any entrenched advantage in the AI space. Rob Arnott says it emphasises the ultra-competitive nature of the industry. He doubts that Nvidia will be “massively successful in the decade ahead.” While the prevalent narrative suggests Nvidia has a “moat” to produce the fast chips demanded, rivals are already catching up. Lisa Su, CEO of Advanced Micro Devices (AMD), recently flaunted AMD’s new M1300 chip. It’s said to outshine Nvidia’s popular H100 model.

Nvidia’s new H200 upgrade will be released later this year. That may regain any lost ground. But it also faces a challenge from stock market rivals and current clients. Its fellow members of the Magnificent 7 want to wean themselves off Nvidia’s supply chain and are working on developing their own products in-house.

Finally, a blast from the past threatens Nvidia’s market share. Intel’s market cap was once $497.6bn bigger than Nvidia’s. It’s now $1.78tn behind. CEO Patrick Gelsinger has a strategy to bridge that gap. He believes Intel will benefit from the next stage of LLMs: deployment. Running these models to churn out text and images requires less powerful and expensive chips. Intel specialises in these simpler products.

Looming AI regulation

Nvidia is benefitting from the great LLM rush. It’s motivated as much by fear as enthusiasm. No CEO can risk being caught with their pants down. Regulation may change that equation.

Things are a little quieter on this front since Rishi Sunak’s first big showpiece summit. It’s also not the sexiest topic for geriatric candidates on the US campaign trail. But it rumbles on nonetheless. Senator Scott Wiener introduced a bill earlier this month to introduce more onerous testing requirements for new LLM models. Increasing concerns around copyright of learning materials and litigious implications of chatbots’ ongoing hallucinations are other financial considerations for model developers.

Things are more gung-ho in China. Angela Zhang’s upcoming book High Wire argues the CCP will put similar concerns aside as it seeks to spur development at all costs. But Nvidia is already restricted from selling certain chips to Chinese firms. Should a regulatory gap open between the US and China, that embargo will likely tighten.

AI Hype

In the meantime, subdued expectations around generative AI may slow Nvidia’s rise. Arnott warns fellow investors about “valuations getting out of hand, based on speculation that the market will grow big faster than it actually does.” He uses the example of Qualcomm during the dotcom bubble. In 1999 it was the world’s hottest stock, far outshining Nvidia now. And it’s been hugely successful over the last 25 years with profits rising 60-fold.

But Qualcomm buyers would have been better off parking their money in an S&P 500 index fund. The internet was transformative but only gradually. Qualcomm could never live up to those inflated expectations.

Similarly, Nvidia may fall foul of hyperbole. Prominent sceptic Gary Marcus points out that we are starting to see LLMs’ capabilities plateau. No model has definitively beaten Open AI’s GPT-4, available over a year ago. Scale is not yet producing better results and is a far cry from the exponential growth once predicted. Even prominent optimists like Demis Hassabis and Bill Gates have acknowledged this in the last few weeks. 

Where are the NVIDIA bears?

Jim Cramer asks as the stock’s surge continues. Sceptics have only been stung thus far. Other constituents of the AI bull market stalled last June but Nvidia goes from strength to strength.

Few doubt that it’s incredibly well-placed to dominate a new AI industry. But there are valid questions as to how such a nascent environment will change. Furthermore, Nvidia’s valuation owes much to a wider bullish outlook. Should this enthusiasm dampen, so will sentiment around its biggest supplier.

And if that narrative subsides, the bull run will slow. On Monday hedge funds cashed out of tech stocks, selling at the fastest price in 7 months.

0Shares

Leave a Reply

Your email address will not be published. Required fields are marked *