Eric Schmidt, CEO of Google until 2011, is many things, but one thing he is not: a financial advisor with a license to provide investment advice. If he were, however, he made it quite clear that the provider of AI training chips NVIDIA is a stock that he believes belongs in every portfolio.
In a speech to students at Stanford University this week, the billionaire angel investor argued that the race to commercialize generative artificial intelligence and other forms of AI is still in its early stages. While the quick bucks may have already been made, he believes there is still plenty of upside in the space.
This is especially true for Nvidia, whose lead in graphics processing units (GPUs) such as the benchmark H100, combined with its dominant CUDA software ecosystem, gives the company a huge lead as the preferred supplier of AI training chips for data centers.
“The sums of money being thrown around are unbelievable,” Schmidt told the students. “I talk to the big companies, and the big companies tell me they need $10 billion, $20 billion, $50 billion, $100 billion.”
Schmidt added that Sam Altman, CEO of OpenAI, even estimates that it will take $300 billion to develop artificial general intelligence, a machine capable of thinking independently.
“If $300 billion goes entirely to Nvidia, you know what to do on the stock market,” Schmidt continued, quickly adding, “This is not a stock recommendation, I don’t have a license.”
AMD cannot yet compete with Nvidia’s dominant CUDA software
Nvidia shares have been the main beneficiary of the AI gold rush, with the company’s share price nearly tripling this year to reach $135 in mid-July. This is due in part to debt-financed “carry” trades, a Momentum game Investors borrow cheap Japanese yen to invest in higher-yielding US growth stocks denominated in dollars.
Due to the recent sharp rise in the Japanese currency against the greenback, shares have fallen from their highs when the company was briefly more valuable than any other company in the world, with a market capitalization of $3.3 trillion.
However, Nvidia still has a chance to reach new highs if it beats earnings expectations when it releases quarterly results later this month.
Based on Schmidt’s conversations with his contacts at major technology companies, he expects demand for his AI training chips to continue to grow robustly, if not exponentially.
Referring to Percy Liang, an AI researcher at Stanford who had to resort to Google’s Tensor Processors (TPUs) due to the lack of availability of Nvidia chips, Schmidt said: “If he had infinite money, he would choose the B200 architecture.”
The B200 is Nvidia’s Next generation training chipIt is so advanced that even its packaging must be done under the controlled conditions of a clean room, not just the production of the wafer itself, says Schmidt.
Lisa Su’s AMD may one day catch up in terms of hardware, but her company’s software ecosystem does not have Nvidia’s user base, and a compiler that translates CUDA into AMD’s own programming language ROCm and helps developers do so does not yet work.
Neither Nvidia nor AMD were immediately available for comment.
AI skeptic Ken Griffin reduces hedge fund Citadel’s stake in Nvidia
Stanford removed the video after Schmidt was criticized for accusing his former employer of squandering its lead in the field of AI. lax work ethic The company allowed its employees to work from home instead of in the office.
In 2017, Google researchers invented the so-called transformer neural network that drives most GenAI models, only to then “put it on the shelf,” as Marc Andreessen put it last month. The name of OpenAI’s own ChatGPT — Generative pre-trained transformer – which was launched in late 2022, reveals its Google lineage.
Schmidt did not comment on the time horizon he has for his investment recommendations, but it is safe to say that as an angel investor who finances various startups, his time horizon is longer than the average on Wall Street.
This was made clear, for example, in a filing by Citadel this week. Ken Griffin’s hedge fund has increased its holdings in Nvidia by two-thirds to $19 million at the end of June after selling about half a million shares. Last month, Griffin told his company’s new potential talent that he doubted GenAI would be as revolutionary as others believe.
Schmidt, whose net worth doesn’t quite match Griffin’s but is still on almost 24 billion US dollarstold Stanford students that he is less targeted in his search for the next AI leader: “Basically, I invest in everything because I can’t predict who will win.”