
The chart below depicts a substantial increase in Nvidia’s quarterly revenues, soaring to nearly $25 billion—over a five-fold escalation since the start of 2024.
Other tech behemoths like Google and Microsoft are observing these numbers with disbelief. Analysts who think Nvidia has peaked may need to revise their forecasts.
ADVERTISEMENT
CONTINUE READING BELOW
“Mainstream commentators often overlook the idea that it’s simply about anticipating the next word. They see only the excitement and a continuation of the status quo; at best, they ponder the possibility of another internet-scale tech revolution,” comments Leopold Aschenbrenner in his series of essays titled The Decade Ahead.
Before long, they will awaken to the reality that is both exhilarating and potentially daunting.
Artificial ‘superintelligence’
AI is evolving towards AGI (Artificial General Intelligence) much faster than expected, allowing machines to surpass human cognitive abilities. Beyond AGI is ASI, or artificial superintelligence.
Nvidia is not the only player in the AI chip industry. Google has launched its Tensor Processing Units to improve machine learning tasks.
Nvidia commands over 70% of the AI chip market, with rivals Intel and Advanced Micro Devices eager to chip away at this vast market share.
It is undeniable that Nvidia has a considerable advantage—often referred to as a moat—but its appealing gross margin of 78% is likely to draw intense competition, which appears to be on the horizon.
The AI chip market is anticipated to reach $400 billion in annual revenue by the decade’s end, welcoming a plethora of new entrants with specialized chips designed for specific applications.
Read/listen:
WEF forecasts rapid growth in AI, big data, and cybersecurity jobs
Nvidia and its partners enhance SA unit trusts
Magic and Warhammer: The impact of wide-moat IP
AI costs are declining faster than any previous disruptive technology.
“The expenses associated with operating artificial intelligence models of similar efficiency have been halved every four months—a trend we expect to persist throughout this decade. In contrast, Moore’s Law in the semiconductor industry has traditionally halved costs every 18-24 months, suggesting that the AI revolution is advancing 4-6 times quicker,” notes ARK Invest.
Microsoft and OpenAI are reportedly collaborating on a $100 billion data center focused on an AI supercomputer named Stargate, set to debut in 2028.
Significant funding is flowing into similar ventures across the US, China, Dubai, and other regions. Microsoft and Google are projected to spend over $50 billion each on capital investments in 2024, with Meta closely following. While not all these expenditures are aimed at AI, the trend is unmistakable.
Companies are more likely to make significant AI investments if they deem the economic returns worthwhile, says Leopold Aschenbrenner.
OpenAI has announced a doubling of its revenue to $2 billion over the six months ending February 2024. If this trend continues every six months, it could reach $10 billion by early 2025.
ADVERTISEMENT:
CONTINUE READING BELOW
Microsoft is already estimated to be generating an additional $5 billion in revenue from AI.
This suggests that significant investments in AI are likely to produce swift returns. As organizations like Microsoft, Google, and Meta begin to derive a major portion of their revenues from AI—expected to happen within just a few years—the demand for capital will inevitably increase.
If Microsoft can convince a substantial portion of its 350 million paid subscribers to pay an extra $100 a month for an AI add-on, revenue growth could surge dramatically.
“The implications are monumental. This would make AI products the primary revenue driver for the largest corporations in America, representing their most significant growth area. Projections for overall revenue growth for these companies would skyrocket,” asserts Aschenbrenner.
“Stock markets will likely respond accordingly; we could see the creation of our first $10 trillion company soon after. At that point, big tech would likely be disposed to fully commit, each investing hundreds of billions at minimum into further AI development. We might even witness our first corporate bond sale surpassing a hundred billion dollars by then.”
The power constraint
The main limitation in the AI race is the power supply. Power generation in the US has only increased by 5% over the past decade; however, the proposed construction of a 100-gigawatt AI cluster would consume approximately 20% of the total US electricity supply alone.
To meet the growing energy demands of AI, tech firms may need to venture into the power generation sector or acquire aluminum smelters to secure their substantial energy contracts.
While clean energy is the preferred choice, a Trump administration is more likely to support US leadership in the AI field to prevent China or a Middle Eastern nation from gaining the upper hand. This could mean tapping into plentiful natural gas reserves and limiting green energy initiatives.
The risks associated with lagging in AI at the state level are immense.
Superintelligence could be weaponized. Unstable nations might use it to threaten annihilation or sabotage rival states, potentially neutralizing any nuclear deterrent. Nations that gain even a slight head start in AGI will hold a significant competitive advantage.
Aschenbrenner argues that the US must secure victory in this race or risk forfeiting AGI supremacy to authoritarian powers.
“These clusters can be established in the US, and we must coordinate our efforts to ensure this occurs domestically. American national security should take precedence over the allure of unregulated capital from the Middle East, complex regulations, or even commendable climate commitments.”
South Africa is trailing in this race. The immense power needs of this emerging sector cannot be met in a country that has struggled with consistent electricity supply for most of the past decade, with the previous year being a rare exception.
Stay informed with Moneyweb’s extensive finance and business news on WhatsApp here.