AI vs. SI

AI is dead, long live Superintelligence! The US president recently signed an executive order directing federal agencies to use the term “superintelligence” instead of artificial intelligence. Or, to put it another way, how rhetoric can be used to regain control of a topic that has captured the attention of investors worldwide. Beyond this announcement, the tech industry continues to move forward with the emergence of a new breed of personal agents, such as Muse, while also running up against the realities of infrastructure constraints, as illustrated by Oracle’s recent setback. In light of these multiple developments, an update to our analytical software seems essential.

It did not take long for the US president to respond to Anthropic CEO Dario Amodei’s call for restraint. On September 29, he convened the six largest US AI players at the White House for a meeting that concluded with the signing of a “morally binding” regulatory agreement. Jay Clayton, former head of the SEC* and current Director of Intelligence, will be tasked with the delicate job of leading the task force. Its role will be to monitor risks associated with superintelligence while ensuring that innovation is not stifled, thereby preserving the United States’ lead in this highly strategic field.

Immediately after leaving this high-profile meeting, Meta owner Mark Zuckerberg stated: “We can all agree on one thing: the American public and our customers want to be confident that what we are building works as we all intended.” Putting words into action, Facebook’s parent company recently launched the virtual assistant Muse, which immediately rose to the top of the US App Store download charts. Designed for the general public, this AI agent does more than simply provide information: it takes action directly, for example by booking a trip or carrying out transactions. OpenAI has followed suit, unveiling its personal agents, Dots, at its fourth DevDay, the company’s annual developer conference. These agents are also continuously proactive rather than acting only on an occasional basis.

Investors were quick to recognise the significance of these developments, once again favouring CPU manufacturers**, the major beneficiaries of this inference economy***, which is highly demanding in terms of computing power. At the same time, they believe these innovations could challenge business models built on inertia rents. This term encompasses industries that rely on friction, such as the difficulty of comparing prices or cancelling subscriptions. As a result, the “consumer inertia” thematic basket developed opportunistically by investment bank Goldman Sachs has fallen by 9% since Muse’s launch was announced.

While innovation among hyperscalers and research labs continues at full throttle, infrastructure is struggling to keep pace. Following Qatar, albeit for entirely different reasons, Oracle has declared force majeure on Project Jupiter, a data centre scheduled to come online in New Mexico in 2028. Press reports cite delays related to obtaining permits concerning air and water quality, as well as connecting the site to a natural gas pipeline. And this is not an isolated incident: 60% of data centres scheduled for delivery by the end of 2027 are experiencing difficulties accessing the electricity grid, according to the Wall Street Journal.

These infrastructure challenges should not prevent US AI star Anthropic from going public by the end of the year. The creator of Claude is targeting a valuation of $2 trillion and recently published a 261-page prospectus reviewed by the Financial Times. The document provides greater insight into the company’s business model through its key financial aggregates. Last year, the company’s revenue increased twelvefold to $4.6 billion, but the picture becomes more complicated further down the income statement: an exponential rise in costs resulted in $8 billion in operating losses! Following an operationally breakeven first half of 2026, the company appears to be forecasting a profitable fiscal year from 2028 onwards.

In the meantime, the US equity market has once again reached a peculiar and potentially risky position. The S&P 500 index is just a stone’s throw from its all-time high, but market breadth remains extremely weak. Indeed, only one-third of the index’s constituents are trading above their 50-day moving average. As a result, market concentration has reached an extreme, with the ten largest companies accounting for more than 40% of the index. This inevitably brings back memories, for those old enough to remember, of the TMT bubble of the early 2000s. There is, however, one major difference: the consensus earnings estimates for these ten companies over the next twelve months account for approximately 37% of the S&P 500’s aggregate earnings.

This is why the quarterly earnings season, which has just begun, is so crucial. It will need to confirm the earnings growth trajectory of the tech giants, as well as their ability to monetise capital expenditure. Otherwise, investors will probably unwind positions accumulated recently in the name of superintelligence…

* Securities and Exchange Commission.

** Central Processing Unit.

*** The phase during which a trained model applies what it has learned to new data in order to generate a response.

Pierre Pincemaille, General Secretary of Investment at DNCA. This article was finalised in October 16th, 2026.

This promotional document is a simplified presentation and does not constitute a subscription offer or an investment recommendation. No part of this document may be reproduced, published or distributed without prior approval from the investment management company.

DNCA Investments is a trademark held by DNCA Finance