The market is having a moment of excitement: this year, the headlines have been about crypto-assets, capital expenditure (capex) linked to artificial intelligence (AI) and the opportunities offered by private asset markets. But are any of these vectors currently in bubble territory?
Pilar Gómez-Bravobond co-CIO at MFS Investment Managementhas decades of experience that allows him to identify flaws in the system that investors should watch out for. During a recent presentation in Madrid, she stressed that there is no red alert currently, while encouraging investors to “make a list of things that bother us and that we don’t fully understand”, emphasizing the importance of expectations compared to the real scope of these three market vectors, especially with regard to AI.
Gómez-Bravo proposed several keys to identify bubbles. First, she stressed the importance of determining whether it is a productive bubble– one that leaves usable assets behind after it breaks up – or not. She gave the example of the dot-com bubble, which left behind infrastructure such as fiber optic cables that continued to be used for years. On the other hand, with assets like gold or cryptocurrencies, the price collapse leaves little or no reusable elements. Another essential point in analyzing a bubble is therefore to assess whether there will be winners after it bursts.
How to Evaluate AI from a Bond Investor’s Perspective
The key to understanding whether there is a bubble around AI – and whether it could burst soon –Gómez-Bravo explained, lies in the ability of companies directly linked to this trend to monetize their capital investments. According to her, current multiples have not yet reached the levels observed during the Internet bubble.
According to his estimates, $1 trillion in profits would be needed to justify current investment levels. Furthermore, many MFS customers expect to see signs of monetization in the next 18 to 24 months.
“The American consumer doesn’t want to pay for LLMs (large language models) and token prices are falling. That’s why the strategy is for companies to pay for their use,” she explained. However, profitability would come more from a reduction in labor costs – through layoffs or hiring – than from a direct increase in revenue.
She also warned of the social risks of AI, particularly due to the high energy consumption of data centers, which increases electricity costs and impacts inflation. “There is a risk of a populist backlash, as the high electricity consumption of these centers affects the utility bills of nearby residents and could spark protests against the construction of new facilities. »
The role of private markets in financing AI
For Gómez-Bravothe concern is not so much about high valuations or increased investment in AI-related infrastructure, but rather the emergence of a closed ecosystem in which the Magnificent Seven finance their operations among themselves. As an example, she noted that OpenAIstill unlisted, announced $500 billion in investments although it remains in the red.
“AI growth is largely financed by private debt,” she explained, noting that only half of AI investments are financed by cash flow. Currently, AI represents more than 14% of investment grade (IG) debt.
The expert’s warning is clear: the bubble could take on a systemic character if the traditional financial system began to participate in it. “When banks start financing private debt transactions, the risk will increase. » She cited examples like J.P. Morgan And UBSboth of which are exposed to failed private transactions such as First brandswhich has recently been lacking.
“It will be crucial to monitor the correlation between bank balance sheets and the private market,” she stressed, pointing in particular to regional American banks. “Private markets are neither good nor bad, but they carry systemic risks, lack regulation and are not always transparent. »
She also highlighted the increase in venture capital fundraising carried out off balance sheet, a sign of fragility which could take time to manifest itself. She further warned of a new accounting problem: data centers are depreciated over six years, while the chips that power them only have a lifespan of two years.
Cryptocurrencies and Stablecoins
Although she clarified that she was not a specialist in the matter, Gómez-Bravo shared thoughts on the rise of cryptocurrencies, particularly stablecoins (digital currencies backed by the dollar), access to which for individual investors has expanded following recent regulations.
The growth of stablecoins, she noted, implies captive demand for Treasuries, and the US government has shown its intention to support this trend by issuing new debt. The only obstacle, she warned, could be the independence of the Fedbecause its high-rate policy puts pressure on the short end of the curve – just as the US Treasury increasingly relies on short-term issuance.
“For now, Fed policy is not a problem, but in the future, the rise of stablecoins could become a threat to Treasury bonds, which serve as collateral for the global financial system.” Gómez-Bravo concluded.
