After boosting the valuation of trillions of dollars in stocks and technology companies, artificial intelligence (AI) is generating a reaction among investors, who are now looking for ways to protect themselves against a possible market reversal. Despite much optimism from many participants, there is an increase in demand for financial tools designed to mitigate losses if the investment in AI does not prosper.
This move of caution is occurring in various areas of Wall Street. Investors are using stock put options and default protection contracts to safeguard themselves against unexpected drops. Simultaneously, corporations that develop and manage data centers are seeking mechanisms to limit losses should the need for computational capacity decrease.
Stephen Solaka, Senior Vice President at Gateway Investment Advisers, told The Wall Street Journal that they are observing considerable demand for loss protection. This apprehension intensified after companies of various sizes failed to meet market projections. Volatility also affected companies directly benefiting from the AI fervor, such as semiconductor manufacturers, large technology firms, and computational infrastructure providers.
On Friday (30), the Nasdaq, a highly technology-focused index, registered a drop of 2.9%, concluding the worst month since March 2025. Alphabet and Intel also posted some of their largest daily declines. Eugene Grinberg, co-founder of Solve, observed that 'there were no exceptions, suggesting a broad market concern for the sector, rather than specific credit events of individual companies.'
Oracle was a notable case, with its CDS spreads increasing by more than one percentage point. The need for protection is not limited only to those who own stocks; companies that lease chips and infrastructure to create AI models are also looking for ways to reduce the risk of falling prices for computing capacity.
Jump Trading, a quantitative trading specialist, recently established a dedicated desk for the computing power market. This operation exploits price disparities between suppliers and also structures financial agreements negotiated directly between parties. A practical example illustrates how this market works: an infrastructure company leasing Nvidia chips could acquire a put option to protect itself against a reduction in the rental cost of these equipment.
If demand for processing falls, rental revenues may also decline. With such a contract, the company would have the right to sell its computational capacity at a predetermined value, minimizing the impact on its results. Jump could act on the opposite side of this transaction, negotiating a cash settlement agreement or even leasing the capacity for resale to another client.
The market already shows significant variations. According to a Silicon Data index, the rental price of Nvidia B300 chips has risen by approximately 56% this year, reaching US$ 7.56 per hour (equivalent to about R$ 41). In a hypothetical situation, Jump could agree to buy this capacity if the price dropped to US$ 4 per hour (approximately R$ 22).
Dave Olsen, President of Jump, mentioned that participants are evaluating more rigorously which companies will succeed and which may face difficulties, which includes analyzing the business models and financial fitness of each company. Susquehanna International Group, another major Wall Street trading firm, has also received inquiries from data centers and private AI infrastructure companies interested in protecting their exposure to semiconductor costs.
One possible approach involves contracts linked to computing capacity price indices. If these values fall due to lower demand or increased supply, such instruments can generate payments that offset losses in other areas of the business. Nick Volbi, head of institutional forecasting markets at Susquehanna, indicated that this type of safeguard also assists companies when negotiating bank loans, as it allows them to present the creditor with an insurance mechanism against a drop in computing capacity prices, aiming for better financing conditions.
The article also highlights interest in forecasting markets related to potential restrictions on data center construction in the United States. The platform Kalshi, for example, offers contracts on which states may suspend new projects before 2028, citing New Hampshire, Michigan, and Maine. However, negotiations in this segment remain minor for now.
Despite the growing interest in protection, hedging strategies still represent a small portion compared to the total amount of investments and bets on AI. For many involved, the sector still holds vast growth potential. Olsen compared by saying: 'I don't think my house is going to catch fire, but I still have insurance. You don't have to be pessimistic to want some protection.'