As investment in AI infrastructure grows, so do concerns about financing costs, companies’ ability to turn spending into profits and the durability of current valuations. If AI-related stocks suffer a sharp pullback in 2027, investors will face a trade-off: limit their portfolios’ dependence on a handful of tech giants without giving up entirely on the potential gains from continued AI demand.
There is a historical precedent for that dilemma. In 1996, Alan Greenspan, who served as Federal Reserve chair under four US presidents, warned of “irrational exuberance.” The Nasdaq Composite went on to nearly quadruple before peaking in 2000 and falling 78%. But investors who left the market too early could also have missed further gains as the bubble expanded.
The current AI trade presents a similar tension. Bond yields for AI infrastructure company CoreWeave at one point approached 15%, prompting investors to reassess its debt burden. At the same time, Nvidia and several large technology companies continue to generate substantial cash flow, and demand for AI has not disappeared. The key uncertainty is when investment will translate into more dependable returns.
BloFin Research recommends reducing portfolio concentration in advance while retaining some long-term exposure. It sets out three approaches.
Reduce reliance on Nvidia and mega-cap tech
Investors who hold S&P 500 index funds may have more exposure to AI-related companies than they realize. In SPY, Nvidia accounts for close to 8% of the portfolio, while the fund’s top 10 holdings make up about 39%. Investors may therefore carry significant concentration risk through index funds, even without directly buying large amounts of AI stock.
BloFin Research suggests shifting some exposure to an equal-weight product such as the Invesco S&P 500 Equal Weight ETF (RSP). At each quarterly rebalance, the fund assigns each S&P 500 company a weight of about 0.2%. After rebalancing, its top 10 holdings have a combined weight of around 2%, far below SPY’s.
The change would not mean exiting the AI theme altogether: large technology companies remain in the index, but the impact of any one company’s moves on the portfolio is reduced. Equal-weight funds are still exposed to a broad market decline, however, and diversification cannot eliminate market risk. RSP had outperformed SPY at points since the start of 2026, but past performance does not determine future results.
Put some AI gains into US Treasuries
US Treasury yields offer another option for investors looking to hold cash or lock in part of their gains. As of October 20, the three-month Treasury yield was about 4.3%, the two-year yield was close to 4.8%, and the 10-year yield was above 5.1%. Yields on Treasury Inflation-Protected Securities were also close to 3% in real terms.
BloFin Research says investors with substantial paper gains in AI stocks could realize some of them and earn interest while waiting for market opportunities. The aim is not to call the exact top of the AI market, but to reduce reliance on one market trend continuing uninterrupted.
Interest-rate risk varies by Treasury maturity. If rates rise further, longer-dated bond prices typically face greater pressure. Short-term Treasuries tend to have less price volatility, but the income available when proceeds are reinvested changes with market rates. Bond allocations therefore depend on maturity and the rate environment; Treasuries should not be treated as a cash substitute with no price risk.
Bitcoin offers a separate growth exposure
BloFin Research’s third idea is to consider Bitcoin. The cryptocurrency fell more than 53% from a peak above $126,000 in October 2025 and dropped below $59,000 in June 2026. It is now trading at about $82,000.
Despite the steep decline, Bitcoin ETF holdings measured in Bitcoin have remained broadly stable. BloFin Research argues that over the past two years, more capital has flowed into AI stocks and related infrastructure, leaving Bitcoin relatively underweighted in institutional and investor portfolios. In its view, that could create a new focus for the market.
“We believe a new Bitcoin cycle is beginning from this base of under-allocation,” the firm said. It expects stablecoins, tokenized assets and AI agents could continue to expand crypto use cases in 2027. That remains a market view, and Bitcoin is not guaranteed to provide protection if AI stocks or broader markets fall.
The focus is on reducing concentration
BloFin Research does not recommend shorting AI stocks to hedge portfolio exposure. Its approach is to maintain participation in long-term growth themes while reducing dependence on a small number of technology companies and preparing for greater volatility. Investors assessing these risks can continue to monitor AI companies’ cash flow and debt, changes in Treasury yields, and holdings in Bitcoin ETFs.