Investors who have watched their favorite AI stocks tumble recently may be wondering if the AI boom is losing steam.
After all, concerns about the enormous costs of building AI infrastructure, rising energy prices, and the possibility of higher interest rates have raised questions about AI companies’ long-term potential.
But according to Andy Goldberg, chief investment strategist at Nomura Asset Management International, the recent sell-off may actually be good news for investors willing to look beyond the market’s short-term volatility.
Goldberg sat down with TheStreet’s Caroline Woods to explain why he believes the selling has been driven more by investor anxiety than any significant deterioration in corporate earnings or the broader economy.
But that doesn’t mean investors should rush to buy every AI stock whose share price has fallen. Instead, Goldberg believes the key is becoming more selective about which companies are best positioned to benefit from the next phase of AI growth.
Here’s where Goldberg sees opportunities in AI stocks, why he’s cautious about certain tech companies, and how he thinks investors can diversify their portfolios to manage risk.
Is the AI sell-off an overreaction?
Goldberg’s explanation starts with market plumbing rather than new information.
Higher interest rates, elevated energy prices, and concerns around the AI build-out have already made investors uneasy, he said. The election became another trigger. When large numbers of investors decide to reduce their holdings at the same time, selling can overwhelm buying, causing stock prices to fall sharply as buyers become harder to find.
Nothing overnight fundamentally changed. Profits didn’t change. The GDP didn’t change. Inflation didn’t change.
Andy Goldberg, when asked whether the sell-off resembled earlier sharp declines that rebounded quickly
That doesn’t mean that every falling stock is a bargain. But Goldberg believes investors should distinguish between stocks that are declining because of widespread selling and those that are falling because of deteriorating business performance.
He pointed to a historical pattern in which six of the market’s 10 best days over a 20-year period occurred within two weeks of its worst days. For Goldberg, that’s a reason to stay invested rather than trying to time a market recovery — something most investing experts warn is an exercise in futility.
Related: Why S&P 500 investors may want to look beyond mega-cap tech
How data-center restrictions could create AI opportunities
As the AI buildout becomes more complicated, Goldberg believes investors need to become more selective about where they put their money.
He pointed to campaign positions in Texas, Ohio, and Georgia that called for pauses on new data center construction while officials assess their potential impacts. Companies depending on future facilities in those states could face delays, making it important for investors to distinguish between businesses planning new facilities and those with operations that already exist.
This distinction echoes a concept that James Demert, chief investment officer at Main Street Research, recently discussed with TheStreet.
Goldberg favors semiconductors as a contrarian pick. He believes they have become oversold amid ongoing shortages of GPUs (graphics processing units), the chips used for intensive AI computing.
Although restrictions on data center development could slow new projects, Goldberg argues they could also make existing facilities more valuable.
He contrasted today’s AI boom with the dot-com boom, when companies built fiber optic infrastructure ahead of demand. Today, he said, operating AI data centers are running at maximum capacity from day one.
A slower rollout could also change investors’ expectations on returns. Building new data centers is difficult because projects require permits, local approvals, materials, power, water, cooling, and fiber-optic connections. These constraints make it harder for companies to accelerate their capital expenditures.
Related: Could a bond fund warn investors about the next tech selloff?
Why Goldberg favors some AI stocks more than others
Goldberg believes investors should distinguish between the broader AI investment opportunity and the individual companies associated with it.
For example, Nomura tends to underweight software companies because AI could disrupt smaller software functions while improving other products, making it difficult to predict which businesses will ultimately benefit.
By contrast, Goldberg favors much of the Magnificent Seven because those companies are large, well-capitalized, and supported by multiple revenue sources.
He’s more cautious about speculative AI IPOs that depend on one single product (like Anthropic).
Goldberg also considers cybersecurity a difficult area to avoid over the long run, although he acknowledged that many companies in the group are expensive based on traditional valuation measures such as the price-to-earnings ratio. His case is based on demand: More AI use can create more cyber risks, which, in turn, creates more need for protection.
Why investors should diversify beyond Big Tech
For investors with significant exposure to the S&P 500, Goldberg believes concentration is becoming an important concern.
He pointed out that 10 stocks account for approximately 40% of the index, a level of concentration he finds excessive.
At Nomura, Goldberg said the firm responds by adding fixed-income investments, including bonds, while diversifying its equity holdings.
He named real estate investment trusts (REITs), utilities, telecommunications, value stocks, and small-cap stocks as areas that were impacted when rates rose but could benefit if rates and inflation cool.
The biggest mistake is that after a big bout of volatility, I sit out on the sidelines for a while.
Andy Goldberg, when asked what retail investors should avoid after the sell-off
Goldberg’s warning about concentration also extends beyond U.S. megacaps. He highlighted consumer staples as a potential defensive holding further on in the economic cycle and said European financial companies can offer dividend yields two to three times those of the S&P 500.
For high-tax-bracket investors who live in high-tax states, Goldberg also mentioned high-yield municipal bonds, which can produce taxable-equivalent yields of 9% to 9.5%, depending on an investor’s tax situation.
The bottom line for AI investors
Investors should begin by deciding whether they want long-term exposure to AI leaders, a tactical position for a broader rally, or both. Short-term investors may favor bonds, while long-term, buy-and-hold investors may be able to hold onto tech stocks through sharp market swings.
Those looking to increase their AI exposure should consider factors such as semiconductor demand, data-center construction constraints, energy availability, and corporate balance sheets.
And for investors looking beyond technology, Goldberg suggests that each new addition should align with your goals and time horizons rather than serving as a knee-jerk reaction to one volatile week in the market.
Related: Why AI investors may favor established data-center owners
Sources
Andy Goldberg on TheStreet · watch the full segment
Cited segments:
Goldberg said simultaneous small trims by many investors can create dislocations and a vacuum in which sellers outnumber buyers. (2:00–3:26)
Goldberg said six of the market’s 10 best days over 20 years occurred within two weeks of its worst days. (3:51–4:42)
Goldberg said he would buy the pullback while becoming more selective in the maturing AI cycle. (4:42–6:48)
Goldberg said data-center constraints could extend the AI build-out and cited a hypothetical shift from a 50% one-year return to roughly 10% annually for five years. (7:53–8:24)
Goldberg said Nomura is adding fixed income and diversifying equity exposure as market volatility rises. (19:27–21:18)
Goldberg said 10 stocks account for about 40% of the S&P 500 and described that concentration as excessive for his approach. (28:05–28:37)
Some or all of this content may have been generated by AI. While we strive for accuracy, AI can occasionally produce incorrect information. If you buy something via one of our links, we may earn a commission.