Wall Street has rewarded investors in 2026, but they’ve questioned the durability of those gains.
Through Sept. 1, the S&P 500 shot up 11.5% year-to-date, the Nasdaq Composite 12.3%, the Dow 9.8%, and the Russell 2000 17.7%. Yet the Nasdaq entered a correction in March as the Iran war drove oil higher, markets rebounded to August records, and September opened with another sell-off.
That said, Morgan Stanley’s head of U.S. public-policy research, Ariana Salvatore, in a CNBC interview, just delivered a pointed message for worried AI stock investors.
The concerns around the AI buildout have shifted beyond chip demand and valuations. Communities are resisting data centers that underpin AI due to higher electricity bills, heavy water consumption, construction-related disruptions, and pressure on strained energy grids.
That tremendous resistance has translated into audits, stricter permitting, and demands that tech companies finance their own infrastructure.
Salvatore doesn’t dismiss the political threat. Instead, her conclusion draws an important distinction between what the backlash might disrupt and what investors may be prematurely writing off.
Morgan Stanley sees AI spending surviving the political squeeze
Salvatore began by breaking down the issue at hand for AI stock investors.
“So it’s remarkable how quickly the public opposition to data centers has become powerful and bipartisan, and politicians are listening.”
She identified three major pressure points: higher utility bills, environmental concerns, including water consumption, and quality-of-life disruption from construction projects.
More Wall Street:
- Wall Street’s AI trade faces its biggest valuation test
- The next Wall Street shift is already underway
- Wall Street sends strong 4-word verdict on the stock market
That said, she believes the political risk is mostly local rather than ideological at this point. The pushback is emerging in Republican- and Democrat-led states, while governors such as Greg Abbott and Josh Shapiro have turned toward tougher oversight.
Yet she isn’t interpreting that resistance as the end of the AI investment cycle. “We think it’s likely that the CapEx story is still intact. We’re still very constructive,” Salvatore said. “We have over a trillion dollars in CapEx from the hyperscalers next year.”
The big differentiator will be timing. “We just think it’s more likely to be conditional,” she said, underscoring “timing delays” and “geographical dispersion.”
In practice, hyperscalers could preserve their overall budgets by postponing projects in politically sensitive regions and redirecting capacity toward areas with greater energy, water, and public support.
A delayed project could push chip, networking, cooling, and electrical equipment sales into later quarters without gobbling up demand. Moreover, geographic dispersion could efficiently redistribute winners across utilities, developers, and infrastructure suppliers.
However, longer permitting schedules and higher community or energy costs could weaken project returns, particularly for more leveraged players.
Overall, though, Morgan Stanley’s message is constructive. “The overall story, we think, is pretty robust,” Salvatore said.
The trillion-dollar AI boom faces its biggest political test
The backlash surrounding the AI boom is growing, and it centers on who absorbs the cost of supplying it.
Morgan Stanley estimates that U.S. hyperscalers will spend a whopping $800 billion in 2026, Reuters noted, and nearly $1.1 trillion in 2027. Similarly, Goldman Sachs projects $7.6 trillion in AI infrastructure investment through 2031.
Energy supply isn’t expanding nearly as quickly. According to PJM, hyperscaler data centers can connect within two to three years, while a new power plant might require four to six years.
Similarly, Pennsylvania argues that data centers generated $29.4 billion, or 46%, of capacity charges across PJM’s four most recent auctions, raising serious concerns that households could finance infrastructure built for Big Tech.
Related: Jim Cramer reveals his 20% rule for winning stocks
Texas shows how political support can reverse.
Governor Greg Abbott ordered an audit of data-center projects seeking ERCOT connections before any are greenlit, Utility Dive reported. ERCOT is looking at 474 gigawatts of proposed load, which is more than five times the state’s record peak demand, with data centers accounting for 90% of that total.
“Simply put, Texans must come first,” Abbott said.
Yahoo Finance reports that Texas is also expecting to forgo $3.2 billion in sales-tax revenue over the next couple of years, prompting state Sen. Joan Huffman to call the cost “extremely concerning” and “unsustainable.”
In addition, Pennsylvania recently removed AI data centers from fast-track permitting, requiring developers to obtain local approval, fund new energy infrastructure, and conserve water.
“These are some of the biggest companies in the world,” Governor Josh Shapiro said. “They can afford to be good neighbors, follow the rules, and do this right.”
Reuters reported that New York went even further, pausing permits for facilities that are using at least 50 megawatts amid nearly 12 gigawatts of queued demand. At the same time, 71% of Americans oppose the building of an AI data center near them.
So the big risk for investors has less to do with demand and more to do with costly projects, delayed equipment orders, and a spending boom that’s contingent on community consent.
The AI trade survives, but stock selection matters more
Undoubtedly, AI has become a critical support for the current bull market.
Over the past three years, through early 2026, according to Yahoo Finance, the S&P 500 gained 76%, compared to 32% for an index excluding AI-linked stocks. That level of concentration means disruption to data-center investment could hit the broader market hard, not just chipmakers.
Morgan Stanley’s message is constructive but conditional.
Political resistance doesn’t erase demand for compute, but it raises the cost of converting spending into capacity. Permitting delays, grid constraints, and water rules could potentially postpone sales, compress returns, and redirect projects between states.
It’s also important for investors to separate demand risk from execution risk. Stronger exposures include profitable platforms and suppliers with contracted backlogs, pricing power, and diversified customer bases, along with fortress-like balance sheets to absorb delays.
Weaker exposures include leveraged developers, speculative utilities, and vendors whose forecasts assume that every planned campus will open on schedule.
So the move isn’t to abandon the AI thesis, but to reduce concentration and stop treating every beneficiary equally.
Companies that can efficiently monetize installed capacity today should remain at the top of investor radars, while investors should stage purchases when valuations reach nosebleed levels, while demanding clear evidence of strong returns.
