Stockbroker examines detailed financial charts on multiple monitors. The best stocks of July 2026 examine companies serving known AI bottlenecks.
Getty
In May, Fidelity Investment’s Director of Global Macro Jurrien Timmer argued that stock prices could continue strong in the second half of 2026—despite the rising inflation risk and the potential long-term effects of the Iran conflict. Morgan Stanley also predicted a 12% gain in the S&P 500 over the next 12 months. Both stock market outlooks cited surging AI-related capital spending as a driver.
As the AI capex cycle continues, companies serving known AI bottlenecks stand to benefit. For that reason, they are the focus for my 10 best stocks to buy now for July 2026.
Well-documented AI bottlenecks include power, memory and cooling systems.
Natural gas is the most popular power source for AI data centers currently, but this could change over time. As power needs increase, clean sources stand to gain momentum. Nuclear energy may have a long-term advantage because it produces power more consistently than solar or wind energy.
Companies to watch include:
Processors are the headlining hardware in AI, and they are capacity constrained. But there’s also a genuine limitation related to memory, specifically, how fast data moves between the processor and storage. High-powered chips can only process as fast as they can access data. Tech companies are exploring several solutions:
Traditional air systems are proving insufficient for modern AI clusters. Liquid cooling is becoming the new standard for its efficiency and effectiveness. A report by Persistence Market Research recently predicted the value of the global data center liquid cooling market would grow to $29.2 billion by 2033 from $5.7 billion in 2026. This is only a single view, but it demonstrates the optimism around this business.
Vertiv and Schneider Electric specialize in liquid cooling.
Some of the companies noted are foreign businesses that trade over the counter in the U.S. Narrowing the list to exchange-traded companies leaves the nine options shown in the table below.
A closer look at each company follows. Metrics are sourced from company reports and StockAnalysis.com. Revenue and EPS data are from each company’s most recent fiscal year and growth expectations are for the current fiscal year.
For more investing ideas, see: Best dividend stocks and 8 index funds to consider for 2026.
Taiwan Semiconductor is the world’s largest chip foundry. The company makes, packages and tests semiconductor devices on contract for the world’s largest technology companies.
A functional AI accelerator requires multiple manufacturing steps, but two to note are the chip itself and what’s called advanced packaging. Taiwan Semiconductor dominates advanced packaging with its chip-on-wafer-on-substrate. CoWoS, as it’s known, connects processors to high-powered memory.
TSM’s advanced packaging capacity is largely allocated to Nvidia. The scarcity could provide pricing power but may also create an opening for competing technologies.
Micron Technology designs and makes memory products including HBM chips for data centers and AI accelerators.
The global supply of HBM chips comes primarily from three companies: SK Hynix, Samsung and Micron. All three have committed their HBM chip capacity through 2026.
SK Hynix is the market share leader, Samsung is working to overcome quality challenges and Micron has momentum. The rising popularity of Micron HBM chips drove the MU stock price up more than 746% over the last 12 months.
Given the supply constraints, Micron has been able to raise prices and shift customer contracts to long-term agreements. The practical outcomes are stronger margins and improved revenue visibility.
Intel designs, makes and packages AI hardware. The Intel Foundry division is responsible for chipmaking and advanced packaging services for hyperscalers like Amazon’s AWS.
Intel Foundry has a small market share compared to Taiwan Semiconductor, but it’s gaining ground amid the advanced packaging supply crunch. Technology companies are increasingly considering Intel’s Embedded Multi-die Interconnect Bridge (EMIB) and Foveros platforms as alternatives to TSM’s CoWoS.
Intel has reportedly secured an order for 3 million tensor processing units by 2028 from Google. There are also reports that SK Hynix and Nvidia are testing Intel technologies for future integration. These developments could propel Intel legitimately into the advanced packaging space.
NextEra Energy operates America’s largest electric utility and one of the country’s largest energy infrastructure development companies. The company generates power from diverse sources, including natural gas, renewables and nuclear.
NextEra is investing in its nuclear fleet and actively pursuing “green data center” contracts. The company has already secured large clean energy agreements with Google and Meta.
NextEra also recently announced its intention to acquire Dominion Energy. Dominion currently powers “Data Center Alley,” a high-density cluster of data centers in Northern Virginia. The acquisition would create the world’s largest regulated electric utility business by market capitalization, serving 10 million utility customers. It would also be a solid win for NextEra’s data center strategy. The transaction remains subject to regulatory approvals.
Vertiv Holdings makes infrastructure technologies that power and cool data centers and other technology environments.
While Vertiv is best known for its cooling solutions, the company is moving to position itself more broadly as an integrated cooling and power provider. The strategy has been effective. In February, Vertiv reported a $15 billion backlog, driven primarily by AI infrastructure demand. In June, Vertiv announced a digital twin capability for its SmartRun product integrated into Nvidia’s AI data center simulation platform.
The digital twin capability allows engineers to design, test and fix data center designs before equipment is installed. This signals Vertiv’s capability as an integrated partner in data center planning.
Williams Companies is a major natural gas processor and pipeline operator. The company handles about 30% of the natural gas usage in the U.S.
Williams Companies describes itself as a one-stop shop to supply power to data centers. To support the positioning, the company is investing in power generation assets to couple with its pipelines. Behind those investments are large, long-term contracts, including one with Meta Platforms and another with an undisclosed investment-grade company. The latter arrangement is a 10-year contract with an option for the customer to extend.
As Williams optimizes its assets and offering, it is positioned to develop a stronghold in data center power supply.
Astera Labs designs and sells purpose-built connectivity solutions to hyperscalers for use in data centers. The technology helps data move more efficiently within AI clusters.
Amid strong demand for Astera Labs’ connectivity solutions, the company increased revenue by 93% to $308.4 million in the first quarter of 2026 versus the prior year. In that earnings release, CEO Jitendra Mohan characterized the AI growth opportunity as “significant” and said the company was “investing to be a leader for rack-scale AI technologies.”
Notably, one customer — presumed to be Amazon —accounted for 70% of Astera Labs’ 2025 revenue. While the company’s Scorpio switching solutions are generating interest from multiple hyperscalers, it remains to be seen how well Astera Labs can broaden its customer base and remain relevant over time.
Cameco Corporation is one of the world’s largest miners of uranium, used to generate carbon-free nuclear electricity in the U.S. and abroad.
Cameco’s growth strategy includes active pursuit of uranium supply contracts tied to the expansion of nuclear power for AI data centers. In 2025, the company announced a partnership with Brookfield Asset Management and the U.S. government to expedite deployment of Westinghouse nuclear reactors to support domestic electricity needs, including data center demand. The value of the deal is at least $80 billion. Westinghouse is co-owned by Cameco and Brookfield.
Deployment of a nuclear reactor can take 10 years or more. The U.S. government has agreed to streamline regulatory approvals as part of the partnership—but Cameco’s monetization of AI power demand is still a longer-term play.
Credo Technology is a primary competitor to Astera Labs, supplying AI-capable connectivity solutions to data centers. The two companies take different approaches but have similar value propositions: to enhance efficiency in high-performance AI clusters.
Like Astera Labs, Credo’s revenue trend reflects exploding demand for data center efficiency from a small number of customers. In the most recent quarter, revenue increased 157% from the prior year to $437 million.
Four customers accounted for 87% of revenue in the most recent quarter. The company expects a new generation of AI cloud providers to enter the market, which will allow for greater revenue diversification in fiscal year 2027.
As AI infrastructure spending continues, the companies that can ease bottlenecks and solve problems will prevail — at least in the medium term. In the long run, relevance for these problem solvers will hinge on ongoing innovation and close collaboration with hyperscalers and hardware makers.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。