What AI Themes and Stocks Will Lead the Next Bull Market?

The AI Revolution

The Iran war has been a significant market distraction recently, but it has not changed the direction of the most powerful growth trends of the next 10 years.

The world's largest technology companies are increasing their investment in artificial intelligence at a frantic pace and are showing no signs of slowing. Global AI-related capital expenditures are projected to exceed $300 billion annually soon, and the estimates keep increasing.

During the last earnings cycle, the Magnificent Seven names were under severe pressure, primarily because of the thesis that their massive investments in AI infrastructure would not yield attractive returns. However, that argument missed a key fact which is that AI investment spending is rapidly evolving, with a focus on greater efficiency. The margins on these investments are improving by leaps and bounds as technology becomes more powerful and efficient.

All the money being invested by hyperscalers such as Alphabet (GOOGL) and Meta Platforms (META) has to flow somewhere, and it is the companies supplying tools, power, and physical infrastructure to support the buildout that will profit enormously.

Meta recently unveiled a new program, the Muse series, that illustrates how AI spending is evolving. The Muse program is Meta's response to ChatGPT and Google's Gemini. Meta is committing billions of dollars to develop its own advanced AI models that are capable of generating text, images, video, and audio. And it is not alone. Every major technology company is also engaged in the race and spending aggressively to avoid falling behind. While Wall Street is worried about return on investment, the hyperscalers are focused on a competitive edge that will produce fat margins.

The AI spending flows directly into demand for chips, data centers, power, and the equipment that connects them all. The smart investment approach is not to bet on which AI platform or model ultimately wins, but to focus on owning the companies that supply the picks and shovels that every competitor in the race has to buy, regardless of who comes out on top.

The Chipmakers

The semiconductor is the engine of AI. Every AI model, from the one answering your health questions online to the one guiding a military drone, depends on specialized processors. Nvidia (NVDA) makes the most powerful of these and remains the dominant player.

Nvidia stock has lost its momentum despite its tremendous growth and margins, mainly because the story has expanded beyond Nvidia. It is still a solid investment but it has some tough competition.

Broadcom (AVGO) designs custom chips tailored to individual technology companies and offers a faster, more efficient alternative to Nvidia's off-the-shelf approach. Marvell Technology (MRVL) makes high-speed cables and connectors that enable chips to communicate at the enormous speeds AI requires. Without it, even the most powerful processors sit idle waiting for data. Advanced Micro Devices (AMD) is the primary alternative to Nvidia for companies seeking greater flexibility in building their AI systems.

Alphabet belongs in this conversation not just as an AI competitor but as an active backer of the physical infrastructure enabling the buildout. Google has taken a 14% ownership stake in TeraWulf (WULF) and backed $3.2 billion in financing for AI data center development, a signal that the largest technology companies are investing in the foundation layer rather than just their own products.

Coherent Corp (COHR) makes fiber-optic components that physically carry data between buildings and across data centers at the speeds AI requires. Indie Semiconductor (INDI) is a smaller name gaining traction in AI processing at the edge, meaning the devices and systems that run AI locally rather than in a distant data center.

The Plumbing

Once you have the chips, you need to connect them reliably. Amkor Technology (AMKR) is a key partner in the packaging process that physically assembles the most complex chips. Current AI chips produce so much heat and require such precise assembly that the packaging has become one of the major bottlenecks limiting how many can be produced. Astera Labs (ALAB) makes the components that manage the flow of data between different parts of a computer system, preventing the slowdowns that would limit AI processing.

Viavi Solutions (VIAV) sits in a unique position here. Every piece of high-speed network equipment has to be tested and validated before it goes into service. Viavi makes the test equipment that verifies AI networks are performing at the speeds they are supposed to reach. As data center networks push toward faster and faster speeds, Viavi is the company that certifies they actually work. Onto Innovation (ONTO) plays a similar quality-control role in chip manufacturing, making the inspection equipment that catches defects in the assembly process before they become expensive problems.

The Buildings and Power

A modern AI data center is nothing like the server rooms of the past. Training a large AI model requires enormous amounts of electricity concentrated in a small physical space, generating heat that must be constantly removed.

Vertiv Holdings (VRT) makes the liquid cooling systems that have become essential for these facilities. Air conditioning is no longer sufficient. Eaton (ETN) supplies the electrical infrastructure needed to deliver power safely and reliably at the scale AI data centers require. Celestica (CLS) builds the actual hardware platforms that house the chips inside these facilities.

TeraWulf and Core Scientific (CORZ) represent one of the most interesting convergence stories in this market. Both companies started as bitcoin miners, businesses that consume enormous amounts of electricity to run specialized computers around the clock. As bitcoin mining became less profitable, both recognized that their power-rich sites, existing electrical infrastructure, and large-scale computing experience were exactly what AI data centers need.

TeraWulf has locked in over $12.8 billion in long-term contracts with Google-backed financing and is building out sites across Texas, Kentucky, Maryland, and New York. Core Scientific has a $10 billion contract with CoreWeave (CRWV) , the AI cloud provider backed by Nvidia, with significant capacity already online and generating revenue. The margins on these AI hosting contracts, running between 80% and 90%, dwarf anything bitcoin mining ever produced.

Comfort Systems USA (FIX) is a less glamorous but highly profitable name installing the mechanical, electrical, and plumbing systems inside these facilities. Every data center being built needs their services.

The Power Grid

You cannot run a $300 billion buildout on a 20th century power grid. Constellation Energy (CEG) and Vistra Corp. (TFC) are the leading providers of nuclear power, which is currently the only energy source capable of delivering the continuous, around-the-clock electricity supply that AI data centers require. Both have become favored plays for investors looking to participate in the AI boom through the energy angle rather than the technology angle.

Solaris Energy Infrastructure (SEI) is a Houston-based power generation company that has moved aggressively into the AI data center market. The company provides modular, scalable power generation equipment that can be deployed quickly where the grid cannot keep up. Solaris Energy signed a 10-year agreement to supply over 500 megawatts of power generation equipment to a major hyperscaler starting in 2027 and has a separate 900-megawatt arrangement already underway. The ability to deliver large-scale power on demand, without waiting years for grid upgrades, is exactly what AI data center operators need right now.

NuScale Power (SMR) is developing smaller modular nuclear reactors that several technology companies are exploring as dedicated power sources for their own facilities. Talen Energy (TLN) has already signed a direct power agreement with a data center operator and is gaining attention as this trend develops.

Keeping It Secure

Every AI system is also a potential target. As AI becomes embedded in critical business and government operations, the security of those systems becomes a priority.

Palo Alto Networks (PANW) is positioning itself as the single platform for AI-driven security across an entire organization. CrowdStrike (CRWD) is increasingly used to protect the actual AI training environments where models are built.

Cloudflare (NET) provides security at the point where AI applications interact with users. SentinelOne (S) is a pure-play AI-native security platform gaining market share. Zscaler (ZS) handles security for the remote access patterns that AI workloads create.

The Bottom Line

The companies solving the physical constraints of the AI buildout, including power, cooling, networking, chips, packaging, and security, are likely to lead the next leg of this trend. The picks-and-shovels names are less crowded and in many cases offer better entry points than the large-cap leaders. Geopolitical risk is real, but the capital committed to this buildout does not pause for headlines. These are the companies making sure the AI revolution actually has somewhere to run.

I own a number of these names but want to expand on this playbook and look for more emerging themes. I will keep you posted.

At the time of publication, Rev Shark was long GOOGL, NVDA, WULF, VIAV, SEI and CORZ.