​If I had to map out the next several years, I'd think about it in terms of where the bottlenecks are likely to move.
Today (still strong):

• AI chips (NVIDIA, AMD)
• Networking (Broadcom, Arista)
• Memory (Micron)
• Semiconductor manufacturing (TSMC)

These companies are still benefiting from enormous AI spending, but expectations are also extremely high.
The next likely beneficiaries (my view):

1. Power and utilities
   This is the area I think has one of the strongest structural tailwinds.
   Every new AI data center needs a reliable supply of electricity, and in many regions the grid is already constrained. That benefits:
   
   • Electric utilities
   • Independent power producers
   • Renewable energy developers
   • Nuclear operators
   • Natural gas generation
   • Battery storage companies
2. Electrical equipment
   A data center isn't useful if it can't connect to the grid.
   Companies making:
   
   • Transformers
   • Switchgear
   • Circuit breakers
   • Power management equipment
   • Cooling systems
   could continue to see strong demand. Firms like Eaton, Schneider Electric, ABB, GE Vernova, and Siemens Energy are often mentioned in this context.
   
3. Grid infrastructure
   Building generation is only part of the story. Electricity has to be transmitted.
   That means investment in:
   
   • High-voltage transmission
   • Distribution networks
   • Grid automation
   • Battery storage
   This fits with your earlier point about Brookfield and Canada's grid modernization.
   
4. Software
   This is where the biggest long-term profits may ultimately end up.
   History shows that the companies providing the infrastructure don't always earn the highest long-term returns. During the internet boom, companies selling networking equipment did very well, but some of the biggest winners over time were businesses that built dominant software and platform ecosystems.
   If AI truly transforms productivity, companies like Microsoft, Alphabet, Amazon, and others could generate much higher earnings from AI services over time—provided they can monetize those investments.
   

One wildcard: roboticsIf AI becomes capable enough to automate physical work, then industrial automation and robotics could become the next major investment theme after software. That could benefit manufacturers of industrial robots, warehouse automation, logistics technology, and related hardware.
My biggest watch itemThe one thing I'd pay the closest attention to over the next five years isn't actually AI models—it's electricity.
The question isn't just, "Who has the best AI?" It's, "Who can supply enough power to run it?"
If AI demand keeps growing as many expect, electricity may become one of the most valuable enabling resources. That doesn't mean every utility stock will outperform, but it does suggest power generation, transmission, and grid modernization could remain important investment themes for years rather than quarters.
That said, it's also worth remembering that markets anticipate trends. If a company is already priced for years of exceptional growth, even good news may not lead to exceptional returns. Looking for areas where the market may be underestimating future demand—rather than simply identifying a good business—is often what separates a strong investment thesis from a popular one.

1. Build phase (where we largely are today)
   
   • Chip designers (NVIDIA, AMD, Broadcom)
   • Semiconductor manufacturers (TSMC)
   • Networking equipment (Arista, Cisco)
   • Optical components and fiber
   • Cooling systems, power equipment, transformers, switchgear
   • Data center construction and engineering
   This is where the bulk of the capital spending is happening.
   
2. Infrastructure utilization phase
   Once the data centers are built, companies still need to fill them with GPUs, replace aging hardware, and expand capacity. Demand for chips probably won't disappear—it may just grow more slowly. GPUs also become obsolete fairly quickly (every 2–4 years for cutting-edge AI), creating an upgrade cycle.
3. Power phase
   This is the area many investors think is still underappreciated.
   AI data centers consume enormous amounts of electricity. That benefits:
   
   • Electric utilities
   • Renewable power producers
   • Natural gas generation
   • Nuclear power
   • Battery storage
   • Grid modernization companies
   • Transformer and electrical equipment manufacturers
   This is one reason companies like Brookfield Renewable, Vistra, Constellation Energy, GE Vernova, Eaton, and Siemens Energy have attracted so much investor attention.
   
4. Application phase
   Eventually, the companies spending tens of billions today want a return on that investment.
   The biggest winners could be software companies that successfully monetize AI through:
   
   • Higher subscription prices
   • Productivity gains
   • New AI products
   • Lower operating costs
   Think Microsoft, Alphabet, Amazon, Meta, Salesforce, Adobe, and many enterprise software companies.
   

One thing I'd add is that these phases overlap rather than occur one after another. For example, Microsoft is still building data centers while simultaneously selling AI services. Utilities are expanding capacity while chip companies continue shipping record numbers of GPUs.
One caution is that markets are forward-looking. By the time everyone realizes that power demand is exploding, utility and power stocks may have already priced in much of that expectation. Investors are always trying to anticipate the next bottleneck.
"Who benefits next?" 

Brookfield Asset Management is acquiring Aypa Power, North America's largest standalone battery energy storage developer, in a US$7 billion deal that significantly expands its role in powering the next generation of AI infrastructure.
The acquisition strengthens Brookfield's US$100 billion AI Infrastructure Program by solving one of the industry's biggest challenges: delivering reliable 24/7 clean electricity to energy-intensive AI data centres. Aypa's 6.5 GW contracted battery storage portfolio and more than 20 GW development pipeline will allow Brookfield to store renewable energy and dispatch it when demand peaks, helping overcome grid constraints and renewable intermittency.
The deal also reinforces Brookfield's partnerships with hyperscalers, including its 10.5 GW renewable energy agreement with Microsoft and 3 GW agreement with Google. Combined with its expanded partnership with Bloom Energy for onsite fuel cells and the launch of Radiant, an NVIDIA Cloud Partner, Brookfield is building a vertically integrated AI platform spanning power generation, battery storage, and compute infrastructure.
The timing could also prove significant for Canada. As a Toronto-headquartered infrastructure giant with longstanding relationships across Canadian governments, pension funds, and public institutions, Brookfield is well positioned as Ottawa prepares to invest billions of dollars to modernize Canada's electrical grid and accelerate the clean energy transition. Large-scale battery energy storage will be essential to integrating renewable power and improving grid reliability, creating substantial opportunities for companies with Brookfield's expertise.
While any future projects would be awarded through competitive procurement processes, Brookfield's acquisition of Aypa strengthens its position to compete for the next wave of grid modernization and AI-driven energy infrastructure investments on both sides of the border.

Ticker is BEP.TO