When investing, your capital is at risk. The value of investments can go down as well as up, and you may get back less than you put in. The content of this article is for information purposes only and does not constitute personal advice or a financial promotion.
Quick Summary for Investors
What’s happening: AI is shifting from experimentation to scaled deployment, shifting value toward infrastructure bottlenecks such as power, cooling, networking, fabrication, and edge compute.
Why it matters: As AI capex becomes physical and operational, companies that enable the “plumbing” often see durable demand before they attract narrative attention.
What the market is missing: Backlog growth, design-win conversion, retention rates, and capacity utilisation often signal structural strength earlier than headline revenue beats.
Key risk to watch: A pause in hyperscaler spending or broader macro slowdown can compress multiples across the entire AI infrastructure stack.
Investor lens: Use these names as targeted exposure to specific bottlenecks. Monitor execution and capital discipline, not just AI headlines.
The uncomfortable truth about AI investing in 2026
The most profitable companies during a tech revolution are often not the most talked about.
Early on, markets favour visible companies. Later, they reward those who are truly essential.
By 2026, AI will focus less on demonstrations and more on real-world use. As companies move from testing to full production, their challenges become physical rather than just creative:
Power delivery limits
Cooling density
Data movement bottlenecks
Yield sensitivity in advanced chips
Network fabric congestion
Security operations overload
Edge devices requiring low-latency inference
Investors who focus only on the biggest companies may overlook those that control these key points. These are the 'dark horses', not unknown, just not fully appreciated.
What defines a true “dark horse” tech stock
A genuine dark horse tends to be:
They are important to the industry but often seen as unexciting.
These companies often:
Sit on critical parts of the AI stack: interconnect, thermal management, test systems, EDA, networking fabric, industrial automation, or edge silicon.
Demonstrate improving customer behaviour before valuation multiples fully reflect it.
Remain overshadowed by higher-profile AI platform names, limiting institutional crowding.
The goal isn’t to find obscure companies, but to spot businesses where performance and customer trust are growing faster than their reputation.
A practical screening framework
Investors can apply five filters:
Is demand structural over the next 12–36 months?
Does the company possess a defensible moat?
Is there a definable catalyst for 12–18 months?
Can it withstand volatility?
Is the market underweighting a key metric (backlog, RPO, bookings quality, design-win conversion)?
When these factors come together, the chances of outsized returns often improve.
Connectivity and semiconductor infrastructure
Credo Technology (CRDO)
Credo solves one of AI’s less glamorous but very important problems: moving data efficiently within dense clusters.
As AI tasks grow, bandwidth and power efficiency become key limits. If data can’t move efficiently, computing resources go to waste.
Investor focus:
Customer diversification
Revenue mix expansion
Margin durability through competitive cycles
Astera Labs (ALAB)
Astera designs connectivity chips used in AI server architectures, particularly where memory bandwidth and low-latency access matter.
Changes to PCIe and CXL standards could open new markets as AI servers become more modular and require more memory.
Investor focus:
Penetration into next-generation AI platforms
Sustained profitability alongside growth
Customer concentration risk
MKS Instruments (MKSI) and Entegris (ENTG)
Both operate inside semiconductor manufacturing ecosystems.
As AI chips use more advanced packaging and smaller designs, there is less room for defects. Managing the process and keeping things clean becomes even more important.
These companies are often valued like typical industrial firms, even though their demand is closely linked to the growth of advanced chips and memory.
Investor focus:
Order trends linked to advanced packaging
Margin stability through capex cycles
Exposure to geographic supply chain risk
Qnity Electronics (Q)
These companies specialise in wafer handling and fabrication. They benefit from the growth in AI chip production, even though they are not in the spotlight.
Investor focus:
Order consistency
Revenue diversification
Capital discipline
The AI data centre backbone
Once companies receive the chips, deployment speed depends on the physical infrastructure.
Vertiv (VRT)
Vertiv provides power and thermal management systems essential for high-density AI clusters.
As racks get denser, liquid cooling and advanced power systems become essential for operations.
Investor focus:
Backlog conversion into revenue
Margin preservation amid capacity expansion
Breadth of hyperscaler demand
Super Micro Computer (SMCI)
SMCI integrates GPUs, networking, and cooling into full AI-ready rack systems.
GPU designers get most of the attention, but rack integrators enable fast deployment.
Investor focus:
Execution consistency
Supply chain resilience
Profitability relative to growth
Brookfield Infrastructure (BIP / BIPC)
Owning infrastructure assets offers another way to invest in AI.
Owning data centres, power plants, and other infrastructure can provide more stable cash flow as AI demand grows.
Investor focus:
Contract duration and stability
Capital allocation discipline
Exposure to rising power demand
Design tools and networking fabric
Synopsys (SNPS)
All advanced AI chips need electronic design automation tools and semiconductor intellectual property.
Companies that make EDA tools benefit as the whole industry moves toward more complex silicon designs.
Investor focus:
Recurring revenue-based
Customer concentration
Cycle resilience
Arista Networks (ANET)
AI clusters need fast, low-latency networks to ensure computing power is fully utilised.
Efficient networks directly affect how much large tech companies get back from their investments.
Investor focus:
AI-specific revenue mix
Competitive positioning
Gross margin sustainability
Hexagon AB (HEXA-B)
Hexagon’s focus on digital twin and industrial software puts it at the centre of AI-driven automation and simulation.
As AI is used more for predictive maintenance and digital transformation in industry, modelling software becomes essential.
Investor focus:
Recurring revenue expansion
Enterprise adoption cycles
Industrial exposure balance
Edge and on-device AI
The next phase of AI growth is happening more at the edge.
Ambarella (AMBA)
This company designs edge AI chips for cameras and cars, where saving power and quick response times are crucial.
Investor focus:
Design win conversion
Automotive cycle exposure
Margin consistency
MicroVision (MVIS)
This company is involved in sensor fusion and autonomous technology markets.
Investor focus:
Contract transparency
Production timeline credibility
Balance sheet durability
Silicom (SILC)
This company makes edge networking hardware that supports distributed infrastructure.
Investor focus:
Customer concentration
Product refresh cycles
Margin resilience
Agentic automation and ad-tech
In the long run, AI will make money mainly by improving workflows.
PubMatic (PUBM)
This company focuses on programmatic infrastructure and automation.
If AI makes digital advertising smoother and more transparent, smaller infrastructure companies could benefit the most.
Investor focus:
Retention and take-rate stability
Adoption of automation products
Competitive positioning within supply-path optimisation
How investors can approach the dark horse category
It’s best to treat dark horse stocks as focused investments, not as main portfolio holdings.
A disciplined framework includes:
Sizing appropriately for volatility
Rotating capital as narratives catch up to fundamentals
Anchoring decisions in customer behaviour and balance sheet strength
Avoiding thematic overexposure
What these companies have in common isn’t hype about AI, but their ability to solve real operational challenges.
As industries mature, the most lasting returns often come from companies that quietly tackle the toughest problems.
By 2026, these tough problems will be mostly found in the underlying infrastructure.
Disclaimer: This publication is for general information and educational purposes only and should not be taken as investment advice. It does not take into account your individual circumstances or objectives. Nothing here constitutes a recommendation to buy, sell, or hold any investment. Past performance is not a reliable indicator of future results. Always do your own research or consult a qualified financial adviser before making investment decisions. Capital is at risk.