Markus@ValueByMarkus
7 Moats for AI Companies - Applied to $IREN
Today I listened to an excellent Y Combinator podcast (linked below) on “The 7 Most Powerful Moats for AI Startups,” inspired by Hamilton Helmer’s book 7 Powers.
It got me thinking: how do these moats translate to AI infrastructure, and more specifically, to $IREN, considering the moats are very different for Infrastructure and application layer companies. Let´s dive in:
Before anything: Speed
In AI infrastructure, speed and execution capability are everything. Early movers who secure land, grid access, vertical integration, and GPU supply chains gain a structural advantage that is incredibly hard to replicate.
This gives them a solid head start over competitors, who now face the choice of building from scratch or competing for scarce resources at much higher costs. Where execution bottlenecks, energy, GPUs, and operational scale are the real constraints, moving first creates a structural edge that IMO $IREN possesses, also enabling the future fast scaling.
Moat 1 - Process Power
The operational excellence to convert energy into usable AI compute at an industrial scale. Building, connecting, cooling, GPU deployment, and overall running data centers with vertical integration efficiently is highly complex: competitors copying the $IREN blueprint could be highly difficult and costly.
Competitors may acquire land, energy grid access (difficult), or GPUs, but replicating the full process is extremely challenging without similar scale, experience, and know-how. IMO, a very strong and one of the most defensible moats.
Moat 2 - Cornered Resource
Cornered resources are those assets that are scarce, hard to acquire, and critical for long-term advantage. For $IREN, the most important are 3GW grid access and land, which form the foundation of its AI compute infrastructure.
While GPU access is fundamental for AI workloads, it is secondary to the energy + land foundation that enables scale. $IREN pre-secured energy and sites provide a multi-year head start, giving them a timeline of roughly 3–5 years before energy constraints will start to fade away.
During this window, the company can deploy facilities, attract hyperscaler partnerships, and optimize operations, creating an ultimate moat that compounds as AI compute demand grows exponentially. IMO, a major moat until, eventually, the energy supply will come.
Moat 3 - Switching Costs
For $IREN, switching costs are present but rather limited. Unlike AI application companies, where deep integration, custom workflows, or valuable data create strong customer lock-in, the $IREN core business here is efficiently converting energy into AI compute at scale.
Some friction exists through longer-term strategic contracts with hyperscalers or enterprise clients, as well as the operational complexity of moving workloads between providers. These factors can make switching costly in practice.
However, the moat is not structural: a well-capitalized competitor with access to energy and GPU resources could, in time, replicate $IREN capabilities. IMO, switching costs contribute modestly, but they are far weaker than the previous moats.
Moat 4 - Counter-Positioning
Counter-positioning arises when incumbents cannot pursue a strategy without harming their existing business. For $IREN, this moat is very limited. While benefiting from the freedom of being a dedicated AI infrastructure player, with minimal legacy constraints.
This allows them to pursue pure energy-to-compute execution and scale rapidly without internal trade-offs. IMO, this moat is secondary and very limited.
Moat 5 - Brand
Brand for $IREN is primarily about credibility with enterprise and hyperscaler clients. The recent Microsoft deal and mentions from $NVDA have strengthened their reputation, signaling that $IREN is a serious player in AI infrastructure.
These milestones help attract partnerships and build trust in their ability to deliver at scale. Ultimately, clients care most about fast, efficient, and reliable energy-to-compute execution, which is operationally complex. Brand helps open doors, but the real defensibility comes from execution, not recognition alone. IMO, limited, but potential moat.
Moat 6 - Network Effects
Network effects are largely absent at the AI infrastructure layer. Unlike AI applications, where more users improve the product itself, $IREN operations don’t inherently benefit from additional clients. There is a limited indirect effect: early adoption by hyperscalers, such as Microsoft, reinforces credibility and signals reliability to other potential clients.
This can help accelerate partnerships and fill facility capacity, but the advantage is small, far outweighed by operational execution and secured energy resources. IMO, a limited moat that might support securing the future scaling.
Moat 7 - Scale Economies
Scale economies are arguably one of $IREN’s strongest moats. Every part of their operations, energy procurement, GPU deployment, facility construction, and day-to-day data-center management benefits from doing more at scale, requiring major CAPEX investments.
As $IREN continues to expand, it gains a long-term advantage through reduced cost per unit of compute output. Larger facilities and more efficient GPUs & GPU integration lower the marginal cost of energy-to-compute conversion. This moat is enduring, not just near-term: potentially, it creates a foundation for future AI infrastructure deployments, positions $IREN to benefit from better GPU generations.
Smaller competitors or new entrants face a material disadvantage trying to match power-to-compute efficiency at similar costs. IMO, this could be a major moat, assuming they can execute the scaling.
Which of these moats do you think are most durable for $IREN over the next 5 years?
Where do you think my reasoning is flawed? Genuinely curious to hear counterarguments.
youtube.com/watch?v=bxBzsS…
Thanks for reading. $IREN
