I'm writing this post so I can encourage discussion so that everyone can help me and others understand economic moats better.
Literal Definition:
An "economic moat" is a company's ability to maintain a sustainable competitive advantage that protects its market share and profits from rivals.
Since AI is so hype, let talk about semiconductors.
TSMC holds near effective monopoly on advanced semiconductor manufacturing, where nearest competitor Intel are way too behind with nowhere near efficient scale and cost-advantage. Regardless who make the best design or favorable chips like NVIDIA, Broadcom, AMD, Marvell, Cerebras, Apple, ARM everyone end up at TSMC, therefore structurally they possess a wider moat than NVIDIA who have more formidable competitors.
Comparing AMD and ARM on CPU, AMD is better positioned to profit as they already have full scale CPU servers on x86 and "acceptable" (despite unnatural deal making OpenAI & Meta) GPUs therefore their stock perform better. However, if we examine business model, ARM CPU architecture have monopolistic market share on smart phone, it preferred design for data center CPUs extremely high barrier to entry due to IP and power efficiency. Therefore, even though AMD is better positioned they automatically have narrower moat than ARM, because their CPU is inferior (losing market share to ARM-based design) and GPUs is inferior to NVIDIA, so they are narrow moat, and ARM is wide moat.
Memory-wise let talk about HBMs players so Micron Technology, SK Hynix, Samsung. Memory in the past is a commodity, but now people argue that HBMs with technological advancements are no longer a commodity. However, I argue it still commodity-like, because even though there are differences between HBMs made by these players, it is negligible and marginal to the point it almost doesn't even matter. The argument about whether Micron or SK Hynix make better HBMs can be extremely subjective and indifferentiable. As three companies that can make almost the same thing, none of these companies can have a wide moat, AT MOST it is a narrow due to scale arguably NONE. These stocks benefit from supply shortage, as that become increasingly resolved in future, there will be less pricing power and long term cash flow compared to NVIDIA, Broadcom, TSMC, ASML. Differentiable meaning like NVIDIA Blackwell is objectively better than the best AMD stuff because computing power and CUDA moat. Server sales like Dell or HPE arguably worser moats than all of these above here.
Semiconductor equipment like ASML, LRCX, AMAT, KLAC all have wide moats, but they are all cyclical businesses as foundries like TSMC, Intel, Samsung, Memory-makers invest large sum for these machines, they will only have servicing revenue for a bit of time. Cyclicality is a minus in investing.
Let move into construction and energy infrastructure, since natural gas is the leading energy form to power data centers this will benefit gas turbines maker like GE Vernova, Siemens Energy, and Mitsubishi this business is narrow moat and somewhat lower form of cyclicality because these turbines take a long time to build. Same thing apply to nuclear reactors maker like BWXT and Westing House, these are narrow moat and lower form of cyclicality, but you could argue that BWXT have wide moat from military nuclear reactors monopoly. Constellation Energy, Vistra Corporation, and other utilities are low quality businesses because of high CAPEX (debt) creating a commodity (electricity not differentiable), so all of them have no moat. However, since the energy demand is tremendously increased it can rerate their valuation similar probably not the same scale as memory.
Construction companies like Sterling Infrastructure, Comfort Systems, and many more are all heavily cyclical. Caterpillar and Deere are also cyclical but less so because of wider moat in construction equipment and backup generators (diesel).
Overall, you absolutely can be extremely successful investing in low moat companies, I bought MU at $100 avg for $2,500 at the age of 20 I found my almost 10 bagger. However, weak moat companies are susceptible to disruption and substitution. Most importantly without pricing power it will be met with margin compression and lower profits, or simply stock will tank.
Wide moat with no pricing power or high level of cyclicality is also a negative mark in investing. For example, ASML is arguably the widest moat company in semiconductor (I bought near $700 sold when it ran up 100%). However, you have to criticize ASML in the cyclical component that is the worse trait about the company creating unpredictable level of cash flows in the future. No pricing power example would be a company creating a commodity. For example, McDonald's or Coca Cola creating food and beverage, Micron Technology creating HBMs, Vistra Corp creating electricity, United Healthcare offering insurance, Costco and Walmart in consumer discretionary. These are all example of commodities or like-commodities, without differentiable products and substitution these no moat (VST), narrow moat (UNH, MU), wide moat (MCD, KO, COST, WMT) have weak pricing power.
The highest quality play like NVIDIA, AVGO, ASML, TSMC are all exposed to CAPEX cyclicality.
This leaves us with a very small universe of the highest quality stocks meaning wide moat, consistent demand (lesser cyclicality), and pricing power (not a commodity). Chris Hohn, TCI management, arguably have one of the strictest guidelines for moat even more than Warren Buffett who coined the term. Even if these stocks are not well known they are or at least the same width of moat as the world biggest companies GE Aerospace (GE), Visa (V), S&P Global (SPGI), Moody's (MO), Ferrovial (FER), Railroads, Google (GOOGL). Some additions, I think follow his guideline would be TSMC (TSM).
Let talk about software companies, since it is frequently mention in this subreddit. I would argue that some software companies that have seen major correction in valuations is completely justified, because the risk of disruption increased dramatically dangering the moat, therefore I think rerating lower is completely appropriate. You could argue that the rerating is overdramatic creating room to run, but to say that it should fully recover is unreasonable. I think Adobe (ADBE), Salesforce (CRM), and Oracle (OCRL) falls into this category where substitution can be a huge factor. I argue that these software plays have wider and durable moats I bought the dips on these and were extremely successful, first is big player cybersecurity (PANW, CRWD, FTNT), chip software (CDNS, SPNS) national security and switching cost, and Microsoft (networking effect, switching cost, but their software subscription is relatively cheap that creating your own is unreasonable).
Thank you for reading my thoughts, if you disagree with me please leave an informative comment arguing against or make your case.