Q2 Equity Outlook: Competitive Advantages in the AI Era

For the better part of two decades, software companies and information services firms have been rightfully viewed as the archetypal quality compounders. Once built, these businesses have been fortresses protected from competitive encroachment, resulting in significant pricing power and offering extraordinary incremental margins with returns on invested capital (ROIC) that only improve with scale. Furthermore, these businesses have taken share from the analog world, offering a seemingly endless runway for growth. As a result, software and information services companies have been accorded premium multiples by the market.

That playbook is now being stress-tested. The arrival of increasingly capable AI agents — from Anthropic’s Cowork and OpenAI’s Codex to OpenClaw — has prompted a broad selloff across software and information services stocks. In fact, the S&P 500 Software & Services Index, encompassing 140 companies, has fallen almost 25% year-to-date. The market is asking a deceptively simple question: If AI can replicate business logic, crawl public data, and compose user interfaces at near-zero marginal cost, what is the terminal value of a traditional SaaS subscription?

We believe the answer is nuanced. Not all moats are eroding equally, and the nature of competitive advantage is shifting in the AI era. Businesses that will compound value over the next decade are those that are investing aggressively today to build the moats of the future. Below, we share our evolving views.

Technology Companies: Capital Light to Capital Intensive

Gone are the days of exploding free cash flow and expanding margins for dominant technology companies known as hyperscalers. Long celebrated by investors for their capital efficiency, these technology behemoths are now some of the most capital-intensive businesses in the world, as they seek to dominate AI technology by building out the infrastructure to support large language models (LLMs) and the associated agents used by consumers and enterprises.

In fact, the five largest hyperscalers are projected to deploy over $700 billion in aggregate capital expenditure this year, an increase of over 60% from 2025. Most of them are expected to see a significant decline in free cash flow, which we believe will be temporary. In addition, a handful are taking on significant debt to finance their AI infrastructure buildout, and we are avoiding these companies. Newcomers like OpenAI and Anthropic are capital-intensive from the onset, burning cash at an unprecedented pace to build durable moats and network effects as well as grab a slice of the market from the formidable incumbents.

As investors increasingly question the ROIC of such massive cash outlays, there are a few useful precedents to look at. In 2015-2016, Amazon was criticized for spending aggressively on cloud infrastructure at the expense of near-term profitability. Amazon Web Services (AWS) ultimately generated substantial free cash flow, became Amazon’s most valuable and profitable business, and continues to lead in the AI era. Furthermore, Amazon also undertook other major investment cycles in the past to build out its now dominant retail and advertising businesses. Another precedent we would point to is Walmart, which went on a multi-year investment cycle to build out its ecommerce business and fortify its brick-and-mortar retail operations, resulting in record results over the past several years.

Today, the hyperscalers are making a similar wager that massive upfront capital deployment will yield durable competitive advantages in AI infrastructure that cannot be easily replicated. We are beginning to see some evidence of this: Google Cloud’s operating income grew a whopping 127% in the past year and its operating margin expanded from 14% to 24%, suggesting these investments are yielding great returns for Google. Similarly, Amazon’s AWS division saw revenue growth of 24% in the most recent quarter — the fastest growth in 13 quarters — while operating margin expanded to 35%.

The Asset-Light Software Model is Under Pressure

If high capital intensity is the story for large technology companies providing the AI infrastructure of the future, the story for software companies is one of gradual moat erosion. The traditional software business model — high gross margins, recurring subscription revenue, and high switching costs — faces a two-pronged challenge.

First, future growth is less certain due to AI disruption, with terminal values increasingly in question. AI agents can now replicate interfaces, encode business logic, and orchestrate workflows across tools in ways that diminish the value of a monolithic software bundle. Classic software moats are weakening. Learned interfaces — the keyboard shortcuts and specialized navigation that kept users locked into various software platforms — dissolve when the interface is simply a natural language conversation that novices can use. Business logic that took years of domain-expert engineering to encode in software can now be written in weeks. Public data access of large databases, once a genuine competitive advantage built on armies of custom document parsers, becomes trivial when the foundation model itself can evaluate and parse these databases in seconds.

Second, ROIC may start to erode as companies lose pricing power. AI has dramatically lowered barriers to entry for new software companies. Thousands of AI-native startups have emerged and received a record $150 billion in funding in 2025. They are tackling problems in different verticals and offering consumption-based pricing at steep discounts to existing subscriptions, eroding the revenue base for traditional software companies. While no clear winner has emerged, it is evident that the premium multiples investors once awarded to traditional software companies for pricing power and 90%+ retention rates can no longer be sustained.

But not all software moats are created equal, and we think it is a mistake to paint the entire sector with a single brush. A useful analogy is the media industry in the early 2000s. The advent of the internet eroded one of the most important moats that legacy media companies enjoyed: distribution. Many incumbents were disrupted. But those that owned truly proprietary content — or pivoted towards it, as HBO (part of Warner Bros.) and a handful of others eventually did — survived and in some cases thrived. A new class of internet-native winners like Netflix and Spotify emerged, distributing their content online directly to consumers. We see a similar bifurcation playing out today in software. The transition period for some will be long and painful, but we believe certain existing players will emerge stronger on the other side.

Reassessing Moats in the Age of AI

To understand shifting moats in the age of AI, it might be helpful to refer to Hamilton Helmer’s “7 Powers” framework, which is widely cited in business analysis. Helmer identified seven aspects of moats, namely Scale Economies, Network Economies, Process Power, Counter-Positioning, Cornered Resources, Switching Costs, and Branding. Some of these deserve more attention than others for the purpose of this analysis.

Cornered Resources have become increasingly valuable. Compute capacity and talent represent the defining cornered resource in this era. Amazon, Alphabet, and Microsoft are investing aggressively to corner available energy, industrial capacity, and semiconductor supply and doling out generous packages to recruit scarce AI talent.

Scale Economies remain powerful and are increasingly important for spreading out the costs of massive AI investments. Hyperscalers are able to amortize their infrastructure costs across their vast customer bases. The economics of training and deploying frontier AI models inherently favor scale — a dynamic that benefits foundational LLM companies like OpenAI with massive user bases — and will create barriers for smaller entrants.

Counter-Positioning is a highly relevant theme in the AI era. Companies like OpenAI and Anthropic and a swarm of AI-native startups have reinvented business models and scaled up at unprecedented speeds. Software is changing from subscription to consumption-based models. This challenges the well-entrenched incumbents, leaving them with a classic dilemma: engage in creative destruction and thus cannibalize the current revenue base, or try to protect the current business and extract maximum cash flows in the face of a secular threat.

Switching Costs warrant a completely new lens for analyzing certain industries. Software companies that have long prided themselves on the stickiness of their workflows are now seeing those workflows increasingly challenged by AI-native startups. When AI agents can rewrite and migrate legacy databases and workflows in days rather than months, the friction of switching diminishes materially.

Portfolio Positioning: Where We See Durable Quality Growth

The shifts in moats discussed above have real implications on how we construct our portfolio, which is focused on quality growth. We are increasingly distinguishing between businesses whose moats are unaffected or reinforced by the AI transition, and those whose moats are threatened by it.

We are overweight the industrial economy, reflecting our view that physical-world moats built through sustained capital investment are becoming more, not less valuable. For companies in the aerospace and rail industries, decades of accumulated capital, engineering expertise, and regulatory requirements have created barriers that AI cannot easily overcome. These businesses benefited tremendously from scale economies and cornered resources that have been built up over generations.

We are also overweight health care. In pharmaceuticals, patents remain the ultimate cornered resource as leading pharmaceutical companies derive their competitive position from molecules that cannot be replicated until patents expire.

Within technology, we favor the infrastructure layer over the application layer. Although current free cash flow challenges warrant close attention, we believe that the ongoing investment cycle from hyperscalers will ultimately translate to accelerating revenue growth and expanding margins, creating widened scale advantages and generating long-term value. We also remain constructive on semiconductor leaders, particularly those with process power and cornered resources in advanced chip design and manufacturing – the kind of moat that takes decades to build and cannot be replicated through software alone.

In software and information services, we are more selective. We favor companies with truly proprietary data assets, regulatory lock-in, and deeply embedded, critical workflows. We are cautious on companies built around public data curation, workflow automation, or business logic that AI agents can increasingly replicate.

Final Thoughts

We continue to apply our quality growth framework to managing portfolios, as we believe owning high quality businesses with strong incremental returns on capital that can compound free cash flow over time should translate to outperformance in the long run.

What is evolving is our definition of a durable competitive advantage, which we have always believed is an essential element of a quality company. The asset-light, high-ROIC model that defined software investing for a generation is not dead, but it is no longer sufficient. In the age of AI, investors must evaluate not just whether a business generates strong returns today, but whether its competitive moat is widening or narrowing, and whether its position is reinforced or undermined by the broader shift toward capital intensity.

To evaluate the structural durability of a company’s competitive advantages, we now seek moats rooted in physical assets, proprietary data, network effects, regulatory lock-in, and large scale capital investment. We are willing to accept temporary declines in free cash flow when companies are investing to build or reinforce their moats, as we believe that free cash flow should eventually improve and grow as their structural advantages compound.

As always, we thank you for your continued confidence in our management.

John Osterweis

Founder, Chairman & Co-Chief Investment Officer – Core Equity

Gregory Hermanski

Co-Chief Investment Officer – Core Equity

Nael Fakhry

Co-Chief Investment Officer – Core Equity