Across both parties, there are minority factions whose agendas center on accelerating economic growth through revitalized industrial policy. The “Abundance”[i] movement on the center-left promotes a pro building, pro supply argument: the U.S. can restore affordability and progress by making it easier to build housing, energy[ii], and infrastructure[iii], which comes by rebuilding government’s capacity to execute. On the right, an “American dominance”[iv] agenda emphasizes strength through energy[v] and reshoring[vi] for strategic industries[vii]. Both aim to transform an economy that is dependent on foreign manufacturing[viii] and repeatedly roadblocks energy[ix][x], housing[xi], and infrastructure[xii] development.

Artificial intelligence is now the focal point of these growth agendas. Yet parts of Silicon Valley, central to these pro-growth coalitions, are underestimating the speed and severity of public backlash against AI, along with deepening mistrust of the tech elite. This shift is turning AI into a political liability and weakening the political mandate that both coalitions need. It is producing a groundswell poised to shape the 2026 midterms and grow central to the 2028 presidential election[1].

Public attitudes toward AI have cooled markedly since the launch of OpenAI’s ChatGPT. Today, more than half of Americans express more concern than excitement about AI’s impact[xiii]. Further polling shows that nearly three-quarters of Americans fear permanent job loss from AI and a similar percentage worry that it will be used to provoke political unrest through deepfakes[xiv]. The implication is a social climate that views AI not as a beneficial tool, but as a threat to employment, identity, and stability.

While the Trump Administration has pushed to accelerate AI deployment[xv] and the corresponding infrastructure buildout[xvi], bipartisan pushback is emerging at the state and local level, especially around data centers. All fifty states have introduced AI-related measures, and more than 100 bills have passed out of over 1,000 introduced[xvii][xviii][2]. Local siting risk is also rising with fewer than half of Americans saying they would support a data center in their community[xix] and Senator Bernie Sanders recently calling for a moratorium on new data center development. Senator Josh Hawley captured the moment succinctly: “These data centers are massive electricity hogs. Someone has to pay for it all. Do not believe any politician who says it will not ultimately be you.“ Together, these positions align with polling that finds majorities of both the public and AI experts are worried the government will underregulate AI [xx].

This populist narrative is increasingly in tension with the public data to date and our own experience within the portfolio. Over the two years since ChatGPT’s release, AI adoption has largely complemented workers rather than replaced them, with jobs and real wages growing in occupations with high AI exposure[xxi][xxii]. Meanwhile, real consumer energy prices have declined since 2010 and are tracking inflation since 2019[xxiii][3], even amid rising demand. We should not dismiss legitimate concerns about concentrated job displacement, misinformation, and local energy bottlenecks. But so far, the realized effects of AI have been counter to the alarmist political narratives.

Prominent voices often frame AI in dystopian terms or as an arms race with China, which may be strategically resonant in Washington but is counterproductive to building durable public support for a technology that can raise worker productivity and advance industry. At the November US–Saudi Investment Forum, Elon Musk drew applause when he told the audience that “work will be optional” and that “at some point currency becomes irrelevant”[4]. Palantir, by contrast, has been the most explicit about public wariness, running ads stating that Americans are being sold an AI future where they are “obsolete or irrelevant” and promising that it will “make Americans irreplaceable”[xxiv]. Notably, Palantir’s business model, which sends engineers into the field to deploy AI, gives it one of the clearest vantage points on the technology’s actual impact on labor and industry.

Our view is that the most durable AI thesis is not replacement, but capability expansion: AI increases what organizations and high agency individuals can accomplish. Across the Logos portfolio, companies are applying AI in insurance processing, chemical sales, cybersecurity operations, small business administration, and professional services hiring, automating drudgery while enabling teams to accomplish more. Delivering that value depends less on model sophistication than on thoughtful productization and well-organized implementation. The strongest founders understand that the economic buyer is often not the end user. They obsess over the jobs to be done and build tools that make workers 10x more effective rather than replace them outright. This is the reality of fast-growing AI software companies, even if it remains politically difficult to communicate.

The political reaction to AI is accelerating faster than its realized effects on work and wages. That mismatch is becoming a defining feature of the investment landscape, one where returns will increasingly accrue to AI businesses that compound value through the expansion of human capability.

[1] For those looking for a prediction market play here, JD Vance may face tough sledding in the Republican primary given his Silicon Valley relationships. A contrast compared to the prediction market odds that have Vance as the favorite.

[2] These create reporting requirements, mandated disclosures and labels, and liability for model and product developers.

[3] Recent increases in nominal energy prices are largely due to inflation for the assets and labor that maintain and replace the aging infrastructure of energy distribution, not data centers.

[4] Let the record show that I like my work more than these people.

[i] Noah Kazis (2025) – Abundance by Ezra Klein and Derek Thompson.

[ii] Biden Administration (2023) – Build a Green Energy Economy.

[iii] The Infrastructure Investment and Jobs Act (IIJA) (2021) – Build.gov.

[iv] Diana Furchtgott-Roth (2025) - Why American Energy Dominance Is a Strategic Imperative.

[v] Trump Administration (2025) – Unleashing American Energy.

[vi] Trump Administration (2025) – Regulatory Relief to Promote Domestic Production of Critical Medicines.

[vii] KCUR (2025) – Could Kansas Panasonic plant get hit in Trump’s campaign against clean energy?

[viii] CHIPS Act (2022) – Public Law 117-167.

[ix] Phys.org (2025) – Opposition to energy transition projects driven by local concerns rather than right-wing populism, finds study.

[x] Reuters (2025) – US terminates financial aid for big Midwest power transmission project.

[xi] Bloomberg (2025) – Mapping a Way Out of the US Housing Affordability Crisis.

[xii] Stanford West (2024) – Investing in the Future of Mobility: The Role of U.S. Local Governments in Building EV Infrastructure.

[xiii] Pew Research Center (2025) – How Americans View AI and Its Impact on People and Society.

[xiv] Reuters / Ipsos Poll (2025) – Americans fear AI permanently displacing workers, poll finds.

[xv] Trump Administration (2025) – Accelerating Federal Permitting of Data Center Infrastructure (EO 14318).

[xvi] Trump Administration (2025) – Removing Barriers to American Leadership in Artificial Intelligence.

[xvii] Kuckuk, Adam. “Governments Walk a Fine Line as Use of AI Expands.” National Conference of State Legislatures. August 19, 2025. NCSL State AI Bills.

[xviii] Brookings Institution (2025) – How different states are approaching AI.

[xix] Heatmap News (2025) – The Data Center Backlash Is Swallowing American Politics.

[xx] Pew Research Center (2025). How the U.S. Public and AI Experts View Artificial Intelligence.

[xxi] Vanguard Investment Strategy Group (2025). AI Exuberance: Economic Upside, Stock Market Downside. 2026 Economic and Market Outlook.

[xxii] Chen, Weixuan, Suraj Srinivasan, and Shervin Zakerinia. “Displacement or Complementarity? The Labor Market Impact of Generative AI.” Working Paper no. 25-039. Harvard Business School, 2024.

[xxiii] Lawrence Berkeley National Laboratory & The Brattle Group (2025) – Factors Influencing Recent Trends in Retail Electricity Prices in the United States (Full Summary).

[xxiv] Palantir 2025 Advertisement.

For the past four decades, U.S. electricity demand has grown at roughly 1% a year[5], a slow and predictable pace that the system has been designed to accommodate. But data center expansion, driven by AI workloads, is now on track to consume nearly all that growth on its own. Estimates put 2024 data center energy use at 4.4% of all US electricity, a 2x+ increase in the last five years[6]. While the go-forward estimates range widely, the next three years are expected to see another 1.5-3x growth in energy use by AI data centers.[7]. Layer on accelerating electrification and industrial reshoring, and suddenly the grid faces an entirely new challenge: produce and sustain 5%+ annual growth for the first time in decades. That is an enormous, resource-straining leap for an industry optimized for stability, not rapid scale.

If you look closely at the data, the picture gets even tougher. Net capacity growth from traditional firm resources (gas, nuclear, and coal) is negative over the next year[8]. Aging plants are being retired faster than replacements come online[9], and new gas development faces mounting constraints from supply chain bottlenecks, permitting, and pipeline opposition[10]. Most of the headline growth is coming from renewables and storage, which together account for more than 100% of net additions. In many ways that is encouraging; solar paired with storage is increasingly capable of acting as a firm resource. But not all of it meets the uptime and stability standards that AI and industrial loads require.

Geographic mismatches between renewables generation, in the sunny or windy parts of the country, and data centers, clustering near fiber and water, compounds the problem. Connecting compute with power through the buildout of transmission and distribution lines is a notorious challenge[11] fraught with NIMBYism and politics[12]. Therefore, pockets of constraint are already forming in regions with heavy AI development.

Investors are now fully aware of the reality that energy is inextricably linked to artificial intelligence. Public markets have gone gangbusters across the core components of the electricity system (utilities, IPPs, developers, grid and transmission infrastructure, and equipment providers). Companies across the electricity value chain have outperformed the market significantly[13].

Private markets investors have also rushed into energy and power opportunities. Electrons are a commodity, but right now they are being differentiated by speed to availability and firmness (the ability to deliver 24/7). That scarcity premium will not last forever. Durable price advantages and quality (uptime, maintenance, etc.) will be required to create sustainable businesses. Still, we expect this supply constraint to persist for years. Even accounting for significant efficiency gains, the forward compute requirements for AI are staggering[14].

Opportunities exist across the power ecosystem. However, it has been surprising to watch as capital chases every developer with access to a site that happens to have grid, fiber, and water in the hope that an AI datacenter will come knocking. These are projects, not businesses. The focus should remain on key bottlenecks to unlocking capacity and system performance, where technology is used to build repeatable businesses with a durable moats.

[1] Bloomberg (2025) – AI Data Centers Are Sending Power Bills Soaring.

[2] Utility Dive (2025) – PJM capacity-price cap settlement shows faster-than-expected power-demand surge.

[3] RCR Wireless (2025) – EIA: U.S. power use to hit record highs in 2025-26 due to data-center demand.

[4] Data Center Dynamics (2025) – Microsoft delaying AI GPU installs for lack of grid capacity.

[5] Our World in Data (2025) – U.S. energy generation.

[6] LBNL (Dec 2024) – United States Data Center Energy Usage Report (2024).

[7] S&P Global Commodity Insights (Oct 2024) – Data-center grid power demand +22% in 2025, nearly triple by 2030.

[8] EIA Electric Power Monthly (Table 6.01) – U.S. net generation by energy source.

[9] IEEFA (2025) – U.S. on track to close half of coal capacity by 2026.

[10] S&P Global Commodity Insights (May 2025) – Gas-turbine lead times up to 7 years, costs sharply higher.

[11] ACEG (July 2024) – Fewer New Miles: Transmission Investment Trends Report.

[12] U.S. Department of Energy (July 2025) – DOE terminates financial assistance for Grain Belt Express.

[13] NASDAQ (2025) – ZAP ETF: Advanced charting and YTD performance.

[14] Goldman Sachs (2025) – AI to drive 165% increase in data-center power demand by 2030.

h/t Roxanne Tully Baine

Venture investing in operationally and/or capital-intensive businesses requires a different foundation than traditional software or consumer models. At Logos, we use a simple, two-pillar framework for scalable disruption called Future Fundamentals. These pillars are: Quantifiable Value and Unit Economics. Together, they inform our evaluation of opportunities and our work with portfolio companies.

Quantifiable Value

Disruption starts with the delivery of superior value. We view Value = Quality / Cost, measured against incumbents and alternatives.

• Cost – Is the solution outright cheaper, or does it lower the total cost of ownership (upfront investment, ongoing spend, working capital, etc.)?

• Quality – Domain and product specific attributes such as throughput, efficiency, speed, yield, durability, uptime, precision, repeatability, resolution, ease of use, and/or safety. Identifying the most important product quality element(s) is an essential part of the analysis.

We group quality-over-cost advantages into three bands:

SpaceX’s drop in launch cost, Illumina’s reduction in genome sequencing cost, and NVIDIA’s GPU order-of-magnitude better performance per dollar than CPUs all reset industry baselines and ignited markets. Step-change gains, like Tesla’s three-fold reduction in EV sticker price or Astranis’ five-fold cheaper satellites, drive market capture. Incremental gains such as Wolfspeed’s modest EV range bump, Redwood Materials’ estimated 30 percent materials cost reduction, and Desktop Metal’s savings on niche parts have struggled to achieve material market penetration.

To achieve venture-scale growth, a business must deliver value that clearly outweighs the frictions of adoption. While value compared to alternatives is critical, the solution must also target a critical point in the customer’s business, either a primary economic lever or an operational bottleneck – a principle we call process centrality. When strong value is applied at critical choke points, it creates the urgency required for rapid adoption and scale.

Unit Economics

The excitement around capital-intensive businesses has obscured the fact that only those with exceptional capital efficiency can generate venture returns. When each dollar deployed drives disproportionate incremental growth, promising technologies become enduring businesses.

To assess capital efficiency, we look at the fully loaded unit economics of a solution which includes:

• Defining the unit of production (e.g. device, job, customer, site, geography, etc.)

• Determining the revenue per unit

• Attributing all costs to delivering a unit

  • Direct variable costs

  • Amortized fixed asset costs

  • Allocated fixed operational overhead

Strong unit economics drive attractive returns to equity, both in unlevered and levered contexts. They need not exist at inception, but must be reasonably attainable as the business grows. Ideally, fully loaded unit economics should be predictable and demonstrate a path towards 50%+ margins. This enables rapid (<2 years) capital recycling and supports non-linear scaling. Predictable economics also unlocks the potential for financing that further accelerates the cycle times of equity capital. Margins above 30% can still be viable, but often require more equity and lead to greater early-investor dilution.

Understanding both the path and the destination of an economic model is essential. The journey from early unit costs to at-scale economics often follows Wright’s Law, with costs falling and quality improving as cumulative production grows.[1]

Three questions frame scalability: Where is the company on its development curve? How far and how quickly can it move along the curve? Which levers, such as innovation or process advantage, will drive that progress?

Other Considerations

Quantifiable Value and Unit Economics do not live in isolation. Greater value creation accelerates adoption, shortens the sales cycle, lowers CAC, and increases the share of value captured.

Even when a company appears disruptive, there are additional dynamics that separate mediocre businesses from great ones.

Understanding the True Benchmark

A common mistake is benchmarking solutions against the replacement cost of existing systems. The correct comparison is often the marginal cost of the installed base, not the replacement cost. Ignoring this can create the illusion of cost advantage when the solution is uncompetitive. Asset replacement cycles, the useful life of existing systems, and the switching costs or residual value will define the adoption window and inform how the product should be priced relative to incumbent options.

Commodities and the Supply Curve

When a company produces a commodity, long-term defensibility and margin potential rest on being the lowest-cost producer. Competing purely on price typically requires durable cost advantages driven by scale, supply chain efficiency, process innovation, or access to advantaged resources. A strong cost position relative to peers allows a company to stay profitable even in down cycles. On the wrong side of the supply curve, the market sets the price, not the company.

If durable cost leadership is unattainable, the only viable alternative is defensible differentiation. This can come from product features that customers care about, such as superior reliability, lower maintenance needs, faster delivery times, or better integration with existing systems. At Logos, this is a critical part of our assessment of energy investments.

Vertical Integration and the Jobs to be Done

A common failure is to mistake a novel technology for a business model. A breakthrough creates outsized value only when applied to a meaningful profit pool, the point in the value chain where the majority of margin is captured. Branded products, critical components, manufacturers with limited supply, or specialized capabilities often extract outsized margins. We assess:

- Where profit pools live in an industry and how they might shift with the introduction of disruptive technology

- What jobs need to be done to deliver and capture value, and decide where the business must integrate or outsource

The goal is to maximize profitability. Sometimes that means taking on more operational responsibility to ensure performance, utilization, or customer ownership. Other times it means resisting the urge to over-build. This is an ROI question where the incremental investment and effort must be compared to the incremental economics captured by integrating another job/part of the value chain.

Future Fundamentals

Future Fundamentals is not a rigid formula but an operating lens. At Logos, we rely on it twice: first, to assess opportunities during diligence; and again, to engage the founders we back. The framework surfaces the economic and technical levers that determine whether a company can scale in capital‑intensive, real‑world sectors. Hardwiring these fundamentals early helps founders avoid costly course corrections as growth accelerates.

Logos

[1] Nagy, B., Farmer, J. D., Bui, Q. M., & Trancik, J. E. (2013). Statistical Basis for Predicting Technological Progress. PLoS ONE, 8(2), e52669. https://doi.org/10.1371/journal.pone.0052669

h/t Roxanne Tully Baine and Claire Goldsmith

Highlights

1. Instant Context & Memory: Years of nested folders, version histories, and document metadata give an AI assistant deep context on user and corporate workflows. Rather than building a file store from scratch, OpenAI could plug into Dropbox’s opt‑in data pipelines and get immediate access to personal and enterprise documents—fuel for personalized models that remember past conversations, drafts, and analyses. People have loved the memory capabilities recently launched by OpenAI. Imagine those but with the last 15 years of your saved data and work.

2. Built‑In Distribution Engine: 700 million registered users (18 million paid and 575K businesses)

3. Exabyte‑Scale Backbone: Dropbox’s Magic Pocket system operates multi‑exabyte storage across proprietary data centers (handling the bulk of user data) and a global edge network. Acquiring this would spare OpenAI years of capital spending and ops build‑out, letting it focus on product.

4. Ecosystem Neutrality: As a platform‑agnostic hub—integrated with Office, Google Workspace, Slack, Zoom, and hundreds of third‑party apps—Dropbox lets OpenAI’s tools plug seamlessly into diverse corporate stacks, avoiding lock‑in and maximizing reach.

5. Defensive: Preempt other players from securing this strategic asset.

Considerations

• User Trust & Churn Risk

  • Risk: Fears over “OpenAI owning my documents.”

  • Mitigant: Historically, Dropbox users have stayed through acquisitions when new capabilities outweighed concerns. If OpenAI enforces strict opt‑in AI usage, honors existing privacy commitments, and demonstrates clear productivity gains, churn should be minimal—and those who do leave are a small fraction of the newly AI‑empowered base.

• Enterprise Compliance & SLAs

  • Risk: Large corporations worry about data governance.

  • Mitigant: OpenAI would inherit Dropbox’s ISO/SOC certifications, HIPAA/GDPR compliance, and existing SLAs. A clear pledge to maintain these standards—and to operate Dropbox as a standalone trust‑neutral subsidiary—smooths the path.

• Financial & Execution Load

  • Risk: A $8–10 billion price tag and running a global storage business.

  • Mitigant: Dropbox is net‑debt‑zero and generates $700–800 million of free cash flow annually. Its infrastructure is already lean—OpenAI’s acquisition would be immediately accretive, funding further R&D and share repurchases.

Key Data Points (Updated - May 2025)

• Scale & Distribution: 700 million registered users, 18 paid seats, 575K businesses

• Data & Infrastructure

  • Multi‑exabyte storage (Magic Pocket)

  • Hybrid cloud: 90% on proprietary data centers, edge network for global performance

• Financial & Ownership

  • Enterprise Value: ~$1.4 B; EV/Revenue: ~4.1×; EV/FCF: ~12.1× (CF from Ops less Capex)

  • Voting Control: Drew Houston holds ~77% voting power via dual‑class shares

  • Ownership Split: 75.7% institutional, 6.2% insiders, 18.1% retail

Trusted data scale, enterprise‑grade storage infrastructure, and a global distribution engine—all primed for immediate AI integration. Acquiring Dropbox could be the most direct path from powerful models to transformative, memory‑enabled AI assistants.

The tariffs amount to a self-induced supply shock, bigger than anticipated and seemingly bludgeoned into place. The impact of tariffs and the uncertainty caused by the current implementation of new trade policy will most likely create economic challenges with slowed capital investment and consumption. Whatever the outcome, the US is pushing for accelerating self-reliance and will have to find a way to replicate the relative cheapness of foreign production.

While tariffs may shield firms from global competition, they risk entrenching inefficiency rather than driving innovation.1 A more constructive path relies on the scale of the U.S. economy to foster internal competition, combined with strategic policy that directs investment and incentives toward innovation. Historically, such approaches have boosted productivity, particularly in the post-WWII era.23 Today, with limits to growth from population, capital, and education, producing more with less—technology driven productivity—is the only sustainable way forward.

As we have discussed, Logos invests in and supports companies that are driving productivity in the United States. Since the mid-2000s, the U.S. has faced a modern productivity paradox: rapid technological progress has coincided with slowing productivity growth,45 a trend that deepened after the Global Financial Crisis.678 Rather than reversing this decline, protectionist policies like tariffs risk compounding it, making the imperative to unlock productivity through innovation and technology even more urgent.

There are early signs that the current technological ecosystem can produce the results needed. Over the past two years, U.S. productivity ticked up beyond its pre-COVID trend.9 We at Logos are witness to tangible examples of AI’s early productivity enhancing use cases. However, we are early in the development and diffusion cycle. History suggests adoption will unfold gradually, and the productivity that follows will lag adoption as complementary technologies advance and industries are reconstructed, much like past general-purpose technologies of electricity10 and IT.11

Market disruptions create an opportunity for new entrants, which are unincumbered by existing business models and balance sheets. It is our expectation that incumbents will focus on cash conservation with the risk of compressed earnings requiring reductions in capital expenditures and R&D. The risk of inflation and a weakening dollar reduce the Federal Reserve’s ability to stimulate the economy with lower rates, further constraining companies with material debt levels (e.g. private equity).

These conditions push companies to seek technology solutions but place a higher bar on the ROI of those investments. Our framework is that disruption comes as a product of better value, which is simply the quality and cost of a product or service. More specifically, we believe that an order of magnitude (10x+) improvement in value is required to drive adoption to justify customer’s making the investment and navigating the challenges of change management.

Productivity will not solve everything, but in the face of mounting headwinds, innovation is the most powerful lever we have.

A brief addition to the quarterly (Q4 2024) update as the DeepSeek news makes waves across venture capital and public markets.

The headline-grabbing DeepSeek R1 model is a stark example of an existing trend in which the cost of AI models is declining rapidly—with demonstrated efficiency gains across inference and training. This has implications for the application and infrastructure layers of the AI technology stack, the core elements of Logos’ investment focus.

Inference: The DeepSeek R1 model is 50%-85% more efficient than the most comparable alternative (OpenAI's o1)1. This magnitude of efficiency gain is not new. R1 and other model advancements continue to reduce the marginal costs of AI-powered applications. In the last three years, we have witnessed cost declines from tens of dollars to single digits cents (Exhibit 1).

Exhibit 1: Cost of Inference (2022-2024)2

Training: DeepSeek R1 was (supposedly) trained for $6M3[3], one to two orders of magnitude cheaper than previous state-of-the-art models (e.g., OpenAI’s GPT-4 at >$100M4 and GPT-5 rumored to be 500M+5). However, this point is highly debated with new information pointing to significantly more capex at DeepSeek’s disposal6. While DeepSeek's methodology—which likely includes both real innovation and "borrowed" data from OpenAI—is a topic of interest, it does not directly impact Logos' investment focus. However, we are happy to discuss technical innovations further if relevant to you.

Takeaways

Jevons' Paradox states that when technological advancements drive more efficient use of a resource (such as chips or energy), the demand for and total consumption of that resource increases. In contrast to the underlying sentiment that destroyed $1T+ of market capitalization7 on Monday, January 27, demand for cloud GPU compute will be massive, necessitating significant increases in power availability and data center capacity.

The conversation on the cost to train DeepSeek R1 is noise in the analysis of the infrastructure needs and the corresponding demands for power and compute. We have always believed that the demand for compute from inference would be orders of magnitude greater than that from training, and we maintain that most inference will occur in cloud data centers due to the memory constraints of edge devices (e.g., laptops, phones) and the networking advantages of data centers.

As we stated during the Fund I fundraise, we believe AI models are becoming commoditized, with open-source driving zero-cost availability. DeepSeek further accelerates this trend, potentially reducing some of the capital barriers that was believed to limit the competitive set. At a minimum, there is another capable AI research lab generating bleeding edge AI models and presenting them as open source.

The application layer stands to benefit most from state-of-the-art models that are freely available and increasingly performant. We are in the early stages of a long-term trend, where improvements in model performance (quality and cost) unlock new AI use cases and propel demand far beyond what is currently observed or even imagined. The opportunity to build applications that harness cheap, ubiquitous intelligence and the need for the infrastructure that powers them has never been greater.

The incentive for humans creating and sharing original content will decline as it becomes more challenging to breakthrough the noise and guaranteed to be scraped by AI models for further training. A doom-loop emerges: AI-generated content proliferates, authentic human content recedes, and the open internet gradually becomes less authentic and useful—a phenomenon sometimes referred to as the “dead internet.”

In response, users may migrate towards trusted, smaller networks characterized by stronger identity verification and selective or gated access. We’ve seen early evidence of this trend across social media (to private messaging)1, forums (to invite-only communities), and ecommerce (to group-buying, live-streaming, and DTC storefronts). While these shifts point to a broader movement, large segments of the internet remain unchanged and vulnerable to degradation from unchecked AI content.

Degradation is the most benign interpretation of what is currently taking place. The internet as a mechanism for predatory behavior is growing rapidly as crime organizations professionalize. AI supercharges this challenge. Sue-Lin Wong (The Economist) has done excellent reporting on internet fraud. Please read or listen to her Scam Inc series and share it with anyone for whom you care.

Informal solutions, such as industry Slack groups or private job boards, currently provide interim relief. However, there is a significant opportunity to build businesses around trusted networks, leveraging strong identity verification and targeted, closed-network dynamics. Companies able to offer compelling services within these networks could quickly establish leadership positions, subsequently expanding their offerings into workflow automation, AI-assisted coordination, and vertical-specific software solutions. We believe this is a second-order way to create vertical AI-powered software businesses.

There is enormous potential in companies solving this trust and truth gap and are starting to invest behind this hypothesis. As AI-generated noise increases, trusted networks will become essential infrastructure for preserving discovery-driven value creation.

This fails to capture the full story. While operational labor replacements will be a temporary value creation mechanism, AI point solutions will eventually compete against similar software offerings and have limited standalone moats. Competition (and subsequent margin compression) is accelerated by AI, as code generation enables rapid software development and replication.

Companies will need to deliver both lower costs and increased product quality to capture and maintain material take rates. AI that replaces labor can and will drive cost savings, but only those businesses that offer transcendent and differentiated product quality will build enduring moats. To achieve these outcomes, rapid feature/product expansion (beyond initial labor savings features) and the creation + usage of previously unusable/unavailable data will be required. With this data, the correct positioning in the value chain, and the features enabled by foundation models, companies will be able to drive incremental volume, revenue, and/or product margin, making their products an inextricable part of a customer’s business.

Therefore, we believe the largest opportunities exist in companies that focus on specific verticals where unique data can be combined with industry insights to deliver meaningfully better customer experiences. Even the best AI models generally do not contain the industry expertise or context necessary for the best outcomes. Vertical businesses can connect to legacy systems to extract and leverage existing data, drive downstream operations, and, over time, expand feature and product offerings that span the full value chain.

An example (that conveniently talks the Logos book) is Mia, which provides AI-powered automotive retail software. At first glance, Mia is a straightforward replacement of labor with its AI-powered customer service for automotive dealers. Alone, this first product solves a material labor and cost challenge for dealerships. Mia’s AI-powered customer service provides 24/7 coverage and connectivity to additional dealer systems that enable automated and seamless downstream operations including scheduling, test-drives, inventory management, service optimization, and financing administration. However, Mia sees these customer interactions as a wedge to capture customer data and build a machine deliver outbound customer communications that deliver high ROI customer retention and marketing interactions. Dealerships spend $500K+ annually1 on marketing across channels to refill the sales funnel due to low consumer retention rates (i.e. repeat purchase) of ~30%2. Successful delivery of these services can make Mia an invaluable part of the full dealer value chain and can transform both the processes and economics of consumer acquisition and retention.

This perspective is not in keeping with the historical outcomes of software. Total software market capitalization in the United States is $6T+3, with the top fifteen vertical software businesses only comprising ~$300B (<5%)4. IT budgets in most industries are <10% of revenue5, with vertical SaaS spend (where applicable) often <1% of customer revenue. As software replaces headcount, we expect to see material expansion of IT budgets driven by vertical software offerings.

The shift toward vertical AI businesses—those that leverage unique industry data to create differentiated, deeply embedded products—will redefine traditional IT budget constraints and significantly expand the total addressable market for software. Identifying these businesses early, as they transform operational economics, is central to our investing approach.