Our ongoing exploration of the hardware development stack, from AI-driven CAE to PCB automation, has consistently pointed us toward a fundamental challenge: the immense complexity of coordinating diverse tools, stakeholders, and workflows across the hardware development lifecycle. While individual design tools have evolved, the job of orchestrating these pieces – managing requirements, test data, manufacturing handoffs, and team collaboration – remains a major bottleneck.

As Western economies pour unprecedented capital into hardware innovation across aerospace, energy, and defense, an entirely new class of hardware companies is emerging. And they are building with fundamentally different expectations around tooling and development speed. The incumbent hardware solution stack fails to meet these heightened expectations – it is fragmented across systems, heavily manual, and lacks real-time visibility. 

As a result, we have seen many emerging hardware companies rolling their own solutions to solve internal and external orchestration across the development lifecycle. Stoke Space’s Fusion, an internal tool that they externalized, is one such effort. This trend, which we have seen inside of several other companies, signals both the severity of existing tooling gaps and validates demand for better solutions.

As such, we see a Stripe-like opportunity to enable and capture a portion of the value created by this new class of companies through the type of critical, but boring, infrastructure that we have deemed “plumbers for reindustrialization” in other research.

We see three primary areas of opportunity for new companies at the orchestration layer:

Test Data & Observability: The proliferation of sensors and testing equipment has created data noise that existing tools can't handle effectively. Real-time analysis of test data, coupled with AI for anomaly detection and optimization – DevOps-like telemetry and monitoring – could transform validation processes that historically relied on manual review and tribal knowledge.

Unified Data & Collaboration Hubs (Next-Gen PLM): The shift toward distributed engineering teams and expansive supply chains (e.g. via outsourcing) has exposed the limitations of current tools. Engineers spend a material amount of their time on non-value-added work like converting files, updating documents, or searching for the latest designs. Modern cloud-based hubs that unify product data (requirements, CAD, tests) could dramatically improve productivity.

Manufacturing Orchestration: The gap between design and manufacturing is a major bottleneck. Tools that automate the translation of designs into manufacturing instructions and provide real-time feedback on manufacturability could significantly reduce iteration cycles and costs.

New platforms built specifically for these emerging workflows – distributed by default, data-intensive by design, and automation-ready from the start – are naturally advantaged.

From a go-to-market perspective, focusing on emerging hardware companies helps orchestration companies avoid legacy processes and tooling and instead focus on shaping modern development workflows. These companies are building complex hardware under intense time (and capital) pressure – they need tools that can keep pace. As these tools prove their value to early adopters, they can expand both vertically (serving larger enterprises) and horizontally (connecting more parts of the development process). 

However, this means our thesis relies on this new class of hardware companies being a durable and scalable customer base. If the end game is dozens of sub-scale acquisitions and a select few successes – leaving today’s incumbent hardware OEMs as the primary market – the entrenchment of existing tooling and orchestration companies (from Siemens to Jama to PTC) will be harder to break.

Similar to what we have concluded in our research into AI-driven CAE, success doesn’t require displacing incumbent tools outright. Rather than competing head-on with entrenched CAD/CAE/PLM systems, new platforms can focus on making these tools work better together – becoming the connective tissue that coordinates modern hardware development. Once established as coordination layers, these platforms position themselves to expand their footprint over time.

The PLM and hardware development tooling market can already be measured in the tens of billions, but we believe the truly transformative companies will win by expanding the market and helping hardware companies iterate and build at the speed of software. This acceleration creates a powerful flywheel: faster development cycles enable more products, which drives increased tool usage and data generation, further improving development speed. Just as software development tools expanded their market by enabling faster iteration cycles, we believe the winners in hardware orchestration will grow the market by unlocking new levels of development velocity.

The risks are real – long sales cycles, integration complexity, and regulatory requirements in sectors like aerospace and defense. But we believe the confluence of market demand (driven by reindustrialization), technological convergence, and incumbent blindspots create a unique opportunity for new platforms to emerge.

The generative AI era presents an interesting paradox – strong confidence in the directional arrow of technological progress (the ever-expanding and evolving LLM blast radius) coupled with significant uncertainty around the economic implications, both macro and micro. Today, the deployment of generative AI solutions into the enterprise, particularly large companies, has started and often exceeds expectations. 

At the same time, there is wide agreement that while early applications are driving positive ROI, most organizations face a significant change management problem in fully incorporating them into existing operational frameworks - “there are no AI-shaped holes lying around.”  For many enterprises and their executives, this has led to a “flight to trust,” with large consulting firms benefitting from enterprise appetite to utilize generative AI. This uncertainty around future enterprise workflows is also furthermore reflected in the observation that most AI startups that have found traction have done so with an anthropomorphized approach, “selling the work” in areas like legal support and accounting – essentially building an end-to-end AI replica of what customers have come to expect from human + software. 

While we think great business can be built there, this can’t be all. We believe that as organizations and society develop a better understanding of AI, we build core workflows around this new paradigm, constantly adapting existing organizational structures and coming up with entirely new ones. We broadly differentiate between resilience and optionality and believe that both areas provide opportunities for interesting models and platforms to emerge. 

Resilience focuses on enabling existing companies forced to adopt AI (in secure) and effective ways to stay competitive. As described above, these companies already have processes and employees. Both might have a hard time adopting. 

As with any complex system, we believe there is a unique opportunity by looking at the smallest unit in organizations - employees. While executives and consultants try to implement AI policies top-down, high-agency individuals (armed with ChatGPT, Claude, Cursor, and the latest tools) are constantly discovering productivity enhancements built around their idiosyncratic workflows, often utilizing these tools without explicit permission. We see an opportunity to push much of the work of identifying enterprise-specific best practices to these forward-thinking individuals and for a novel platform focused on this bottom-up approach to AI resilience to emerge.

In the process, such a platform could kill two birds with one stone. It provides a starting point for better data controls and security processes to manage risk while helping companies understand the financial implications (productivity improvements, cost structure changes, etc.) of continued AI deployment. 

Furthermore, monitoring and visibility in AI use by employees help enterprises gain insight into best practices (making AI fit into existing holes) that can be rolled out across the organization. The big opportunity that emerges from this wedge and model for enterprise trust building is that such a platform positions itself as we move toward a world of “spontaneous software” and possibly, “AI as the last employee” – similar to how Workday came to define ERP for the “digital transformation” era. 

Optionality focuses on building companies around novel organizational structures with a view to the upside, native to AI workflows and not possible before. 

This is an extension of what we previously wrote on “spontaneous software and SaaS on-demand”. In line with a recent post from Nustom that draws parallels from the autonomous vehicle market to propose the idea of a self-driving startup, we believe that there is a massive opportunity here for companies that operate like mobile game studios, making use of the reality that software is increasingly cheaper to write and startups cheaper to run with AI getting better and more capable at both. We expect these companies will excel at rapid experimentation and iteration, consistently positioning themselves ahead of the capability curve to try to catch lightning in a bottle (hits-driven) or/in combination with being long-tail driven with a large number of small cashflow-generating businesses under one roof. 

We believe that digital olfaction and a better understanding of how molecules make us smell are among those areas. Scent, due to its complexity, is our least understood sense. Novel approaches to machine learning, such as GNNs, have been proven to cut through this complexity and beat the human nose at scent profile classification based on molecule structures. Osmo, the company at the forefront of this research, has proven that it can utilize this understanding to develop novel scents. It is reasonable to assume that this technology will enable the faster development of novel molecules at lower cost and at scale. 

In 2023, the global aroma chemicals market was valued at approximately USD 5.53 billion (unclear if this also includes produced chemicals vs. only IP). This market is essentially dominated by a few large players:  Givaudan, Firmenich, IFF, and Symrise. All these players are fully integrated, meaning they both develop scent molecules (IP) and manufacture them. It is unclear how much value is in the pure IP, but some tailwinds could favor the emergence of a novel AI-enabled player focused on novel IP. In 2023, a class action lawsuit was filed against major fragrance companies Givaudan, Firmenich, IFF, and Symrise. This followed multinational (Switzerland, Europe) antitrust investigations into suspected price-fixing by these companies initiated earlier the same year. Moreover, there is a marketable shift in the industry to focus on sustainable molecules that don’t require scarce resources and have no negative health effects. 

The ability of AI to generate such molecules that either have novel scent profiles or are similar to existing ones without having negative externalities (e.g., associated with production health) is likely a unique fit for these AI models. We expect that to create maximum value, such a model (or suite of models) would likely need to be capable of 1) the ability to model molecule interaction to create a whole scent, 2) an understanding of constraints (e.g., toxicity, costs) and 3) the ability to assess the producibility of these molecule sets at scale. 

Moreover, we see substantial potential for market expansion. Suppose these AI systems are capable of identifying, mapping, and predicting the behavior of scent molecules, given certain hardware advancements (essentially a chip capable of detecting, analyzing, and recreating scent) are made. In that case, several new application areas emerge: These can span from environmental monitoring to medical diagnostics, where AI can detect disease biomarkers through molecular analysis to consumer applications such as capturing, reproducing, and sharing scent online. White is hard to quantify, but it is reasonable to assume that there is substantial option value.

As old space assets deteriorate and more are launched, collisions are inevitable, particularly as a result of space debris. Satellites already make 100,000+ maneuvers per year for collision avoidance. Losses due to collisions in low earth orbit were estimated at ~ $100m in 2020. Since then, we have at least tripled the satellite population.

While responsiveness is improving (e.g. edge GPUs to enable on-board autonomy), hundreds of billions of dollars in assets are and will be exposed without i) precise monitoring, ii) proactive defense systems (beyond trying to move out of the way), and iii) adequate financial risk management (i.e. insurance models). 

While it is easy to forget amid the current space market momentum, the industry is still walking a fine line – something that seems to have motivated Elon’s all-in effort in the direction of Donald Trump’s election. As the nuclear industry has demonstrated, the public perception pendulum in highly sensitive industries can swing toward the negative (for decades) with only a small number of high-profile failures. Space debris is the type of problem that, left unchecked, poses a Three Mile Island-style risk for the industry. Also like most “waste” related problems, there is often not a strong enough incentive for any single actor to solve it until it is too late. 

The fragmented incentives and control mechanisms in addressing space debris are evident in the current regulatory frameworks. 

The United States is a patchwork of policies, with agencies like the FAA, FCC, and NASA each taking different approaches to limiting or removing space waste. Europe’s approach has been more comprehensive with the European Space Agency (ESA) developing the "Zero Debris Charter," aiming to prevent the generation of new orbital debris by 2030. As of October 2024, 110 countries or entities have joined the charter, and discussions are ongoing with major satellite operators for participation. 

Despite these initiatives, the absence of a binding international legal framework leads to a "tragedy of the commons" scenario, where individual actors may lack sufficient incentives to invest in debris mitigation (opting instead to accelerate commercial advances amid increasing competition, resulting in increased collective risk.

International cooperation around debris is also threatened by geopolitical posturing. Without better visibility and defense mechanisms, nation-states will always have plausible deniability around the destruction of important satellites and space infrastructure (“it wasn’t us, it was debris”). Since even 1mm fragments of debris can disable a satellite entirely, this is not too much of a logical leap.

We believe that solving the problem of space debris creates an interesting wedge for companies to eventually become critical infrastructure for space security and risk management.

The generative AI era presents an interesting paradox: strong confidence in the directional arrow of technological progress (the ever-expanding and evolving LLM blast radius) coupled with significant uncertainty around the macro- and microeconomic implications. Acknowledging this uncertainty, we expect three things to happen as we move toward a world of increased AI and agent usage by organizations and, possibly, a trend towards “AI as the last employee.”

1. Data and information will be processed much faster, leading to real-time insights and decision support. 

2.  The metabolic rate of organizations is set to go up as feedback loops from planning to action become faster.

3. Organizations will face substantially different resource and capital allocation decisions. Both require an orchestration, planning, and decision layer purpose-built for these changing dynamics. 

All of the above requires an orchestration, planning, and decision layer purpose-built for and enabling these changing dynamics. As a result, we see an opportunity to build an AI-enabled business planning platform with substantial optionality to become an integral part of the roll-out and management of increasingly powerful AI systems. Giving organizations on steroids what Cisco called its ‘virtual close’ advantage more than 20 years ago. 

Over the last 20 years, "game tape" and statistical sophistication have revolutionized athletics (and other highly quantifiable professions like enterprise software sales) by enabling precise feedback loops, accountability, and recognition. AI is now driving a similar transformation of the broader professional universe where the relationship between inputs and outputs is often harder to grasp. Professionals have always valued mentorship and coaching. But access has historically been limited by cost and scale (hence “executive” rather than “employee” coaching. AI promises to democratize this type of performance enhancement (and an organization’s ability to measure it) in the same way that companies like Synthesis address Bloom's Two Sigma problem in education. 

Our hypothesis is that “constellations of (AI) experts” – deployed across every facet of professional development and organizational performance – will become as fundamental to career success as mentors and coaches are to elite athletes today. Several converging catalysts are making this possible. 

• The mass market deployment of co-pilots and proto-agents has rapidly normalized AI-human collaboration. More than 60% of physicians now use LLMs to check drug interactions and support diagnosis – early evidence of adoption for high-leverage decision support. 47% of GenZ employees say ChatGPT gives better career advice than their boss – signaling dissatisfaction among young workers with the status quo.

• The proliferation of audio/video capture in professional settings generates rich data to help these systems better understand and improve performance. People increasingly operate with the assumption that every call is recorded, while younger employees regularly go viral for sharing layoff videos online. 

• The economics of AI are reshaping both organizational structures and individual incentives. Companies are shifting from fixed to variable cost models, flexing labor (human and agent) up and down based on demand.  This, in turn, is shifting how workers are measured and compensated. As a result, professionals must proactively adapt to succeed in this new paradigm where human judgment and AI capabilities become increasingly intertwined.

We see several areas where the “constellations of AI experts” will be professionally valuable. In each of these categories, we expect the most successful platforms will combine automated interactions, human experts in the loop, and content/validation that come together to create holistic systems of improvement. 

• Organization-wide solutions that integrate deeply with company context to provide AI-powered coaching and feedback loops. While employees have shown a willingness to trade privacy for better tools, trust and security guardrails are essential. 

• Individual-focused platforms that grow with professionals throughout their careers, combining performance enhancement with credential creation in an increasingly fluid labor market. 

• Solutions for high-turnover industries that capture and distribute best practices to improve training efficiency and retention (e.g. frontline audio-first interfaces)

• SMB owner enablement systems in areas like the skilled trades and family medicine, to make it possible to i) capture and transmit tacit knowledge (streamlining hiring/training while increasing terminal value) and ii) help operators compete without needing to bring in expensive consultants or PE expertise. 

These are, to be clear, highly divergent use cases that necessitate different product philosophies, business models, and competencies from the companies building solutions. However, they share important characteristics, namely that they all represent opportunities to use AI and better data to make professional tacit knowledge, action, and context visible and measurable, unlocking precise intervention to help individuals (and by extension teams and companies) grow into their potential. 

For companies deploying AI in legacy industries, a key question is whether to enable incumbents by selling them solutions or compete with them by taking a more full-stack approach. The trade-offs between these two models is something we have started to explore through our AI-enabled services analysis and this piece on when to compete with and when to sell to incumbents in an industry.

Recently, several firms have shared public theses on the opportunity for emerging AI companies (or vertical market software companies) to capture additional value in a given value chain by fully integrating via the acquisition of assets – as opposed to selling solutions to incumbents or taking a more organic (build instead of buy) approach to going full stack.

Slow, which helped finance the $1.6b acquisition of parking operator SP+ by software company Metropolis, calls the model “Growth Buyouts”. Equal Ventures, which recently opined on the opportunity for such a model in insurance calls it “tech-enabled consolidation”. Vertical market software investor Tidemark calls the approach “tech-enabled vertical roll-ups”. Re:Build Manufacturing calls its technology-driven manufacturing roll-up model an “American Kieretsu”. 

Our current hypothesis is that while the AI-enabled acquisition of services businesses (with venture dollars) may not be wise, there is a significant opportunity for emerging AI, software, and robotics companies to capture more value and develop value chain control by acquiring legacy assets in physical industries. 

For decades, venture capitalists have been involved in what seems like Sysyphosian tasks: digitizing businesses that operate assets in the real world. For many reasons, from software shortcomings to incentive problems and structural challenges in GTM to a lack of skilled labor on the consumer side. We see a trend for novel ML models to solve the first of these problems by being able to operate assets end-to-end without much human input. Yet the latter challenges remain. Therefore, AI-native companies addressing these problems are prone to leave value on the table and, due to slow adoption, likely are slower at training and developing their models, preceding a lot of additional value. Therefore, AI-enabled asset ownership represents one path to achieve this. 

Sequence matters for companies that go down this path. Companies should prove they can build technology and deliver ROI (for early customers or via a smaller-scale organic full-stack approach) before embarking on buying distribution via M&A. The only cases where early M&A can be attractive are cases where smaller targets that are structurally very similar to large-scale targets in the market can be acquired for amounts smaller than traditional GTM. Initially, these businesses have venture risk profiles, and only after the second or third large acquisition should they be derisked and become predictable/repeatable enough for investors with a lower cost of capital – Infra, PE, etc. – to consider participating. By reaching this point, venture investors will have seen highly attractive returns. 

Initial Hypothesis on Key Conditions

Below is an initial hypothesis for when it makes sense for a company to vertically integrate via acquisition as opposed to doing so organically or remaining a software/technology vendor to a given industry:

• The company must have a demonstrated “production advantage,”; i.e., a clear technological or product edge that creates compounding value in an industry. Companies leveraging exclusively off-the-shelf technology likely lack the differentiation to deliver venture-scale outcomes even with strong operational execution and financial engineering. If a PE fund working with Accenture can go after an opportunity or if human labor is cheaper on an efficiency-adjusted basis it is unlikely to be a VC case. If solving the problems requires a combination of world-class technologists AND operators, this becomes an interesting opportunity for venture-style risks and outcomes. 
  

• Customers have proven structurally unable to adopt and deploy a company’s solution to its most productive extent. Alternatively, they seem unwilling to pay for the full value of it. This can be due to various reasons, from lack of scale to incentives leading to stasis in a given market (“if my competitor doesn’t innovate, I don’t have to”). We should be able to pinpoint a structural issue – and generally point to evidence from a company’s own experience –  with a given market/customer base to be certain the ineffectiveness is not a product issue. 
  

• Building on the previous criteria, companies spearheading the buy-out strategy should be building technology that changes the core way an asset is operated, transforming the economics of the business/industry. Most likely that is where existing operators are (somewhat paradoxly) least incentivized to adopt technological disruption. This is what makes the Metropolis acquisition of SP+ such a compelling application of this approach. SP+ has 3,000+ parking locations around the world where the core customer experience (paying for parking) can be largely automated. While the “work around the work” (maintenance, security, etc.) still requires people, the ROI around the primary transaction is much easier to understand than situations where the AI solution is helping people deliver the primary solution more efficiently (e.g. home services models, legal services, etc.). 
  

• Likely, there is a sweet spot around the level of complexity that goes into operating an asset that makes it a fit for AI-enabled acquisition. Complexity can either stem from the core value proposition being complex, several offerings being performed at the same asset leading to compounded complexity or the “work around the work” being significant (e.g., for regulatory reasons). Too little complexity at the core value proposition becomes a PE case; too much and the operational overhead reduces the leverage associated with improving the margins of the core asset. Ideally, the complexity/problems across holdings within the same space should be the same (e.g., parking lots), and skills easily transferable. We should be able to pinpoint these levels of complexity and identify assets/problems where they meet the sweet spot. 
  

• The category a company is operating needs to have acquisition targets that are operating at scale (ideally businesses worth USD 1B+ with additional value creation in the several hundred million – further analysis on this needed). Buying assets operating at scale that can be fully optimized and automated via AI is substantially more attractive than rolling up locally-delivered services businesses. Again, this is what makes the SP+ acquisition so attractive, SP+ has 3,000+ parking locations around the world that likely are all run very similarly. Ideally, solutions deliver not only cost savings but also growth opportunities. We are also interested in companies with a view on how the integration of software and legacy assets will unlock increasing ecosystem control and turn the business into an industry operating system. 

• Companies must have advantaged access to talent across functions. It is rare for a founder or founding team to understand “what great looks like” in areas where they have not had direct experience. A team of software engineers is likely unfamiliar with what makes a great industrial CFO or service-business COO. As a result, we may expect the pool of founders well-equipped to build such a business to be small. We have seen this play out at companies like KoBold Metals, which combine highly scientific founding teams with business acumen. 

These criteria still don’t fully answer why/when it is better to grow a full stack solution via acquisition rather than a more organic approach. One primary reason a company would choose to grow via acquisition is if the geographic footprint and surrounding “infrastructure” of an industry will look largely similar in the future as it does today. In such cases, the speed of distribution created by acquisition is enough of an advantage to overcome the accompanying cultural and organizational complexity that could be mitigated with a more organic strategy.

To use the Metropolis example, should we expect the footprint of the parking industry to be more or less the same in 10 years as it is today? While autonomous vehicles may make some impact on the margin during that time period, the inertia of the built environment probably means we should expect the flow of traffic and parking to remain relatively the same (airports, stadiums, commercial centers, etc). 

A counter-example is H2 Green Steel, which has raised multi-$B to produce steel with 95% lower emissions than traditional steelmaking. Because of the fact that the company’s steel production depended on access to ample clean energy, the company couldn’t just acquire and transform underperforming steel facilities despite the similarity in equipment, permitting, and human capital needs. Thus, to transform the industry around their vision, the company was forced to build a more organic approach. 

Companies also might pursue a buy instead of build strategy when the technology can be easily integrated with existing assets and infrastructure, substantially reducing time to value for a given solution. 

There are likely several other criteria in support of (and against) the strategy of vertically integrating via acquisition which need to be explored in further depth. 

Across the world, advanced policing systems are delivering for forward-thinking law enforcement and border control agencies. Solutions like Flock solve over 700,000 crimes annually, making departments more efficient, while drones drive faster and safer responses. As resistance to US technology persists, expanding EU internal security budgets and increasing evidence that these systems work will push Europe to seek out homegrown solutions.  

Fire and disaster response, helping mitigate the €77b in annual losses from natural disasters and protect human livelihood, represents another critical opportunity area. New entrants combining predictive modeling of weather and the built environment with proactive intervention capabilities will capture value by closing Europe's critical response gaps.

Finally, intelligence agencies are approaching a breaking point: drowning in multi-modal data (from video to financial transactions) while inter-agency coordination fails. Companies that bridge European fragmentation while navigating privacy mandates will become essential infrastructure, enabling real-time response to physical and digital threats.

We see an opportunity for a new class of European "Public Safety Primes" to establish themselves in the market. The Axon playbook – now a $45b+ company built through methodical expansion by expanding from tasers to body cameras to a comprehensive digital evidence platform – shows what's possible. The company has effectively zero penetration in Europe, and local players like Reveal Media and Zepcom remain subscale. Winners will start narrow with a must-have product, earn trust through measurable impact, and expand across the public safety landscape as system-wide advantages compound.

The unbundling of the firm has been in flight for decades. As the internet enabled increased access to global labor markets, outsourcing to lower-cost countries exploded. The proliferation of cloud computing and mobile took this a step further, making it possible to externalize an increasing number of key operational functions and allowing for more asset-light business models. This prompted a thesis several years ago that the rise of “One Person Companies” remained an underrated opportunity. 

The next step in the evolution of the firm will build on this but will come at the problem from a different direction. It will be defined by the rise of One-Person Companies. Creators and knowledge workers will access external services that provide the capabilities to start and scale a firm and then re-bundle them in unique ways around their core skill set. They will monetize by selling products, services, and expertise to an owned audience that their core skill set has helped them build.

New platforms and infrastructure providers will emerge to support the tens of millions of individuals capable of building successful One-Person Companies along with the billions of consumers and businesses that will support them. More generally, the rise of the One Person Companies will inject dynamism into the broader economy and will play a role in driving more inclusive innovation.

AI – particularly agentic solutions capable of proactively understanding and executing end-to-end business workflows – represents the next leap in this evolution. As several investors and operators have observed, AI is empowering small groups more than ever before and new businesses across the economy (i.e. not just tech startups) are building from inception with AI literacy as a core competency. According to Gusto, roughly a fifth of businesses created last year said they were using generative A.I. to more efficiently carry out things like market research, contract reviews, bookkeeping, and job postings.

Current and Future Market Structure

In complex, non-commodity service categories like strategy consulting, law, investment banking, and wealth management – where key individuals inside of large companies often already “run their own book” – we believe these forces create the opportunity for further fragmentation; i.e the “creator economy-ization” of professional services.

A study cited in a 2015 Forbes article about million-dollar solo consulting businesses indicates this opportunity is not new. 

The U.S. Census Bureau found that of 3.2 million "nonemployer" professional services businesses in the U.S., there were 7,276 firms that brought in $1 million to 2.49 million in revenue in 2013, the most recent year for which statistics were available. And 321 superstars brought in $2.5 million to $4.99 million.

For the sake of simplicity throughout the document, we will refer to these companies as Service OPCs, though there is of course no reason why it must be a single person.

In practical terms, we believe we are entering a period where an even larger number of individuals or small teams with a differentiated skill set or value creation mechanism (product) will increasingly be able to leverage the marketplace (instead of “the firm) for distribution and operational capacity to build profitable and durable OPCs.

This thesis rests largely on the idea that some elements of human judgment are inherently non-scaleable / automatable (similar to our thesis around where value is captured in AI-driven content creation) and thus that the dynamics of the market will tend more towards specialization – thousands of small, profitable “winners” – rather than winner-take-all. 

A services Mittelstand rather than Silicon Valley concentration.

We are interested in what the agentic technologies that drive end-to-end workflow execution will look like and what the coordination mechanism across those autonomous services will be for Service OPCs. Without both of these things becoming a reality in parallel, the overhead of identifying and managing dozens of end-to-end AI agents (some of which will inherently be more end-to-end than others) while growing a client base and playing the most important role of delivering the differentiated service (even if some elements are made more efficient through automation) is likely enough to push promising OPCs back into the arms of McKinsey or Kirkland Ellis.

Effectively, we believe there is a Shopify-like opportunity to “arm the rebels” and build an ecosystem-level operating system for the AI-driven services transformation – combatting empire-building incumbents who view AI as a solidifier of their market positioning and what are sure to be dozens of overfunded venture-backed businesses promising to be “the next Goldman Sachs”.

Product and Technical Hypothesis

By engaging at the aggregation and coordination level, we are interested in answering the question of how a platform might “package” a business around an OPC’s core skill set to help it grow beyond its pre-AI agent potential. 

While we want to avoid being overly prescriptive in our analysis at such an early stage, we believe that for such a platform to represent a comprehensive – and attractive – alternative to the firm for Professional Service OPCs, it would possess some or all of the following characteristics (features), listed roughly in order to how they might be sequenced from a product perspective:

1. Functional Automation (Operational Capacity) – This pillar would serve as an "Agent Store," featuring both proprietary workflows and third-party end-to-end AI agent solutions. It would offer OPCs end-to-end functional agents for various business operations, such as:

• Contract management

• Financial management and forecasting

• Compliance and risk assessment

• Resource allocation and project management

• Continuous learning and skill development

• Marketing and public relations

• Legal execution

It is also interesting to consider how such a store could provide a distribution channel for third-party developers of specialized AI solutions like Devin (for software development) or Harvey (for legal services) or the seemingly dozens of AI agent companies launching each week (a quick scan of the most recent YC class highlights how prevalent this model has become for early stage companies. 

These developers would be incentivized to use the platform due to its focus on going beyond simply offering access to agents but, helping OPCs “package” a specific set of agents around the skills and ambitions of the company, which brings us to the next pillar of the platform. 

2. Organizational Coordination (The AI COO) – The AI COO acts as the central nervous system of the OPC, ensuring all parts of the business work together seamlessly. Key functionalities include:

• Automated integration between functional agents (the Bezos API Mandate on overdrive)

• Workflow optimization across all business functions

• Stakeholder communication management

• Strategic decision support

• Continuous improvement engine for business processes (i.e. vetting and implementing improved solutions or workflows autonomously). 

This pillar is critical in attracting and retaining both OPCs and third-party AI solution providers. For OPCs, it offers unprecedented operational efficiency and is the core enabler of leaving the firm behind for good. For AI solution developers, it ensures their tools are integrated effectively into the OPC's operations, maximizing utility and, long-term revenue potential.

With these three pillars working together, such a platform aims to create a robust ecosystem that not only supports individual OPCs but also fosters a network of AI solution providers. This symbiotic relationship between OPCs, the platform, and AI developers has the potential to drive rapid innovation cycles and expand the market in the same way Shopify has done in e-commerce for physical goods. 

Antithesis

While we have a reasonable degree of confidence that the end state of the unbundling of the firm will look something like what we have laid out above (“Shopify for professional services” is likely a primitive analogy for what we will have in 2050), there are several reasons to be wary of the thesis. Much of this hinges on market timing as well as the common question of whether this will enable truly novel business models to emerge that incumbents are structurally unable to compete with.

• We may be underestimating incumbent entrenchment, particularly around trust and social signaling, and their ability to adapt. “Nobody got fired for hiring McKinsey, Goldman, etc.”. While not apparently on the operational level (yet), incumbent consulting firms have been among the biggest beneficiaries of the generative AI explosion. 

• Regulatory, compliance, and legal structures may change slower than the technology,  Sectors like law and finance are heavily regulated. OPCs might face significant hurdles in meeting compliance requirements without the resources and infrastructure of larger firms, potentially limiting their ability to operate in certain (high-value) areas.

• The complexity of integration (i.e. the AI COO) may be substantially more complex than we have described. The reality of seamlessly integrating multiple complex AI systems could be far more challenging and error-prone than expected, leading to inefficiencies or significant mistakes.

As Mark Rubenstein wrote in “The Business of Benchmarking”, universal standards are usually unassailable. The risk for companies that manufacture them is less that their moat is crossed and more that their castle becomes irrelevant. We believe the current geopolitical, economic, and technological moment is creating a once-in-a-generation opportunity to successfully counterposition and eventually (finally!) displace the global ratings and benchmarking oligopoly.

Several forces are converging to create this opportunity. First, Great Power Competition is fundamentally reshaping global trade and industrial activity. The push for energy independence, secure supply chains, and strategic autonomy is driving massive investments in decarbonization and reindustrialization. Reconfigured trade flows and industrial priorities demand new frameworks for understanding risk and opportunities. Second, the growth of sensor networks, connected devices, and geospatial systems has created unprecedented visibility into physical world operations and trade flows. This proliferation of data  – from factory floors to shipping lanes – provides granular, real-time insights that were previously impossible to capture. Finally, advances in AI and machine learning allow us to process and derive meaning from complex, multi-modal data at the scale and speed demanded of modern trade. 

We've seen the fundamental transformation of key commodity markets firsthand through our investment in KoBold Metals. Better collection and analysis of physical world data is revolutionizing resource discovery and development. Meanwhile, geopolitical machinations are accelerating the reconfiguration of global supply chains and trade routes, creating urgent demand for new frameworks to understand and price physical world assets. Traditional frameworks – built for a different era of global trade – are increasingly misaligned with markets that require real-time, granular insights to make decisions.

Success in this market isn't about attacking the incumbent oligopoly directly. Through counterpositioning, the opportunity lies in building for the new industrial economy with a model native to the speed and complexity of modern trade. Winners will start narrow, building density of data and trust in specific verticals, before sequencing alongside their customers' evolving needs to develop new pricing and risk infrastructure for the physical economy.

Today, the AI boom is drawing parallels to the 1990s dot-com and telecom bubbles, with massive capital expenditures, promises of exponential growth, and early productivity gains. On the horizon, the potential of general-purpose robotics even resembles the iPhone-driven mobile revolution that followed almost a decade after the dot-com bust. 

But the differences between the two eras are equally striking. Today, incumbent technology companies possess more structural power over the ecosystem than 30 years ago, suggesting perhaps less overall volatility at the expense of dynamism – i.e. The Magnificant Seven serve as “hysteria dampeners” thanks to massive balance sheets and distribution networks. And while opportunism in AI is certainly present, the technical barriers to entry needed to build a competitive foundation model (and the pace of customer feedback) are substantially higher than building an ISP during the DotCom frenzy.

However, the most consequential difference between the two eras may be the central role of energy – namely the re-emergence of nuclear power – in today's AI boom, particularly with the backdrop of rising great power competition and the ever-present specter of climate change.

Unlike the telecom infrastructure of the dot-com period (and data centers in today's cycle), which serve singular purposes, the expansion of nuclear infrastructure addresses multiple critical challenges. First, it promises to play a significant role in solving for the energy intensity and reliability demands of AI data centers. This is a problem we are looking at from several angles – nuclear and other forms of production, efficiency (via our industrial action layer research), and finally via an exploration of better distribution and grid resilience technologies.  

Beyond serving AI data centers, nuclear power (along with the other categories highlighted), meets the vast need for clean baseload power to accelerate decarbonization and the push for increased energy security amidst heightened geopolitical risk. 

As a result, nuclear’s positioning as a one-size-fits-all solution to many of our most pressing concerns – and thus its (theoretical) resilience to the fortunes of any single macro factor – makes it an attractive “picks and shovels” opportunity perfectly in line with the three major economic supercycles of the next several decades (AI, climate, and national security) – provided the current momentum can overcome generations of cultural and political baggage.

This baggage is complex, equal parts social, economic, and regulatory with each reinforcing the other in a vicious cycle. High-profile accidents and proliferation risk have dominated the social consciousness for 40+ years. This, in turn, influences regulation which increases the red tape related to the development and deployment of safer, more effective modern solutions. As process knowledge is lost and talent leaves the industry, costs spiral even higher and we are left with the current state of affairs. 

Despite its history as a safe, clean, and cost-effective technology, nuclear costs have jumped 33% higher while the cost of new solar generation continues to plummet (over 90%). The latter is, to be clear, a positive – we are pro-energy abundance more than we are wedded to a single approach – and showcases the power of technological learning rates if unleashed. 

Current and Future Market Structure 

Today, the narrative – and the fundamentals – around advanced nuclear are finally shifting towards the positive across several dimensions.

• Russia’s full-scale invasion of Ukraine provided a clear focal point for the importance of energy security and the role nuclear energy can play in decoupling from Russia. Between 2021 and 2022, the percentage of Europeans at least somewhat in favor of nuclear energy jumped substantially – in Germany from 29% to 44%, in France from 45% to 55%, and in the UK from 39% to 53%.

• Energy has become the core bottleneck to scaling AI. While GPU scarcity dominated the preceding couple of years, everyone from Mark Zuckerberg to Sam Altman believes the next substantial step forward requires energy-related breakthroughs. As a result, big tech companies have become the “urgent buyers” necessary to drive market transformation. Microsoft’s actions signal a clear belief in the long-term need to expand nuclear capacity. Its recent 20-year power purchase agreement with Constellation, which will revive the Three Mile Island facility, is as symbolically important as it is economically.   

• The capital markets are responding swiftly to this step change in demand, with financial institutions including BofA, Morgan Stanley, and Goldman backing a COP28 goal of tripling nuclear capacity by 2050. The commitment to triple capacity also came with support from the US, UK, Japan, Sweden, and the United Arab Emirates.

• Regulatory support has not been limited to COP commitments. In the United States, for example, President Biden recently signed into law the ADVANCE Act, which aims to streamline licensing, promote R&D investment, and contribute to workforce development.

• This follows on the heels of (tepid) progress on the deployment front. In the United States, Vogtle 3 and 4 are finally online, years late and billions over budget. Still, the finalized designs, established supply chains, and trained workforce should contribute to less expensive future deployment. This summer, TerraPower began construction on its advanced reactor facility in Wyoming. Meanwhile, SMR certification is building momentum in France as companies like NewCleo and Jimmy Energy look to deploy first-of-a-kind reactors by 2030.  
  

• Finally, the characteristics of SMR and AMRs coupled with the aforementioned demand dynamics have ignited venture interest in the space. Smaller form factors that can be deployed more flexibly and with a broader range of financing options have eased some concerns about the venture scalability of such projects. As a result, dozens of companies have been funded in recent years. Today, over 100 SMR and AMR designs are being developed at various stages and with different timelines across the world. 

Key Early Assumptions / Potential Catalysts 

The improved environment around nuclear power leads us to a few critical questions, based on important assumptions about where scarcity and abundance sit in the nuclear value chain.

• Assumption 1 → The timelines for most advanced nuclear projects are at least 7+ years out, likely longer if history is a guide. This may not align with our time horizon unless we can identify intermediate value inflection steps that create opportunities for exit, etc. similar to the life sciences ramp-up.

• Assumption 2 → The crowded nature of SMR/AMR technology development (abundant capital and attention at that part of the nuclear value chain) and the lack of a clear set of winners should push us to areas of relative scarcity where solving key bottlenecks can accelerate the market overall (i.e. turning those hundreds of companies into customers immediately).

So where is there scarcity in the market that aligns with our investment horizon and approach? Three areas stand out initially, with each meriting a separate deeper dive should conversations with experts, operators, and founders push us in a certain direction. 

Fuel → Russian (and increasingly Chinese) dominance of key fuels and processing steps risks putting the West in a vulnerable position, which echoes the overreliance on cheap gas that contributed to the Ukraine invasion and the European energy crisis. Significant efforts are underway to de-risk the nuclear fuel supply chain. The US Congress passed a bill to limit the import of Russian uranium, while the “Saparro 5” (Canada, Japan, France, the UK, and the US) announced plans to jointly invest over $4b to boost enrichment and conversion capacity over the next three years. 

The biggest risk to decoupling from Russia has been HALEU (high-assay low-enriched uranium), which advanced reactors are being developed to run. Until the Russian invasion of Ukraine, Russia was the only place with a plant licensed to produce this material. Companies like Centrus Energy and a still-under-the-radar Founders Fund-backed startup are targeting this bottleneck, which could be an important enabler of the broader advanced nuclear space. 

Project Development → Over the last two years, much of my work has centered on how to best help “emerging industrial” companies scale more effectively. While my early assumptions were largely centered on the need for new financing models, the critical bottleneck to deploying new projects across energy, materials, and manufacturing often turned out to be capable of on-the-ground project development and execution. Given the deterioration of the nuclear talent base across most Western countries, this problem is even more severe. 

A key problem with effective (i.e. on time, on budget) project development is the fragmentation of the subcontractors needed to build a project end-to-end. Companies are aiming to solve this through a reactor-agnostic platform for nuclear project execution. Through a comprehensive oversight strategy, which includes taking direct control over supply chain management, sourcing, workforce coordination, and financing required for constructing power plant fleets the company hopes to do for nuclear what SpaceX did for launch. Others are building fully modular, factory-made systems innovating on the delivery and development model rather than at the reactor level. 

Waste → Waste remains perhaps the most politically and socially charged aspect of nuclear energy, leading to decades of information warfare despite the relatively boring nature of the problem. Historically, the commercial incentives to store waste weren’t particularly attractive, making it a largely political endeavor. 

Today, countries and companies around the world are starting to see opportunities to turn waste from a burden into an opportunity through recycling.

We think the most interesting venture cases will emerge in the context of 1. consolidation and 2. maintained concentration. In the context of 3, it is interesting to explore the next-order effect of this, (see Marketplace for AI-enabled Services and Products)

Independent of the structural market outcomes, the occurrence of AI-enabled service businesses requires specific conditions to emerge and thrive. We differentiate two types of requirements.

First, necessary conditions are always necessary for these businesses to pursue opportunities. Many of these opportunities, however, may become commoditized, leading to numerous profitable but modestly sized businesses, typically in the $10 million revenue range (i.e., fragmentation).

Therefore, for market outcomes 1. and 2. and venture-scale outcomes to occur, opportunities must have the potential for significant competitive advantages, or "moats." These moat conditions are likely present in only a small subset of AI-enabled opportunities.

Our primary objective is to identify and focus on the most promising categories where both the necessary and moat conditions are met. These categories represent the most attractive opportunities for substantial growth and success in the AI-enabled service sector.

Necessary Conditions

• Objectifiable Outcomes: Objectifiability in outcomes is crucial to 1) training the models and 2) lowering transaction costs with customers. 

• Easy to Hand Off from/to customer: Easy hand-off is critical to lower transaction costs and make sure the business can scale without integration work, etc. 

• Technology Maturity: Technology utilized to deliver services needs to be sufficiently mature, or there needs to be a clear path for technology to be mature. In the beginning, human labor supporting the delivery of services is fine, but there needs to be a clear path to attractive unit economics with close to 90% automation. 

• Value-based Pricing Possible: The thesis depends on a company's participation in the upside of what is 1) generated at a lower cost or 2) made possible due to a new service. It is critical for the economics that the service provider can sufficiently capture the value generated to ensure top-notch margins that are improving as the technology matures. 

Moat Conditions (at least one or two of them need to be true)

• Enabling Technology Moats: Using off-the-shelf LLMs will not lead to sustained technology moats. Technology moats will emerge where the quality of the service offered relies on developing and integrating with other types of ML / AI, which will lead to some initial barriers to entry from a technical perspective.

• Improvements in Technology from Feedback Loops: Building on the previous point, ​​another source of possible moat is that technology improves through feedback loops, as services are delivered to customers and lessons are learned. This means that the market leader can improve its technology fastest, leading to winner-takes-most situations. 

• Sustained Unique Customer Access: Efficient customer acquisition for a lot of these businesses will be key to delivering top-notch profitability margins in the long term. Those categories where certain types of companies have unique access to customers (e.g., due to industry pedigree) are likely to be attractive. Especially when paired with the technology feedback loop condition outlined. 

• Liability Underwriting: The assumption is that many of these service businesses will have to take liability for correctly delivering these services. Suppose liability is a crucial aspect of the offering. In that case, the amount of risk that can be taken in this context is a function of the cash a company has on its balance sheet to serve as a buffer for potential failures and can, therefore, be more aggressive. 

• Regulatory Moat: Benefit from licensure requirements and other regulatory hurdles, which act as a natural barrier to entry and provide a stamp of trust and credibility. However, it is unclear whether this is actually the case. Lawyers require licenses, but barriers to entry are fairly low and based on academic performance. If the underlying models are similar, won’t everybody or nobody get approved? 

• Brand / Trust: Services businesses inherently are built on a high degree of trust and brand recognition. This trust and brand enable customers to know that somebody can be held accountable and their choice isn't questioned by their bosses or investors (e.g., hiring McK, BCG, or top 4). It is likely that the same dynamics play out here and that this can be a source of sustained competitive advantage.