Portfolio Commentary
Beyond the AI Bubble
Executive Summary
Artificial intelligence has become one of the most polarizing investment themes in global markets. Comparisons with the dot-com bubble have become increasingly common as private AI valuations reach unprecedented levels and billions of dollars continue to flow into large language model (“LLM”) developers and AI infrastructure.
This has naturally led some investors to ask whether our portfolio, comprising Hologen, Partex, Harbinger Motors, RCB Nanotechnologies, and AllJoined, faces the same risks.
Our view is no.
- They are not investing billions to train general-purpose AI models.
- They are not dependent on selling AI compute.
- They are not competing to build the next ChatGPT.
Instead, AI serves as an enabling technology within businesses built around proprietary assets, industrial capabilities, scientific expertise, or specialized datasets. Their long-term value depends primarily on solving real-world commercial problems rather than winning the race to build the largest AI model.
Consequently, although market sentiment toward AI may fluctuate, we believe these companies are materially less exposed to the factors currently driving concerns about an “AI bubble.”
| Characteristic | Frontier AI / LLM Companies | Our Portfolio Companies |
|---|---|---|
| Primary Business | Build and commercialize foundation models (LLMs) | Use AI to improve products in healthcare, industrials, transportation and neuroscience |
| Role of AI | AI is the product | AI is an enabling technology |
| Revenue Source | API usage, subscriptions, enterprise AI services | Drug discovery, diagnostics, EV manufacturing, industrial recycling, neuroscience tools |
| Competitive Advantage | Scale of computing infrastructure and model performance | Proprietary data, patents, manufacturing, scientific know-how, specialized hardware |
| Capital Requirements | Multi-billion dollar AI training and data-center investments | Capital directed toward laboratories, manufacturing, clinical development and commercialization |
| Exposure to AI Spending Cycle | High | Moderate to Low |
| Dependence on AI Hype | Significant | Limited. Commercial value driven by industry-specific demand |
| Valuation Drivers | Future AI adoption and market leadership expectations | Technology, IP, commercial partnerships, assets and execution |
| If AI Sentiment Weakens | Valuation multiples could compress materially | Business fundamentals remain tied to underlying end markets |
| Investment Thesis | Success depends on becoming a dominant AI platform | Success depends on solving real-world problems in large industries |
The AI Bubble Debate
Today, concerns surrounding an AI bubble generally focus on a relatively narrow group of companies:
- Frontier LLM developers
- AI infrastructure providers
- GPU manufacturers
- Data-center developers
- Companies whose valuations are based primarily on expectations of future AI demand
These businesses require enormous capital expenditure, consume vast computing resources, and often trade at valuations that imply decades of future growth. This is where comparisons with the dot-com era originate.
However, applying this same framework to every company using AI risks oversimplifying an increasingly diverse technology landscape.
“AI Companies” and “Companies Using AI”
One of the biggest misconceptions in today's market is treating AI as a single investment category. It is not. There is a significant distinction between companies whose entire business is building AI models and companies that use AI to improve products or services.
Think of electricity. The companies that built power plants were very different investments from manufacturers that later used electricity to improve production. Likewise, cloud computing created enormous value, but not every successful software company had to own a data center.
Artificial intelligence is following a similar pattern. Our portfolio largely consists of companies that apply AI rather than sell AI itself.
Key message: Our investments are application-layer businesses that leverage AI as a tool, not frontier AI companies whose value depends primarily on developing increasingly larger language models.
Why Our Portfolio Is Different
Although each investment operates in a different industry, several common characteristics distinguish them from companies typically associated with AI bubble concerns.
AI Is an Enabler, Not the Product
None of these companies generate revenue by selling large language models or AI computing capacity. Instead, AI enhances an existing commercial business. For example: Hologen applies AI to drug discovery and diagnostics. Partex integrates AI into pharmaceutical research. Harbinger leveraged its commercial vehicle manufacturing and autonomy software to enter unmanned defense ground vehicles. The company licenses its autonomy software to third parties too, and is now building generators for data centers re-using its electric motors from commercial vehicles. AllJoined develops neural decoding technology supported by AI algorithms. RCB Nanotechnologies is fundamentally an advanced materials company rather than an AI company.
If AI progress slowed tomorrow, each business would still retain its underlying assets, intellectual property, customer relationships, and commercial opportunities.
Proprietary Assets Create Barriers to Entry
Much of today's generative AI ecosystem relies on models that can increasingly be replicated. Our portfolio companies derive competitive advantage from assets that are substantially harder to reproduce. Examples include proprietary clinical datasets, patented industrial processes, manufacturing capabilities, pharmaceutical partnerships, specialized scientific expertise, and unique hardware platforms.
These advantages are accumulated over years rather than purchased through additional computing power.
Revenue Comes From Real Industries
Each company serves an established end market: healthcare, pharmaceuticals, transportation, industrial recycling, neuroscience. Demand in these sectors exists independently of investor enthusiasm for AI.
The commercial question is therefore not “Will AI remain popular?” Instead, it is: “Can these businesses create measurable value for customers?” That is a fundamentally different investment proposition.
Valuations Reflect Company-Specific Fundamentals
Many of today's highest-profile AI companies command extraordinary valuation multiples because investors expect dominant future market positions. Our portfolio companies were acquired at valuations that we believe better reflect their stage of development. In several cases, investments were made through negotiated secondary transactions rather than highly competitive financing rounds.
While valuation alone never eliminates investment risk, purchasing businesses at more moderate entry prices provides a larger margin of safety than investing after valuations have already expanded dramatically.
| Company | Core Industry | Primary Value Driver | Role of AI |
|---|---|---|---|
| Hologen | Healthcare & Biotech | Proprietary NHS clinical data, diagnostics, drug development | Accelerates biological discovery |
| Partex | Drug Discovery | Healthcare datasets, pharma partnerships, IP | Identifies therapeutic opportunities |
| Harbinger Motors | Commercial EVs, Unmanned Defense Vehicles & Generators for Data Centers | Mission-critical vehicle platforms; picks and shovels manufacturer for the defense technology and defense AI ecosystem | Power for AI data centers, plus Physical AI in defense drones and commercial EVs |
| RCB Nanotechnologies | Circular Economy | Patented recycling technology and industrial process | Minimal; operational optimization |
| AllJoined | Neurotechnology | Brain-computer interface hardware, proprietary datasets | Neural decoding algorithms |
Lessons From the Dot-Com Bubble
The dot-com bubble is often remembered as proof that technology investing is dangerous. History tells a more nuanced story. Many internet companies failed because they lacked viable business models. Yet the internet itself transformed the global economy. Amazon, Google, Salesforce, and numerous other technology leaders emerged after the bubble burst.
The lesson is not that every technology investment fails. Rather, markets eventually distinguish between businesses built on speculation and businesses built on sustainable economic value. We believe the same distinction will likely emerge within artificial intelligence.
Some AI companies will disappoint. Others will become enduring businesses. Our investment process seeks to identify companies whose value is supported by intellectual property, commercial partnerships, differentiated technology, and industry expertise, not merely excitement surrounding AI.
| Dot-Com Bubble Characteristics (1999 to 2000) | Our Portfolio |
|---|---|
| Internet itself was the investment thesis | AI is only one component of the investment thesis |
| Companies often had little or no revenue | Companies are building products for established industries |
| Valuations driven primarily by internet excitement | Investments made at negotiated private-market valuations |
| Limited barriers to entry | Proprietary IP, scientific expertise and specialized assets |
| Business models often unproven | Clear commercial applications in healthcare, mobility and industrial technology |
Conclusion
The current debate surrounding an AI bubble is both understandable and healthy. Markets periodically become overly enthusiastic about transformative technologies, and artificial intelligence is unlikely to be immune.
Nevertheless, we believe it would be inaccurate to view our portfolio through the same lens as frontier AI laboratories or companies whose valuations depend primarily on continued enthusiasm for large language models.
Our investments are concentrated in businesses where AI is a powerful tool, not the sole source of value. Their competitive positions are supported by proprietary assets, industry expertise, intellectual property, commercial relationships, and exposure to large real-world markets. While no early-stage investment is without risk, we believe these characteristics provide a fundamentally different risk profile from the companies most frequently cited in discussions of an AI bubble.
panaceavc.com/research