The Pre-IPO Mirage: Why Neuralink and Anduril Represent Structural Fragility, Not Innovation

The announcement that MSX's Pre-IPO III fund has taken positions in both Neuralink and Anduril was met with the usual chorus of approval. The narrative writes itself: from the human brain to the battlefield, AI's final frontiers. But as someone who has spent the better part of three decades dissecting the gap between cryptographic promise and operational reality, I see something else. I see a fund manager constructing a narrative portfolio designed for LP acquisition, not for technical diligence. The front-runner didn't get ahead by being first; they got ahead by understanding the latency of the system. And in this case, the latency is between the hype cycle and the actual balance sheets. Let me be precise about what we know. The original analysis, which I have reviewed, contains exactly two information points: MSX has introduced Neuralink and Anduril into its Pre-IPO III fund, and the stated thesis is 'AI's next frontier.' That is it. No technical details. No valuation breakdowns. No exit strategy. The rest of the report is extrapolation based on public knowledge about these two companies. This is not diligence; this is narrative construction. And narrative construction, in my experience, is the first sign of a fragile investment thesis. Context is critical here. We are in a bull market for AI-related assets. The SEC's regulation-by-enforcement approach has created a vacuum where narrative often substitutes for substance. In this environment, Pre-IPO funds are proliferating, each claiming access to the next big thing. The MSX fund is positioning itself at the intersection of two of the most emotionally charged sectors in technology: brain-computer interfaces and autonomous defense. Both sectors have massive TAM narratives. Both have governments and militaries as potential customers. Both have founders with cult-like followings. And both have fundamental technical and economic questions that remain unanswered. Let me start with Neuralink. The technology is impressive in its ambition. The N1 implant with 1,024 channels, the R1 surgical robot that can avoid blood vessels, the FDA breakthrough device designation. These are real achievements. But let's look at the numbers. As of late 2025, there are exactly three human patients. The company has no revenue. Its valuation is estimated between $8 billion and $10 billion. That is a P/S ratio of infinity, and a P/R&D ratio of roughly 80 to 100 times. This is not an investment; it is a lottery ticket with a very specific set of technical risks. The core issue is the transition from proof-of-concept to production. In my 2017 audit of the EOS codebase, I identified a race condition that could allow infinite token minting under specific block producer configurations. The response from the community was dismissive. They were focused on the price action, not the code. Neuralink faces a similar dynamic. The technical community is focused on the promise of high-bandwidth neural interfaces, but the reality is that invasive BCI technology has a long history of failure. The electrodes degrade. The body rejects the implant. The signal attenuates over time. These are not hypothetical concerns; they are documented phenomena in the neuroscience literature. Neuralink's own data, limited as it is, has shown signal degradation in at least one of its three patients. And then there is the competitive landscape. Synchron, with its non-invasive Stentrode, has already implanted 10 patients. Precision Neuroscience has 30. These companies have less bandwidth, but they have significantly lower risk profiles. The market may not care about 1,024 channels if the alternative is a safer procedure with 16 channels that actually works reliably. A bug is just a feature that hasn't been exploited yet, and in the case of invasive BCI, the bug is the body's immune response. Now let's turn to Anduril. This is a different beast. The company has real revenue, projected at $1 billion for 2025, with a year-over-year growth rate of approximately 150%. It has contracts with the US Department of Defense and allied governments. Its Lattice software platform is genuinely innovative, representing a shift from hardware-centric to software-defined defense. The acquisition of Callum for $1.3 to $1.4 billion adds manufacturing capability. On the surface, this looks like a more rational investment. But let's dissect the fragility. Anduril's valuation is estimated between $14 billion and $28 billion. At the midpoint, that is a P/S ratio of roughly 21 times. Palantir, its closest comparable, trades at 30 to 50 times sales. So Anduril is cheaper on a relative basis. But this ignores the concentration risk. Anduril's revenue is heavily dependent on US defense contracts. If the political winds shift, if the defense budget is cut, or if a traditional player like Lockheed Martin or RTX successfully pivots to a software-defined model, Anduril's growth narrative collapses. The company is also facing ethical and regulatory headwinds. The development of autonomous weapons systems is a lightning rod for international criticism. The UN has been debating a framework for lethal autonomous weapons for years, and while no consensus has been reached, the direction of travel is toward greater regulation. There is also a deeper structural issue. Anduril's Lattice platform is designed for deterministic, real-time decision-making in contested environments. This is fundamentally different from the general-purpose, creative AI that companies like OpenAI are building. The partnership with OpenAI for counter-UAS systems is interesting, but it also highlights the tension between these two paradigms. Can a system that is designed for creativity be safely integrated into a system that requires absolute determinism? In my 2025 analysis of AI-crypto integrations, I identified a flaw in the Chainlink API design that allowed AI models to manipulate price feeds through synthetic data injection. The same class of vulnerability applies here. If an AI model can be tricked into making a wrong decision in a financial market, it can be tricked into making a wrong decision on a battlefield. The stakes are higher, but the fundamental problem is the same: AI systems are not reliable in high-stakes, adversarial environments. The contrarian angle, and I always try to find one, is that the bulls might be right about the long-term direction. The convergence of AI with biology and defense is inevitable. The question is not whether it will happen, but when and at what cost. Neuralink's technology, if it can overcome the safety and durability hurdles, could revolutionize the treatment of neurological conditions. Anduril's software-defined approach could make defense systems more responsive and cost-effective. The potential is real. But potential is not a business model. In my 2021 analysis of Axie Infinity, I calculated that the protocol's treasury was insufficient to cover potential sell-offs, estimating a 90% crash probability within 18 months. The community downvoted me into oblivion. But the crash happened. The same dynamic applies here. The valuations of Neuralink and Anduril are based on a future that may not arrive on the timeline that the Pre-IPO fund requires. The takeaway is not that these are bad companies. It is that the investment thesis is structurally fragile. MSX is not investing in technology; it is investing in a narrative. The narrative is compelling, but narratives are not balance sheets. The fund's exit strategy depends on IPOs that may be delayed, valuations that may correct, and technologies that may fail. The LP who invests in this fund is not buying exposure to AI's frontier; they are buying exposure to the fund manager's ability to exit before the music stops. And in a bull market, the music always stops. The question is whether the exit door is open when it does. Based on my analysis, the door is not just closed; it is locked, and the key is in the hands of FDA regulators, Pentagon procurement officers, and the unpredictable dynamics of the human immune system. I would not bet on that combination. But then again, I have been called a cold dissector. I prefer to call it a realist.