Internal Memo: To All Portfolio Managers, Re: The $500 Billion Hallucination

FROM: Chief Investment Office
DATE: October 6, 2025
RE: Market Positioning in Light of Recent Developments
CLASSIFICATION: For Internal Distribution Only

---

Colleagues,

Three days ago, OpenAI closed a secondary share sale that values the company at $500 billion. Half a trillion dollars. To put that in perspective: that's more than Costco, more than Visa, more than every single major automaker combined. This valuation milestone came just as Nvidia hit fresh all-time highs, the S&P 500 crossed 6,700, and chip stocks collectively added roughly $200 billion in market cap on nothing but vibes and the words "artificial intelligence."

I need you to understand what we're witnessing here. We are watching the final act of a valuation cycle that has completely severed from the constraints of traditional financial analysis. OpenAI does not trade publicly. It has no quarterly earnings calls. The $6.6 billion in secondary shares that changed hands represent liquidity for insiders—not capital raised for operations, not proof of revenue growth, not validation of profitability. Just a price tag agreed upon by people who need to believe the number.

And the market ate it whole.

Here's what happened in the 72 hours following the OpenAI announcement: AMD climbed 2.7%. Nvidia added nearly 3%. SK Hynix jumped. Tokyo Electron surged over 4%. Every single semiconductor name with even a tangential connection to AI infrastructure rallied on the assumption that if a chatbot company is worth $500 billion, then surely the companies selling them shovels must be worth proportionally more. The logic is circular. The conviction is absolute. The risk is invisible.

But let's run the numbers for a moment. OpenAI's valuation is approximately 50x what many analysts estimate their 2025 revenue might be. Fifty times. For context, Nvidia—which actually prints cash, actually has a moat, actually dominates its market—trades at around 20x forward sales. The market is pricing OpenAI as if it will not only maintain dominance in a crowded field of competitors but also expand margins, scale infrastructure, and somehow monetize chatbots at rates that would make Big Tech blush.

This is not analysis. This is theology.

Now, I'm not here to moralize about bubbles. Bubbles are just momentum trades with longer time horizons. Our job isn't to call tops—it's to allocate capital intelligently and hedge tail risk. But what concerns me is the degree to which the entire technology complex has become a single, undiversified bet on a narrative that cannot be empirically tested in real time. When chip stocks rally because an AI company you can't invest in gets a new valuation, you're not trading on fundamentals. You're trading on faith.

And faith is not an asset class.

Here's the structural problem: AI infrastructure spending is real, but it's also front-loaded. Companies like Microsoft, Google, and Meta are pouring tens of billions into GPUs, data centers, and power infrastructure. That's great for Nvidia today. But at some point—maybe 2026, maybe later—the marginal return on the next data center declines. The next GPU farm doesn't unlock a new capability; it just processes more of the same prompts. And when that happens, capex contracts. Not catastrophically, but enough to reset expectations.

The market hasn't priced this in because the market doesn't believe it can happen. Right now, every investor in the AI trade believes they're early, that they'll see the slowdown coming, that they'll rotate out before everyone else. Statistically, this is impossible. Psychologically, it's inevitable.

So what do we do?

First, we acknowledge that momentum can persist far longer than rationality would suggest. The S&P 500 is up nearly 17% year-to-date. China large-caps have surged 36%. These are not markets that respect valuation concerns. If you fought the tape in January, you've underperformed for nine months. Don't fight it now.

Second, we diversify away from single-factor exposure. If the portfolio is overweight semiconductors because we believe in AI, fine—but make sure the rest of the book isn't also levered to the same thesis through cloud infrastructure, SaaS multiples, or tech-heavy indices. The last time a single sector dominated this completely was 1999. You remember how that ended.

Third—and this is critical—we start building positions in assets that benefit when narratives crack. Gold has rallied 40% this year, not because of inflation but because of uncertainty. Bitcoin just crossed $116,400 on the same impulse. These aren't rational hedges; they're emotional ones. But in a market driven by narrative rather than cash flow, emotion is the variable that matters most.

Finally, we watch the data. Not the hype cycle, not the secondary share sales, not the analyst notes from banks with AI investment banking mandates. We watch revenue growth at the hyperscalers. We watch capex guidance. We watch GPU shipment volumes and power consumption trends. Those are the early warning signals. When the marginal AI data center stops getting funded, you'll see it there first.

I'll say this plainly: I do not know when this reprices. Neither do you. Neither does anyone else, no matter how confident they sound on television. What I do know is that $500 billion valuations for companies with uncertain business models do not persist indefinitely without delivering results that justify the price. And right now, we're in the phase where the price justifies itself simply by existing.

That works until it doesn't.

Stay sharp. Stay hedged. And for the love of fiduciary duty, don't let anyone on your team pitch OpenAI as a comp for anything.

Best,
CIO