Memo to Investors: On the Unserious Business of Taking AI Seriously

FROM: Your Portfolio Manager's Conscience
TO: Anyone Still Pretending Valuations Matter
RE: The $500 Billion Question
DATE: October 4, 2025

We need to talk about OpenAI.

On Wednesday, the company closed a $6.6 billion secondary share sale that values it at half a trillion dollars. Not revenue. Not market cap after years of profitability. Valuation. For a private company that's never turned a profit and exists primarily to burn through compute credits and Azure capacity.

Five. Hundred. Billion. Dollars.

To put this in perspective: OpenAI is now worth more than SpaceX, which actually launches things into orbit and brings them back. It's worth more than ByteDance, which has 1.5 billion monthly active users on TikTok. It's worth roughly the same as Visa, a company that processes $14 trillion in payments annually and has a monopoly on turning plastic rectangles into money.

What does OpenAI do? It makes a chatbot that occasionally hallucinates legal citations and writes mediocre marketing copy.

The buyers in this deal—SoftBank, Thrive Capital, MGX, T. Rowe Price—just signed checks that imply ChatGPT is going to capture economic value equivalent to the entire payment processing infrastructure of the global economy. Show your work, gentlemen.

And you know what happened when this news broke? The Nasdaq 100 climbed to a fresh record. The Philadelphia Semiconductor Index rallied. Tech stocks celebrated like someone just proved the business model works. Nobody stopped to ask whether a company with negative cash flow and a cost structure that requires a $300 billion Oracle cloud deal to function might be priced for a future that doesn't exist yet.

Meanwhile, across the market, reality tried to intrude.

Palantir—darling of the defense tech revolution, AI avatar for the MAGA-industrial complex—got clipped for 7.5% after Reuters surfaced an internal Army memo describing the battlefield communications system Palantir built with Anduril as riddled with "fundamental security problems." The memo, written in September by the Army's chief technology officer, flagged issues including uncontrolled data visibility, no ability to track user activity, and modules harboring over 200 vulnerabilities. The Army called it "very high risk."

Palantir refuted the claims. Said the memo was outdated. Anduril echoed the same defense. The stock recovered slightly. But here's the thing: these are the companies promising to disrupt legacy defense contractors by moving fast and breaking things. Except when you're building systems for soldiers in combat zones, "breaking things" isn't disruption. It's catastrophic failure with body bags.

The S&P 500 closed at 6,715—a fresh all-time high. The Dow climbed 0.17%. The Nasdaq Composite rose 0.39%. None of this should be happening during a government shutdown that's left the Bureau of Labor Statistics dark, the Fed flying blind, and ADP's private payrolls report showing a 32,000-job contraction in September.

But it is happening. Because markets in October 2025 have decided that none of the old rules apply anymore. Economic data? Optional. Profitability? Quaint. Valuation discipline? For boomers.

What we have instead is a risk-on frenzy powered by the belief that AI will save us all, that trillion-dollar valuations are justified by vibes and potential, and that somehow the same Federal Reserve that spent two years hiking rates to kill inflation will now pivot dovishly just because we really, really want them to.

OpenAI at $500 billion isn't just a number. It's a symptom. It's what happens when money has nowhere else to go, when yields are compressed, when pension funds and sovereign wealth funds are so desperate for growth that they'll pay anything for exposure to the Next Big Thing. It's the sound of capital chasing narrative over numbers.

And the narrative right now is that AI will change everything, so why bother doing the math?

Here's some math anyway: If OpenAI is worth $500 billion, it needs to eventually generate profits that justify that valuation. Let's be generous and say it trades at a 30x P/E—reasonable for a high-growth tech company. That means OpenAI needs to earn $16.7 billion in annual profit. Not revenue. Profit. After paying for compute, salaries, regulatory compliance, and whatever legal fees it takes to fend off copyright lawsuits from every media company on earth.

Does anyone seriously believe that's happening in the next five years? Ten?

Or is this just 1999 with better graphics?

I'm not saying AI isn't transformative. It is. I'm not saying OpenAI won't be a significant company. It will be. What I'm saying is that $500 billion is a bet on perfection—on flawless execution, on zero competition, on regulatory tailwinds, on Moore's Law holding forever, on enterprise adoption accelerating at exponential rates, and on OpenAI capturing the lion's share of an industry that doesn't have established unit economics yet.

It's a valuation that assumes everything goes right and nothing goes wrong.

And if 2025 has taught us anything, it's that things go wrong all the time. The government shuts down. Jobs reports get canceled. Battlefield communication systems ship with 200 vulnerabilities. Markets climb to all-time highs while the data that's supposed to guide them sits locked in a furloughed office.

So when you see OpenAI at $500 billion and the S&P 500 at record highs and Bitcoin pushing $120K and everyone celebrating the AI revolution, just remember: the last time capital got this euphoric about a technological paradigm shift, we ended up with Pets.com and a three-year bear market.

This time might be different.

But it probably isn't.

END MEMO