The Day Tesla Stock Fell 14%, Musk Talked Abundance

Opening

Reader, something strange happened at Tesla’s Gigafactory in Texas on July 23rd.

The evening before, Tesla had reported second-quarter earnings per share of $0.33, badly missing the $0.44 consensus, and the stock plunged 14.5% that day. Yet inside that very factory, Elon Musk sat down for a 90-minute interview with The Economist and said: “We are heading toward an age of stunning abundance.”

Five days later, the mirror-image scene played out. On July 28th, Mark Zuckerberg published an op-ed in The Wall Street Journal titled “The AI Future Is for Everyone.” The next day, Meta reported a 14% drop in net income, and its stock fell nearly 10% in after-hours trading.

One man talked abundance the day after a crash; the other unveiled a philosophy the day before an earnings report. Here’s the bottom line up front: the AI safety blueprints tech titans rolled out this July are business plans before they’re safety documents. Today, let’s lay all three blueprints on top of each other and read them together.

---

Three Blueprints in Two Weeks

The first blueprint appeared on July 14th. Google DeepMind’s Demis Hassabis posted a piece on X titled “A Framework for Frontier AI.” Its core idea: build an AI standards body modeled on FINRA, the self-regulatory organization¹ (SRO) that oversees the U.S. securities industry. Industry would fund it, independent technical experts would review new models 30 days before launch, and the scheme would start voluntary before hardening into a requirement for U.S. market deployment. Sam Altman and Satya Nadella voiced support immediately, reports emerged that the Treasury Secretary had been involved in shaping the design, and the White House was said to be reviewing it.

The second blueprint came in Musk’s July 23rd interview. He revealed he’d spent hours on the phone with Hassabis before the post went up, but cut straight to the point: no need for a formal body. Leading AI companies should simply talk biweekly and show rivals their new models 1-2 weeks before launch. His logic: governments lack the technical understanding to judge frontier models, but competitors have it, and competitors have every incentive to delay a rival’s launch, so they won’t quietly look away if they spot a problem. It’s a design where competition itself becomes surveillance. Government stays in reserve as a last resort, stepping in only when the industry signals it’s needed.

Underneath this proposal sits a premise that’s pure Musk. In this same interview, he said AI will surpass the sum of human intelligence within 5 years, and that the odds humans will still hold the reins in 10 years are low. The intelligence gap between AI and humans, he argued, will exceed the gap between humans and chimpanzees — and it’s hard to picture chimpanzees controlling humans. He didn’t walk back his old claim that there’s a 10-20% chance killer robots end humanity, either. Asked “would you get on a rocket that had that same chance of exploding,” his answer was “yes.” If you can’t stop it, the thing to do is lower the odds and enjoy the ride. All delivered with a note of self-deprecation: he founded OpenAI to keep Google in check, then sued the very OpenAI he no longer trusted — and somehow every road still leads back to acceleration.

The third blueprint is Zuckerberg’s July 28th op-ed. His question isn’t whether superintelligence arrives, but who gets access to it. He writes that the truly dangerous idea is concentrating power in a few hands in the name of safety — the answer, instead, is giving everyone their own personal superintelligence. He contrasts a courtroom where only one side has a superintelligent lawyer with one where everyone does, arguing that distributed power naturally lets people check each other, and check big institutions too.

An institution. Mutual surveillance. Distribution. The three designs contradict one another, yet agree perfectly on one thing: the pen that writes the rules should be held by them, not by government.

If this dynamic feels familiar, there’s a reason. In 2017, Zuckerberg called Musk’s AI warnings “pretty irresponsible,” and Musk fired back that Zuckerberg’s “understanding is limited.” 9 years ago, the two fought over whether AI was dangerous. In 2026, both have shelved that question and are fighting instead over who’s qualified to write the rules. The entire axis of the argument has shifted.

---

Every Blueprint Resembles Its Author

Now let’s read each blueprint again, laid over its author’s own profit-and-loss statement.

Start with Hassabis’s institutional model. Pre-launch review, standards, compliance — all of it costs money. And the players best positioned to absorb that cost are exactly the leading firms with the lawyers, policy teams, and capital to spare. The burden of a delayed launch during review also falls harder on the chaser than the leader. An institution sounds neutral, but the moment a threshold appears, it favors whoever has the stamina to clear it.

Now Musk’s mutual surveillance. Having folded xAI into SpaceX, Musk is the challenger in this race — by his own admission, “Anthropic is currently in the lead.” From that position, “getting to see a rival’s model 1-2 weeks early” is both a safeguard and, for a challenger, the sweetest possible information advantage. The power to flag a problem and delay a rival’s launch is a bonus on top. His incentive design might well be the right one. But it’s worth remembering who ends up wearing the referee’s armband in this scheme.

Zuckerberg’s distribution argument sounds the most philosophical of the three, but laid over Meta’s business structure, it’s the most finely engineered. Meta doesn’t make money selling models; it makes money selling ads on a distribution network that 3.6 billion people use every day. The more the model layer becomes a commodity, the shallower the moat gets for rivals selling closed models — and the higher Meta’s position, as the owner of the distribution network, rises. The claim that “concentration is dangerous” is also, conveniently, a shot aimed squarely at the closed frontier labs.

Reading the op-ed again with a fact-checker’s eye reveals an interesting gap: nowhere in the full text does the phrase “open source” or “open weight” appear. The promise stops at “giving everyone personal superintelligence.” That’s not a pledge to release model weights — it’s a pledge to distribute access through Meta’s own products. Sure enough, on the earnings call the next day, Zuckerberg said Meta would “resume” releasing open-source models “at some point,” but that it would mix open and closed approaches, adding there’s “no dogma” here. An anti-concentration manifesto coexists, without contradiction, with concentration into Meta’s own channel.

Is any of this strange? Economist George Stigler argued in a 1971 paper that regulation isn’t a product of the public interest — it’s a good that industry acquires and designs to serve its own interest. The living specimen of this insight, known as regulatory capture², is none other than FINRA itself, the very model Hassabis proposed. Its predecessor, the NASD, was an institution the securities industry designed for itself under the Maloney Act of 1938. It’s not that each proposal happens to resemble its proposer. This is simply how institutions are born.

---

Philosophy on the Earnings Calendar

Let’s return to the two opening scenes. Why those particular dates?

Tesla’s second quarter saw solid deliveries of 480,000 vehicles, but profit badly missed expectations. SpaceX, which had gone public on June 12th in the largest IPO ever, had also failed to hold its first-month high. In the very week both his companies faced market skepticism, Musk spoke of superintelligence in 5 years and abundance in 10. Switching from the language of the quarter to the language of the century when the quarter’s numbers turn against you — that’s an old communications trick.

Meta’s structure is more clear-cut. Second-quarter revenue rose 28% to $60.8 billion, beating expectations, but costs surged 55%, driving operating income down 8%. Legal costs of $2.4 billion and restructuring costs of $1.2 billion piled on simultaneously, and net income fell 14% to $15.8 billion. This year’s capex guidance was raised at the low end, to a range of $130-145 billion. Quarterly free cash flow came in at $784 million — effectively zero for a company this size, meaning the cash it generates is being swallowed whole by data centers.

Then, on the next day’s earnings call, Zuckerberg summoned the op-ed himself. After noting he’d just published a piece on an optimistic future, he said Meta is the only major company that has made “putting superintelligence directly in people’s hands,” rather than centralizing it, its top priority. His defense for the spending was demand: he said offers to buy Meta’s existing compute at a premium keep coming in. CFO Susan Li backed this up, saying the industry has historically under-built relative to AI demand. The philosophy that ran in the newspaper the day before came back the next day as the justification for $130 billion in spending. The contrast sharpens when set beside Microsoft, which reported 43% growth in Azure that same day and rose 3% after hours. The company already seeing AI spending convert into revenue spoke in numbers. The company that hasn’t yet spoke in philosophy.

Don’t misread this. I’m not saying the philosophy is fake. Zuckerberg’s distribution thesis and Musk’s abundance thesis are both beliefs the two men have repeated, near-sincerely, for years. But when that belief gets deployed on an op-ed page or an earnings call is a separate variable — and this July’s timing was synchronized with the earnings calendar to a remarkable degree.

---

Oswald’s Lens

In building GTM strategy, I’ve worked on industry position papers aimed at regulators more than a few times, and it left me with a reading habit that stuck. A regulatory proposal that an industry volunteers is always doing two things at once: shrinking the government’s rationale for intervening, and staking a claim on the coming rules before anyone else can.

Read this way, all three blueprints are textbook cases. So when I read documents like these, I ask two questions before I ask whether the content is right or wrong. First: whose hand ends up holding the pen in this design? Second: what weakness of the proposer does this design conveniently paper over? Hassabis’s institution converts the leader’s speed burden into a shared burden under the name of “review.” Musk’s mutual surveillance fills in the challenger’s information deficit. Zuckerberg’s distribution swaps the burden of the frontier race for the distribution-network race he’s already won.

There’s a clear reason this isn’t someone else’s problem for Korean readers: we already chose the opposite path. Under Korea’s AI Framework Act, which took effect on January 22nd, the government unambiguously holds the pen. But depending on whether the U.S. lands on Hassabis’s institution, Musk’s gentleman’s agreement, or Zuckerberg’s market laissez-faire, the global compliance terrain Korean companies will face looks completely different. This isn’t just a PR battle among other companies — it’s the trailer for our own regulatory environment, and it should be read that way.

---

Closing

To sum up, three things.

First, over two weeks in July, three AI safety blueprints emerged — an institution (Hassabis), mutual surveillance (Musk), and distribution (Zuckerberg) — and all three leave industry holding the pen.

Second, each blueprint precisely mirrors its proposer’s market position. As regulatory capture theory tells us, this is simply how institutions are born.

Third, the timing of each announcement was synchronized with the earnings calendar. Narrative rings loudest exactly when the numbers turn unfavorable.

The next thing to watch is the White House. Which blueprint gets stamped with approval will be act two of this story.

Reader, which of the three — institution, mutual surveillance, or distribution — do you think is most likely to actually get adopted? Pick one, tell me why in the comments, and I’ll gather the responses for the next issue.

---

💬 Share your prediction with Reader in the comments — I’ll fold the results into the next issue. 📨 If you have a colleague wrestling with AI regulatory strategy, pass this along.

---

📎 References & Further Reading

Primary sources

• The Economist, “An interview with Elon Musk”, July 2026. ··· This is today’s primary source — the 90-minute interview containing the mutual-surveillance proposal, the chimpanzee analogy, and the rocket exchange. Paywalled content.
• Mark Zuckerberg, “The AI Future Is for Everyone”, The Wall Street Journal, July 28, 2026. ··· The original text of the distribution thesis. Worth checking yourself whether the phrase “open source” really is absent.
• Demis Hassabis, “A Framework for Frontier AI and the Dawning of a New Age”, X, July 14, 2026. ··· The original text of the FINRA-model proposal. You can also see the supportive replies from Altman and Nadella.
• Meta Investor Relations, Q2 2026 earnings release and conference call, July 29, 2026. ··· See in the original how the op-ed’s philosophy got recalled on the earnings call.
• Tesla Investor Relations, Q2 2026 Update, July 22, 2026. ··· The earnings released the night before the interview.

Background

• George J. Stigler, “The Theory of Economic Regulation”, The Bell Journal of Economics and Management Science, 1971. ··· The original text on regulatory capture theory. Even just the introduction shows where today’s lens comes from.
• Fortune, “Wall Street is helping shape Google DeepMind CEO’s pitch for AI industry oversight”, July 21, 2026. ··· Reporting on the White House review of Hassabis’s proposal and the Treasury Secretary’s involvement.

Past issues worth reading alongside this one

• Why Did Jensen Huang Praise a Chinese Model? ··· Uses the same method of translating a CEO’s public remarks into self-interest. A companion piece to today’s issue.
• Why Does the New Fed Chair Talk Like a Startup Founder? ··· Covers the structure of narrative being used as currency in markets.

---

📝 Glossary

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at OBF (Oswarld Boutique Consulting Firm). He teaches statistics and data analysis, including business data management and business analytics, while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master's from Korea University's Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

1. Self-Regulatory Organization (SRO): An industry-run regulatory body that operates in place of a government regulator. FINRA, which oversees the U.S. securities industry, is the classic example — funded by member firms while also supervising them. This is exactly the model Hassabis proposed adapting for AI. ↩

2. Regulatory capture: The phenomenon where regulation ends up designed and operated to serve the interests of the industry it regulates, rather than the public interest. Theorized by George Stigler in 1971, work for which he won the 1982 Nobel Memorial Prize in Economic Sciences. ↩