OpenAI had another terrible week

A lawsuit from Apple, an Oracle downgrade, and a price war.

57%

Of Americans placed a bet last year. Only 38% read a book.

1. OpenAI had another terrible week

2. Market broadening is a myth: 50% of the S&P 500 is AI

3. Netflix earnings disappoint as engagement slips

---

OpenAI Could Miss Its 2030 Revenue Forecast by 70%. Here’s Why

It was another terrible, horrible, no good, very bad week for OpenAI. The company was caught selling advanced AI models to Chinese firms blacklisted by the Pentagon, its first AI device leaked (it’s reportedly a movable speaker), and according to new Emarketer estimates, OpenAI’s ad business is on track to miss its own forecast by 95%.

That’s not all. Apple sued OpenAI last week, alleging its consumer hardware plans are the product of stolen intellectual property. S&P Global Ratings also downgraded Oracle’s debt to BBB-, just one notch above junk status, citing OpenAI as a “key credit risk.” Plus, DeepSeek is reportedly preparing for an IPO and could file as early as this year. A successful public debut of a cheaper Chinese AI model provider could make it even harder for OpenAI and Anthropic to attract capital.

Taken together, these issues all call into question OpenAI’s ability to hit its revenue projections and fulfill hundreds of billions of dollars in contractual obligations to compute providers and chip companies.

First, Apple’s lawsuit could put OpenAI’s entire hardware business on hold. Apple alleges that OpenAI poached over 400 Apple employees, extracted confidential information from them, and then tricked Apple’s suppliers into doing proprietary work for OpenAI that they didn’t have permission to do. Apple is asking the court for monetary damages and an order for OpenAI to return or destroy all misappropriated property.

Second, the AI price wars have begun, and Chinese firms like DeepSeek are the biggest threats. Open-source Chinese models now account for nearly 50% of enterprise token usage on OpenRouter, a marketplace for AI models. That’s up from just 4.5% in the first half of 2025.

In response, American firms are cutting their prices dramatically. Last week, Meta announced a new model, Muse Spark 1.1, that is up to 75% cheaper than OpenAI and Anthropic. Under industry pressure, OpenAI released a model undercutting itself by 80%.

In a worst-case scenario, if the Apple lawsuit shuts down OpenAI’s hardware business, ChatGPT advertising revenue is as low as EMarketer’s projections, and price wars force OpenAI to reduce their model pricing by 80%, OpenAI’s 2026 revenue would fall by 40% and its 2030 revenue would fall by 70%.

That scenario would be disastrous for a company that is, in an ideal case, covering only about 80% of its cash burn by 2030.

This would also impact when OpenAI becomes cash flow positive. According to internal projections OpenAI will become cash flow positive by 2030. In this downside case, it would instead lose $165 billion that year.

OpenAI CEO Sam Altman tried to calm investor concerns with a tweet, but his statement ultimately amounts to a promise to “do the right thing.” Whatever that means.

The best business model in history is IP theft. Second best: giving 80% of a product’s value at half the price. That’s what DeepSeek and the rest of the Chinese open-weight models are trying to do right now.

America has made a massive bet on AI, and China just showed up with a near-frontier product for a fraction of the cost. This becomes the next geopolitical football once Trump figures out what’s happening.

The Market Isn’t Broadening — It’s Just Getting Better at Hiding AI

Investors keep hearing that the stock market is broadening. But is it? The deeper you look, the harder it is to argue that stocks, bonds, and even alternative assets are now one big bet on AI.

This pattern is most evident across the equity markets. AI-related stocks account for more than 50% of the S&P 500 by weight, and if you removed AI and energy from the S&P this year, the S&P would be in the red.

AI is the hidden catalyst driving returns across seemingly unrelated sectors. For example, 3 of the 4 top performers in the S&P 500 real estate sector are real estate investment trusts (REITs) that focus on developing AI data centers.

Utilities companies are benefiting from AI’s skyrocketing demand for power. U.S. electricity demand jumped to a record high last year, with data centers accounting for around 50% of the demand growth.

Industrial stocks are soaring on construction demand from building AI data centers. In fact, for the first time since 2021, industrials in the S&P 500 are trading at a higher forward price-to-earnings multiple (26x) than technology companies (24x).

The financial sector hinges on AI, too. Big banks are collecting record fees from AI company IPOs and M&A activity and record trading revenues from market hype around AI. The Financial Times’s Robert Armstrong went so far as to write: “It’s not too much of a generalisation to say: the big banks are a straight AI play now.”

Even 52% of the Russell 2000 small-cap index’s return in the first half of the year came from AI-related companies.

Emerging markets are no different. South Korea and Taiwan have accounted for 75% of emerging-market returns, and most of those gains came from three suppliers of AI semiconductor chips: TSMC, Samsung, and SK Hynix.

And in Europe, just nine AI-winners have accounted for about 47% of returns to the Stoxx Europe 600 this year.

Torsten Slok, Apollo’s chief economist, articulated the implications of this dependence succinctly: “This AI thing better work out.”

A Real Estate Investment Trust (REIT) is a company that owns, operates, or finances real estate — apartment buildings, hotels, or, increasingly, data centers. Many REITs trade publicly like stocks, so buying a share means buying into a professionally managed real estate portfolio. REITs are required to distribute at least 90% of their annual taxable income to shareholders as dividends.

CNBC pundits own stocks, so they’re always going to make the case for why everyone else should buy more. But don’t be fooled: The market isn’t broadening; it’s just finding new ways to buy Nvidia.

Everything’s becoming an AI stock. That’s not necessarily bearish, but investors are kidding themselves when they call that “broadening,” like it means diversifying away from AI. It doesn’t. Buying “AI-adjacent stocks” and calling it broadening is like ordering a Diet Coke with your Double-Double from In-N-Out. Be clear: You still bought a cheeseburger.

Of the Big Tech names, who’s least dependent on AI? Apple. Apple’s stock is up 60% over the past year and just surpassed Nvidia to again become the most valuable company in the world. Amazon, still AI related but more diversified than the other hyperscalers, is up 11% over the past year. Microsoft, ground zero for AI, is down 23%.

If I could go long one basket of stocks, it’d be GLP-1s. If I could go short one, it’s AI. But to be clear: I’m not telling you to hold gold bars or cash. I’m always in the market — you never know how fast or how irrational it’s going to run. But you should understand how exposed the market really is to one sector.

I’m a big fan of index funds and passive investing: Put your money in and let the market do the work. But now we have to ask what true diversification actually means. Putting your money in the S&P doesn’t do that job anymore, which means you have to start doing some homework.

The question is: Can you find sectors with real distance from AI?

One sector I’d point to: healthcare. It was one of my picks at the start of the year, and I’m sticking with it. AI hasn’t touched it yet — which means the real returns might still be ahead. But finding these sectors is the hard question investors are facing right now.

Netflix’s Engagement Slips as Competitive Pressure Mounts

Netflix reported disappointing second quarter earnings. Revenue grew 13%, short of expectations, and the streaming giant spooked investors by reporting weak engagement data — and then announcing it would start publishing engagement metrics less frequently. The stock fell as much as 8% Friday.

Netflix once boasted of its transparency; now, that claim seems almost ironic. In Q1 2025, Netflix stopped reporting quarterly subscriber counts, telling investors to instead focus on engagement. Last week, the company decided to cut its What We Watched engagement report from twice a year to once a year, starting in 2027.

The final biannual engagement report looked weak. Total viewing hours rose just 2%, while the subscriber base grew an estimated 10%, implying an 8% decline in per-subscriber daily engagement.

Netflix has been facing increasing competition from short-form video providers, specifically YouTube. In response, it added “Clips,” a TikTok-style scroll feature surfacing short content from its own library, video podcast deals with Spotify and Barstool, and new licensing deals with outside publishers (BuzzFeed, Condé Nast) that bring new short-form content onto the platform.

Netflix has lost more than $250 billion in market value over the past year, and fellow streaming giant Disney has lost nearly $50 billion. Both are well-managed companies that are growing revenues and subscribers, raising prices — and being punished for it. This raises an important question: Is streaming just a bad business? Or have Netflix’s and Disney’s creative juices run dry? Let us know your take in the comments.

In the next six months, OpenAI is going to acquire Sierra, an enterprise AI company, and install Bret Taylor as CEO. Sam Altman will be moved up to chairman. Altman is an innovator, not an operator, and Bret Taylor may be the best enterprise software operator of his generation.

1. Teens give their verdict on Britain’s social media curfew

2. The Conspiracy World Cup

3. Investors now value Eli Lilly more like a tech giant than a drugmaker