China restarted robotaxi permits and its companies are racing into Dubai, Abu Dhabi, and Singapore. Cost matters: Baidu builds vehicles for $29,000 using EV supply chains while US firms refine premium markets at home. Strategy follows economics.
Good Morning from San Francisco, Microsoft had an interesting week on the transparency front. Australia's competition watchdog sued,
Australia's regulator says Microsoft pushed 2.7 million subscribers toward pricier AI plans by hiding a cheaper option inside the cancellation flow. The case tests whether essential software providers must disclose alternatives as clearly as upgrades.
Microsoft buried OpenAI losses in a $4.7B catch-all line without disclosing stake size, fair value, or revenue mechanics—despite a partner now worth $500B. Investors can't map exposure or returns. The gaps are fixable and overdue.
Investors love the upside story. They can’t see the accounting.
Microsoft’s partnership with OpenAI is now central to its valuation narrative, yet the money trail remains opaque. In its latest annual report, Microsoft swept OpenAI-related losses into a $4.7 billion “other, net” line, without breaking out the number, the stake’s fair value, or the mechanics of revenue sharing—despite classifying OpenAI under the equity method. That leaves shareholders guessing about both exposure and return.
Three gaps dominate. First, Microsoft hasn’t disclosed the carrying amount of its OpenAI stake, even as it confirms “reciprocal revenue-sharing arrangements.” Second, OpenAI doesn’t appear in Microsoft’s related-party footnotes—unusual given the equity-method label implies significant influence. Third, the company offers no detail on how those reciprocal flows hit revenue, cost lines, or cash. These are fixable omissions; they’re also material ones.
Essential Points
• Microsoft reports only $6B total for all equity-method investments despite OpenAI's $500B valuation suggesting stake above $100B
• OpenAI revenue share dropping from 20% today to single digits by decade's end, shifting Microsoft returns to equity value
• OpenAI signed $300B Oracle deal plus arrangements with Nvidia, AMD, Broadcom—reducing Azure dependence Microsoft models assume
• Missing disclosures: ownership percentage, carrying amount, related-party table, revenue recognition policy, and sensitivity analysis
The scale argues for clarity. After a recent secondary sale, OpenAI’s private valuation reached roughly $500 billion. If Microsoft owns a meaningful slice, the implied fair value could be enormous relative to the $6 billion Microsoft reports for all equity-method investments combined. That contrast is precisely why analysts have pressed for related-party transparency.
The partnership’s economics are also moving. Reporting indicates OpenAI is reducing the share of revenue it passes to commercial partners over time—from about 20% today toward single digits by decade’s end. For Microsoft, less revenue share means fewer near-term, contractual inflows—and greater reliance on either Azure consumption tied to OpenAI workloads or on long-dated equity value. Without clear disclosure of how those flows are recognized, investors can’t map the shift to margins or cash.
Equity-method accounting dulls the signal. Microsoft records its share of OpenAI’s profits or losses against the investment’s carrying amount, capping losses once that book value hits zero; new funding lifts the cap. In practice, rising OpenAI operating costs or infrastructure prepayments can widen losses before scale kicks in, while revenue-share terms and Azure consumption rates determine Microsoft’s offset. The problem isn’t the method; it’s the lack of note-level detail to gauge the sensitivity.
OpenAI’s compute strategy adds another layer. The company has diversified beyond Microsoft—signing a reported five-year, ~$300 billion agreement with Oracle and pursuing multi-year, milestone-based arrangements with Nvidia, AMD, and Broadcom. Whatever the ultimate spend, the direction is clear: supplier optionality, not single-cloud dependence. For Microsoft, that could temper the Azure uplift otherwise assumed in bullish models.
A credible fix requires plain English and basic numbers:
None of this requires revealing competitive secrets; it simply aligns a material relationship with standard public-company disclosure.
Two markers arrive almost at once. First, Microsoft’s Oct. 29 earnings call is an opportunity to move OpenAI into the related-party footnote and quantify the stake. Second, as OpenAI’s new financing and capacity agreements progress, investors should expect cleaner breakouts of Azure consumption versus third-party clouds—so the market can separate “AI halo” from booked revenue. Both steps would lower the discount rate investors quietly apply to opaque AI stories.
Microsoft needs OpenAI for product velocity and cloud demand. OpenAI needs Microsoft for distribution and credibility. The financial plumbing between them, however, is too murky for a relationship of this size. Better notes—not bigger promises—will decide whether the market keeps giving Microsoft the benefit of the doubt.
Q: What is equity method accounting and why does it matter here?
A: Equity method means Microsoft records its share of OpenAI's profits or losses each quarter, reducing the investment's book value when OpenAI loses money. Losses stop once the carrying amount hits zero, unless Microsoft adds more funding. The method signals Microsoft has significant influence—typically 20-50% ownership—which triggers related-party disclosure rules Microsoft isn't following.
Q: How much has Microsoft actually invested in OpenAI compared to what the stake might be worth?
A: Microsoft invested roughly $13-13.75 billion since 2019, according to Wall Street Journal reporting. OpenAI's recent $500 billion private valuation suggests Microsoft's stake could be worth over $100 billion at fair value. Yet Microsoft reports only $6 billion total for all equity-method investments combined, creating a massive valuation gap investors can't reconcile without more disclosure.
Q: What are related-party disclosures and why is this a problem?
A: Related-party rules require companies to disclose transactions with entities they control or influence, including dollar amounts, pricing terms, and outstanding balances. This lets investors judge whether deals happen at fair market rates. Microsoft labels OpenAI an equity-method investment (which means influence exists) but excludes it from related-party footnotes—a contradiction that hides transaction details.
Q: What happens to Microsoft if OpenAI's losses keep growing?
A: Under equity method rules, Microsoft records its share of OpenAI's losses each quarter until the investment's carrying amount reaches zero. After that, losses stop hitting Microsoft's income statement unless it provides new funding, which resets the cap. Rising OpenAI spending on infrastructure could widen near-term losses while returns stay tied to long-term equity value or Azure consumption.
Q: Why does OpenAI's $300 billion Oracle deal matter for Microsoft's returns?
A: Microsoft's business case assumes OpenAI workloads run on Azure, generating cloud revenue alongside the equity stake. OpenAI diversifying to Oracle, Nvidia, AMD, and Broadcom means less Azure consumption than bullish models assume. Without disclosure of how much OpenAI actually spends on Azure versus competitors, investors can't separate the "AI halo" from real revenue.
China restarted robotaxi permits and its companies are racing into Dubai, Abu Dhabi, and Singapore. Cost matters: Baidu builds vehicles for $29,000 using EV supply chains while US firms refine premium markets at home. Strategy follows economics.
OpenAI's company knowledge mode connects workplace apps to ChatGPT—but the real test is whether enterprises will expose their entire institutional memory to AI. The feature points toward governed knowledge bases, yet arrives with manual toggles and gaps Microsoft solved months ago.
Anthropic secured Google's largest chip deal—up to 1M TPUs worth tens of billions—while keeping Amazon as primary partner. The rare multi-cloud strategy gives the startup leverage both clouds typically demand for themselves. Can neutrality scale?
OpenAI bought the team behind Sky, a Mac assistant that actually controls your apps. It's a move from answering questions to doing work—and puts pressure on Apple's automation story. Integration timing and privacy details will decide if it sticks.
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
