OpenAI's Annualized Revenue Grew Nearly 80%, Yet the Market Is Disappointed; AI Reshapes McKinsey Hiring with a Greater Focus on Art, History, and Music Talent

This edition of The New York Times DealBook is a comprehensive financial newsletter: it discusses AI companies and capital markets while also examining corporate M&A, household expenses, satellite communications, and how AI is transforming work and hiring.

Here are the six key takeaways worth noting from this edition:

• OpenAI's annualized revenue grew nearly 80%, so why is the market still disappointed? Newly disclosed figures came in lower than previously circulated levels, sending shares of related companies like Nvidia, Oracle, and Microsoft downward.
• AI has changed McKinsey's hiring criteria. The firm has expanded its consideration of candidates with backgrounds in art, history, and music, placing greater value on novel creativity while also prioritizing the ability to learn how to tackle unfamiliar problems and collaborate with others.
• Why does Starbucks want to buy Chipotle? Acquisition rumors pushed the target company's stock higher while leaving the buyer's shareholders concerned about the deal's price tag.
• When borrowing becomes more expensive, how do living costs change? A federal deficit, a 7.4% mortgage rate, and rising oil prices impact the government, homebuyers, and businesses respectively.
• Why is SpaceX spending billions on spectrum? Direct-to-cell satellite technology is bringing new opportunities and competition to the telecom sector.
• What value are massive bonuses and podcast contracts buying? Goldman Sachs' equity incentives and Spotify's renewal of its flagship show illustrate two distinct ways of allocating capital.

Growth of Nearly 80%: Why Did It Still Disappoint the Market?

First, let us clarify the metric of annualized revenue (run-rate).

Annualized revenue converts a recent revenue pace into an annualized scale. A common example is multiplying the most recent month's revenue by 12; it answers the question, "If business continues at this pace for a year, roughly how large would it be?", whereas actual annual revenue answers "How much has already been realized over the course of the year?"

There is also a distinction between operating revenue and cash collected: service revenue is typically recognized as performance obligations are fulfilled, while customer payments may arrive earlier or later. Neither the annualized metric nor actual annual revenue represents profit after deducting costs.

Data disclosed by an OpenAI spokesperson to DealBook shows:

• At the end of last quarter, OpenAI's annualized revenue was approximately $28 billion;
• At the end of this quarter, this metric climbed to approximately $50 billion.

Expressing the figures in units of $100 million, the change between the two quarter-end levels is: (500 - 280) / 280 ≈ +78.6%, which is close to 79%.

Following this revision of the numbers, shares of Nvidia, Oracle, and Microsoft fell.

Since the annualized revenue level grew by nearly 80% compared with the end of last quarter, why did related stocks still decline?

Two things must be viewed separately: the growth compared with last quarter, and the shortfall relative to previously reported figures.

An OpenAI spokesperson told DealBook that the $70 billion annualized revenue figure previously given to investors was inaccurate. Using the previously reported $70 billion as a baseline against the current quarter-end figure of $50 billion, again in units of $100 million: (500 - 700) / 700 ≈ -28.6%. The $50 billion is approximately 29% lower than the previously reported $70 billion; this does not mean that revenue over the past quarter dropped by 29%.

The newsletter notes that the rapidly rising valuations of AI companies rely on ever-increasing revenue forecasts. Stock prices also reflect investors' expectations for the future. If a larger revenue scale had previously been anticipated, the new figures could prompt them to re-estimate the company's future revenue and returns. Consequently, business growth and declining share prices can occur simultaneously.

Investors are also seeking independent ways to estimate revenue. The newsletter notes that data from platforms such as Sensor Tower and Ramp has been used to build financial models. However, partial consumer spending data is not the same as a company's complete books: payment or usage activity observed through a single channel can only directly reflect the sample it covers. Extrapolating a sample to estimate a company's total revenue still requires assessing omitted channels, sample representativeness, and revenue recognition standards.

Why Does Revenue News from Model Developers Affect Cloud and Chip Companies?

There are operational links between model services, cloud computing, and semiconductor chips:

1. Frontier model developers (e.g., OpenAI, Anthropic): Advance foundational large model research and commercial applications, offering APIs and subscription services to enterprises and consumers;
2. Compute and cloud service providers (e.g., Microsoft, Oracle): Provide large-scale cloud computing resources and data center infrastructure support to model developers;
3. Underlying chip and hardware suppliers (e.g., Nvidia): Supply specialized accelerator chips and server hardware to cloud service providers.

The internal transmission mechanism of this industry chain operates as top-down demand pull and commercial payments: End customers pay for model products and services → Model companies incur compute expenses → Cloud service providers procure hardware from chip and equipment makers based on compute demand.

If investors believe that a model company's future revenue will be lower than previously assumed, they may re-estimate how much compute it can purchase and how long cloud providers will take to recoup their infrastructure investments. When expectations for future expansion shift, chip procurement forecasts can change along with them.

Security Controversies Facing AI Labs

Former Anthropic researcher Jacob Coxon claimed on the PBD Podcast that he is "90% confident" that China has placed personnel inside Anthropic and OpenAI.

This controversy unfolds against the backdrop of US-China AI competition. The newsletter also notes that Donald Trump opposes stricter regulation of American AI companies, advocating instead for accelerated innovation to maintain the lead over Chinese competitors.

A person familiar with Anthropic responded that the company believes there is no evidence to support this allegation, pointing to its personnel vetting, access restrictions on critical technology, and suspicious activity monitoring. According to the newsletter, OpenAI adjusted its security protocols following allegations of model copying by Chinese competitors such as DeepSeek. A person familiar with OpenAI described measures taken by the company, including tighter restrictions on model access, fingerprint-based entry for sensitive areas, and employee vetting.

The newsletter also noted that three former OpenAI safety employees stated they were fired for prioritizing safety over the company's short-term interests.

Several other business and fundraising developments involve compute infrastructure and product competition: Nvidia-backed Australian data center operator Firmus withdrew its public listing plans due to market volatility and may reportedly turn to private investors to raise $3 billion; the Financial Times reported that SoftBank is seeking to raise up to $100 billion from Middle Eastern investors for AI investments; and Meta reportedly accelerated the release of Muse in response to competitive pressure from startup Instinct.

Starbucks Acquiring Chipotle: Is Bigger Necessarily Better?

Another news item provides an example of evaluating corporate growth: what value does acquiring another company actually add?

According to a Financial Times report citing people familiar with the matter, Starbucks has worked with advisers in recent months to explore the possibility of acquiring Mexican fast-casual chain Chipotle (with a market capitalization of around $41 billion); Semafor previously reported that Chipotle had retained investment banking advisers to defend against potential acquisitions or activist investors, noting Wall Street speculation regarding a potential bid by Starbucks. If the transaction proceeds, it would be the largest M&A deal in restaurant industry history, though significant uncertainty remains surrounding the acquisition's prospects.

Following the disclosure of the acquisition rumors, Chipotle's shares rose 6%, while Starbucks' shares dropped by as much as 6.7% intraday before recovering somewhat. Notably, Starbucks' current CEO, Brian Niccol, previously served as Chipotle's CEO for an extended period.

Some Wall Street analysts expressed skepticism about the potential transaction. Sharon Zackfia, an analyst at investment bank William Blair, stated that she sees no obvious revenue synergies from combining the two companies and pointed out that an attempt to acquire Chipotle could heighten investor concerns that Starbucks' own growth potential is limited and that the turnaround plan led by Niccol is losing momentum.

This rumor surfaced against the backdrop of an active broader M&A market. The newsletter lists announced deal volume in the first half of the year at $2.77 trillion, an increase of 48% year-over-year, with nearly half of that total coming from transactions valued at $10 billion or more. While some companies seek new growth through acquisitions, higher borrowing rates mean the financing cost to purchase the same company increases. According to the newsletter, rising borrowing costs in the third quarter have begun to slow M&A activity.

Why did the same news drive the stock prices of the two companies in opposite directions? In takeover negotiations, target company shareholders may expect the buyer to offer a price above the original stock price—an acquisition premium; buyer shareholders, on the other hand, must weigh the purchase price, financing, and integration costs. The two sides focus on different considerations when assessing whether a deal is worthwhile.

Whether an M&A deal creates value depends on what can be saved and what additional sales can be generated after merging:

• Cost Synergies are primarily realized through sharing and integrating back-office resources and supply chains, such as unifying IT systems, consolidating administrative support functions, or pooling property leasing and procurement, thereby saving back-office administrative costs;
• Revenue Synergies depend on whether the merger directly drives sales growth—for example, whether customers visiting to buy coffee will also purchase burritos as a result of the combination, or whether the two brands' loyalty programs and store networks can genuinely drive incremental foot traffic.

Only if the cost savings and incremental revenues are sufficient to cover the purchase price, financing interest, and integration costs can the transaction create value.

Starbucks stated that its team remains focused on executing the Back to Starbucks strategy.

AI Reshapes McKinsey Hiring: Greater Focus on Art, History, and Music Talent

AI has transformed McKinsey's talent screening criteria. Firm leader Bob Sternfels stated in an interview that the firm places greater value on "novel creativity" and has begun broadening its consideration of candidates with academic backgrounds in fields such as art, history, and music.

For job seekers, this means: in addition to demonstrating the knowledge they have already mastered, they must also show how they track down clues, learn new things, and collaborate with others when encountering problems with no ready-made answers.

Why Are Art, History, and Music Backgrounds Receiving Greater Attention?

Sternfels offered a specific basis: McKinsey administers a novel problem-solving test, and some candidates with these academic backgrounds performed better. McKinsey has broadened the scope of majors it considers and uses such tests to observe how candidates tackle novel problems.

"The ability to learn new things" is another distinct screening priority. He said McKinsey places more value on the ability to learn new things than on whether candidates have merely mastered the subjects they studied. The firm designed simulations where established patterns cannot be applied, observing how candidates pick up clues along the way and learn while doing; in this type of test, performance showed no correlation with academic major.

The two tests evaluate creative problem-solving and learning in unfamiliar situations respectively. When preparing for interviews, a clearly articulated experience can explain: what was initially unknown, what clues were uncovered, how judgments were adjusted, and how the problem was ultimately resolved.

Why Is Retail Work Experience Also Worth Noting?

Sternfels noted that collaborating with others has become more important. He pointed to two observational signals: whether an applicant has achieved success in team sports, and whether they worked retail jobs during their teenage years.

Retail experience has a very specific characteristic: young employees have no positional authority, yet they still need to interact with people. Relying on a title to demand cooperation is entirely different from securing cooperation through communication. In a résumé or an interview, clearly explaining how one understood others' needs, navigated disagreements, and achieved cooperation can make the value of this kind of experience evident.

Letting AI Pressure-Test Your Thinking

Sternfels sets aside one hour each week to learn about AI, seeking guidance from technical experts and newly hired business analysts. In his personal workflow, he built a dedicated Agent named Contra.

When preparing speeches, he feeds the draft into Contra and asks two questions:

• What am I missing?
• What is the core counterargument to this content?

This method also serves as a practical, transferable reference for professional readers in their daily work: when preparing important presentations, drafting business proposals, or writing project initiation documents, beyond having models assist in drafting or polishing text, one can also use them as an auxiliary tool to test reasoning, proactively checking for logical blind spots before finalizing documents.

[Prompt Extension Based on the Interview Approach]
Please review the draft proposal below:
1. List the facts and assumptions that the core conclusions rely upon.
2. Point out the weakest arguments; state directly if there are no obvious flaws.
3. Present the strongest counterarguments and specify what additional evidence I need to provide.

Draft proposal:
[Paste your content here]

Managers Must Also Understand the Tools in New Hires' Hands

Equipping junior consultants with extensive AI tools has not only increased their working speed, but also posed new organizational challenges for consulting firm management. Sternfels pointed out that an important management challenge currently faced is how to help partners and senior leaders understand that team members now command unprecedented digital tools.

Deficits, Mortgages, and Oil Prices: How the Cost of Money Affects Daily Life

The US federal budget deficit for the most recent fiscal year was approximately $2 trillion. The deficit is the amount by which spending exceeded revenue during that year. The newsletter emphasizes that the concern also lies in the fact that this fiscal gap is widening; CBO figures cited by the CRFB also reflect this magnitude.

How to raise tax revenue cannot be evaluated solely through individual tax rates. The Budget Lab at Yale's Taxes at the Top interactive tool allows readers to combine multiple policy measures and compare the revenue and distributional impacts of complete packages. For instance, adjustments to capital gains taxes and corporate income taxes may interact, and taxpayers will alter their investment and reporting behavior; thus, combined revenue is not necessarily equal to the sum of separate estimates. The purpose of the tool is to make these interactions comparable.

When Mortgage Rates Are High, Why Are Buyers Asking for Price Cuts Instead?

The newsletter reports the average US 30-year fixed mortgage rate at 7.4%; existing home sales fell 1.2% year-over-year in the most recent month. Zillow economists forecast that the year-over-year decline in the fourth quarter could reach 3.5%.

High interest rates first increase borrowers' monthly payments. To understand this through a teaching example: borrowing $400,000 through a fully amortizing, 30-year fixed-rate mortgage, monthly principal and interest is roughly $2,398 at 6%, and roughly $2,770 at 7.4%—an additional $370 each month. This example calculates loan principal and interest only.

After monthly payments rise, some buyers can no longer afford original prices, while others choose to remain on the sidelines, potentially reducing the pool of buyers sellers encounter. Consequently, the newsletter notes that some buyers are using high interest rates as leverage in negotiations, demanding price reductions from sellers. For the same home, buyers look not only at the listing price, but also at how much they can afford to pay each month.

When Tankers Are Attacked, Why Does It Impact Oil Prices?

The newsletter reports that oil tankers were attacked near Qatar and the UAE, with Brent crude trading near $103 per barrel; Houthi militants also claimed responsibility for an attack on Riyadh's airport, disrupting flights.

Oil prices are influenced both by current supply and by market expectations of future supply. Attacks along shipping lanes stoke buyers' concerns about whether crude oil can be delivered safely and on schedule, thereby heightening worries over supply disruptions. Fuel is a vital input for transportation and manufacturing, so rising prices can also push up costs for airlines, shipping, and businesses.

SpaceX Buys Spectrum: Why Does It Concern Traditional Telecom Carriers?

The newsletter reports that SpaceX reached an $8 billion deal to purchase wireless spectrum. Musk simultaneously expressed an extremely optimistic vision for the company's long-term valuation. Following the news, shares of T-Mobile US, Verizon, and AT&T fell in after-hours trading.

Spectrum refers to the range of radio frequencies used for wireless communications. For satellites and mobile phones to transmit and receive signals to and from each other, they must use appropriate frequency bands; spectrum usage rights are therefore a vital resource for mobile communications businesses.

Starlink's technical overview describes a more direct connectivity model: satellites equipped with Direct to Cell capability carry cellular communication equipment, allowing them to communicate directly with LTE phones and integrate into carrier networks like a roaming partner. (LTE is a cellular communication standard commonly used by mobile phones.)

This provides satellite networks the opportunity to extend connectivity into areas not covered by terrestrial cell towers. For traditional carriers, satellite operators could serve as partners to supplement coverage, but they could also introduce new competition. What makes the spectrum transaction noteworthy is its relationship to future communications services.

Goldman Sachs Equity Incentives and Spotify's Renewal: Who Is Being Paid For What?

According to a Bloomberg report cited in the newsletter, roughly 20 Goldman Sachs executives are expected to receive equity awards totaling over $500 million, with CEO David Solomon expected to receive around $100 million. This development followed record equity trading revenue at Goldman Sachs.

Another substantial sum comes from the content industry: the Wall Street Journal reported that Joe Rogan renewed his contract with Spotify for approximately $250 million. The newsletter also noted that when he first signed with Spotify in 2020, the contract was valued at around $100 million, and his show remains the most popular podcast in the United States.

The former awards equity, whose value fluctuates with the corresponding shares; the latter is a contract value, reflecting what a platform pays to secure a partnership with a leading show. Looking at both stories side by side reveals two approaches to capital allocation: how a corporation enables management to share in the value it creates, and how much a platform is willing to pay to aggregate an audience. The figures from Rogan's two contracts provide a sense of scale for comparison.

What Other Business, Policy, and Recommended Reading Items Were Included?

At the end of this edition, the newsletter also listed these news briefs and recommended reads:

• Apple Shifts M&A Leadership: Bloomberg reported that Apple is planning to replace M&A head Adrian Perica with Steve Smith, a longtime lieutenant to new CEO John Ternus.
• Microsoft H-1B Visa Eligibility Suspended: Citing The New York Times, the newsletter reported that Microsoft's access to the H-1B visa program for foreign tech workers was suspended, with Vice President JD Vance accusing the company of abusing the system.
• European Sovereign Debt Concerns: The newsletter recommended an article explaining why investors are concerned about a potential French debt crisis.
• American Express Hit with Anti-Money Laundering Fine: Citing the Financial Times, the newsletter reported that the Office of the Comptroller of the Currency (OCC) fined American Express $350 million for critical deficiencies in its anti-money laundering program.
• Former NFL Player's Mathematical Breakthrough: John Urschel, a former Baltimore Ravens professional football player and current MIT professor, may have achieved a major research breakthrough in mathematics, according to The Washington Post.
• Commercial Airline Pilot Safety Risks: A Wall Street Journal report examined safety risks posed by pilots and the difficult challenges the aviation industry faces in addressing them.