Microsoft's third-quarter Form 10-Q, filed with the Securities and Exchange Commission on April 30, contains the 45 percent figure that the analyst call surfaced a day earlier. It does not contain it on a cover page. It contains it on page nine, inside the management's discussion section, written as narrative — the kind of paragraph an investor reads only after they have already read the headline they came for. [1]

The paper's May 1 account of the analyst-call carve-out that put OpenAI at roughly forty-five percent of Microsoft's $627 billion remaining performance obligation treated the number as the disclosure event. It was the moment the share went from inference to attribution. The 10-Q clarifies what kind of disclosure event the company chose it to be. There is no 8-K. The amended partnership Microsoft announced April 27 — the rewrite of exclusivity, the introduction of an AGI carve-out — appears in the periodic filing only. [2]

This matters because the 8-K is the form a company files when something has happened that a reasonable investor would need to know before the next quarter's report. A counterparty concentration of roughly 45 percent in a $627 billion backlog is, by any traditional materiality test, an 8-K candidate. Microsoft did not file one. It chose the routing.

The 10-Q paragraph itself is careful. It describes the amended agreement, references the carve-out under which Microsoft "cannot, or chooses not to, support the necessary capabilities," and notes that ex-OpenAI commercial RPO grew 26 percent year over year. [3] The 26 percent figure is the load-bearing fact, because it is the number from which the 45 percent share is back-calculated; the company never quite says "OpenAI is forty-five percent of the backlog." It says the things from which that number follows.

That is a familiar architecture in technology disclosure. The headline number is set up to be derivable rather than stated. An investor who reads the press release sees a record backlog. An investor who reads the analyst-call transcript hears the 26 percent ex-OpenAI growth rate. An investor who reads the 10-Q narrative sees the carve-out language. Each layer is true. Only the last layer reveals the concentration.

The Om Malik write-up published May 1 is the one piece of independent press that walked the read in order. [4] Bloomberg and the Motley Fool covered the amendment and the backlog as a stock story. [5] [6] CNBC covered the rates implication. [7] None of them put the 8-K-versus-10-Q choice at the center, because none of them treats disclosure routing as a story. On X, the analyst Kakashii pinned it down: page nine of the 10-Q, no 8-K, the architecture is the disclosure.

The reason the architecture matters is that the 45 percent figure functions differently in different forms. In an 8-K, it is a current-event disclosure that any subsequent regulatory or counterparty action can be measured against. In a 10-Q narrative paragraph, it is a contextual datum inside a quarterly review, useful for valuation models but not legally framed as material on its own. A securities-law professor at the University of Pennsylvania, asked by Reuters in a separate context last year about similar architecture, described the practice as "compliant minimalism" — meeting the disclosure obligation by placing the disclosed fact at the lightest available form. [8]

There is also a counterparty question. OpenAI is preparing for a public offering inside an environment where the Tumbler Ridge complaints filed last week allege a twelve-engineer safety team was overruled, and where Florida's attorney general has expanded a criminal probe into a second homicide case. [9] [10] Microsoft's Apr 30 8-K, had it been filed, would have entered the same discovery and disclosure environment as those complaints. The 10-Q narrative does not.

The remaining question is whether the analyst desks treat the architecture as the story. So far they have not. Wedbush's note Friday continued to model the 45 percent as a known input rather than a new disclosure. Morgan Stanley referenced "the OpenAI exposure framework Microsoft has discussed." The framing is consistent with the routing. The 10-Q narrative reads to the sell side as a confirmation, not a filing event.

The paper's position from May 1 was that Microsoft's analyst-call carve-out was the disclosure that made counterparty concentration a number rather than an inference. That position holds. Today's revision is narrower: the company has now chosen the form of the disclosure, and the form is the lightest one available to it inside the existing periodic-reporting calendar. The 45 percent share is on page nine. The 8-K shelf is empty. The architecture is the answer.

What the next filing event tests is whether OpenAI's eventual S-1 — or any further amendment to the partnership — produces a disclosure event Microsoft cannot route into narrative. An IPO counterparty disclosing its own customer concentration is not Microsoft's choice. It is OpenAI's. That filing, when it lands, is the next test of whether the 45 percent number can stay on page nine.

-- THEO KAPLAN, San Francisco