Microsoft Beat Every Number. A Lease Reclassification Moved $15 Billion Off The Reported Capex Line. Its Auditor Has Been In The Seat Since 1983.

GOVERNANCE WATCH

By InsidEntity Editorial Desk · Jul 30, 2026 · 14 min read

Company Risk Rating: 3.30, Good

Auditor Independence pillar: 0.00

The quarter Microsoft reported on 29 July 2026 was the strongest in the company’s history by almost every measure a shareholder tracks. It was also the quarter in which the company disclosed two changes to how it will account for the largest capital programme in corporate history. Both changes are legitimate. Both are disclosed. Both sit squarely inside the category of management judgment that an audit committee and an external auditor exist to test.

Microsoft’s external auditor has held the engagement for forty-three years.

1. What was reported

The three-quarter sequence matters more than any single print, because the market’s reading of Microsoft reversed inside it.

Full year FY26: revenue of $331.8 billion, up 18 percent, and operating income of $155.2 billion, up 21 percent. Azure crossed $100 billion in annual revenue for the first time.

Editorial note: the Q4 Intelligent Cloud segment revenue figure carried in the original draft ($39.3bn) conflicted with a second figure found in verification ($37.8bn). Rather than publish either uncorroborated, that line has been pulled pending a direct check against Microsoft’s own segment disclosure in the 10-K.

Two features of this table deserve to be separated out before anyone draws a conclusion from it.

The GAAP line is doing work the operating business did not do. Q2’s 60 percent GAAP earnings increase included a $7.6 billion gain on Microsoft’s OpenAI holding following its recapitalization, worth $1.02 of the EPS figure, confirmed against Microsoft’s own Q2 FY26 disclosure. Q4’s GAAP result included a $3.2 billion gain on Microsoft’s Anthropic investment, worth $0.33 of EPS, and a further gain on the OpenAI position worth $0.07, not a loss. Neither reflects cash received or a product sold. Both reflect a revaluation of holdings in privately held companies, where there is no market price to observe and the carrying value rests on estimation. Non-GAAP is the cleaner comparator across these three quarters, and non-GAAP still shows acceleration: 24 percent, 21 percent, 23 percent.

The Azure trend inverted the narrative. Q2’s print triggered a share price fall of roughly 10 percent on concern that Azure was decelerating while AI capital expenditure compressed margins. On the Q2 call, management guided Q3 Azure growth to 37 to 38 percent in constant currency. It came in at 40 percent, then Q4 guidance of 39 to 40 percent was itself beaten at 43 percent. Shares rose about 7 to 8 percent after the Q4 release. Hood has since guided 45 percent constant currency Azure growth for Q1 FY27, above the roughly 41 percent analyst consensus at the time of guidance.

The forward indicator behind that reversal is commercial remaining performance obligations of $678 billion, up 8 percent sequentially, described by management as evidence that the infrastructure spend is contracted against real demand rather than speculative.

Set against that: fourth-quarter free cash flow of $19.64 billion, down 23 percent, against $41 billion of reported capital expenditure and finance leases in the same quarter. Capital spend exceeded free cash flow. On the reported figures that is the first quarter in which that has been true since before the AI buildout began.

2. The two accounting changes, stated precisely

Both take effect at the start of FY27. Neither is retroactive to the quarter above. Both were disclosed by CFO Amy Hood on the earnings call.

Change one: useful life. The estimated useful life of data centres and office buildings, structural and infrastructure assets, extends from 15 years to 25 years, which management attributes to operating history and expected use of the assets. The disclosure names data centres and office buildings specifically, not servers or networking equipment, which carry their own shorter depreciation schedules and are unaffected. That distinction matters: extending the life of the building that houses the hardware is a more defensible estimate than extending the life of the hardware inside it. Hood’s own characterisation is that the change affects only the timing of future depreciation and is expected to deliver minimal benefit to FY27 operating income.

Change two: lease classification. More future data centre leases will be structured as operating leases rather than finance leases. Hood identified this, not the depreciation change, as the larger effect. The mechanism is simple and she stated it plainly: finance leases are included in capital expenditure, operating leases are not.

The consequence. Outside the useful life impact, calendar 2026 capital expenditure expectations are unchanged. Including it, the expectation is adjusted to approximately $175 billion, against roughly $190 billion previously guided.

Roughly $15 billion of planned infrastructure investment shifts from owned, finance-leased assets to operating-leased assets, reducing reported capital expenditure while increasing operating lease obligations. Under ASC 842, operating leases remain on the balance sheet through right-of-use assets and lease liabilities. The economic obligation does not disappear. It changes where it is presented. The data centres still get built. The obligations still exist. They are recorded elsewhere.

This is worth stating carefully, because the distinction is the whole point. Microsoft has not hidden anything. The change was announced on the call, in advance of the period it affects, with the mechanism explained. Every element of it appears to be within the accounting standards. A reader who assumes wrongdoing here has misread the article.

What has happened is narrower and more interesting. Two estimates that determine how the largest capital programme in corporate history appears in the accounts have both been revised in the direction that makes the programme look less intense on the reported capex line, in a period when capital intensity had become a central focus for investors. Useful life extended means depreciation spread thinner per year. Leases reclassified means reported capital expenditure lower. Neither changes the cash, the concrete, or the underlying risk.

Most accounting controversies do not arise because accounting standards are violated. They arise because accounting standards permit judgment. Governance exists to ensure those judgments remain objective, consistent, and capable of withstanding independent challenge. Estimates like the two above are precisely what independent assurance is for.

3. Why this is a governance question even when everything is compliant

Judgments of this kind are not audited in the sense of being recalculated. They are challenged, or they are not. An audit committee asks whether a 25-year life is supportable for the buildings and infrastructure it now covers. It asks whether lease structuring is being driven by economics or by presentation. It asks what the fair value of a private stake in an AI laboratory rests on, and whether the resulting gain belongs in this quarter’s earnings.

Compliance is not the end of that inquiry. An estimate can sit comfortably inside the standards and still be the wrong estimate. The standards define the boundaries of what is permissible. The audit committee and the external auditor exist to test whether management’s choice within those boundaries is the right one for shareholders. That is why governance matters even when, especially when, nothing alleged is improper.

Whether those questions get asked with force depends on the independence of the people asking. Which brings us to the pillar.

Deloitte & Touche’s own signed audit report states it plainly: “We have served as the Company’s auditor since 1983.” That is three years before Microsoft’s 1986 initial public offering, and forty-three years of continuous engagement, predating, not merely spanning, the entire public life of the company.

Microsoft does not treat this as a weakness. Its proxy statement makes the affirmative case, identifying two benefits of an auditor with an extensive history with the company: higher quality audit work and accounting advice arising from Deloitte’s institutional knowledge of the business, its accounting policies, its financial systems and its internal control framework; and operational efficiencies with a resulting lower fee structure because of that familiarity.

The audit committee also discloses genuine mitigations. Lead engagement partners rotate on a five-year cycle under SEC rules, a prior rotation is on record beginning with the FY2020 audit, and the committee selected a new lead partner beginning with the FY2025 audit, consistent with that cycle. All services, audit and non-audit, require committee pre-approval. The policy prohibits the auditor from providing bookkeeping or financial systems design work. Only independent directors serve on the board’s committees. Eleven of twelve board nominees were disclosed as independent.

Those are real controls and they should be weighed. They are also, every one of them, controls on the individuals and the services. None of them addresses the tenure of the firm, because none of them can. A relationship of forty-three years is a commercial relationship of forty-three years regardless of which partner signs.

The governance concern here is not that a forty-three-year relationship implies a compromised audit. It is that long tenure increases the risk of perceived familiarity influencing judgment, independent of whether actual independence has been impaired, and regulators and governance frameworks generally treat both as relevant, since investor confidence depends on the perception as much as the fact. And the company’s own stated benefit contains the difficulty. Lower fees arising from familiarity is a description of an efficient relationship. It is also a description of a client the firm has strong reason to keep.

4. The arithmetic

InsidEntity’s Company Risk Rating weights four pillars 40% Director Independence, 20% Director Capacity, 20% Auditor Independence, 20% Shareholder Influence, each scored 1 to 5. Until a company’s directors complete the independence questionnaire, the pillars are weighted equally and Director Independence carries a 5.00 default. Under this methodology, Auditor Independence is scored on a tenure ladder: 1 to 3 years scores 5, 4 to 5 years scores 4, 6 to 7 years scores 3, 8 to 9 years scores 2, 10 years scores 1, and beyond 10 years scores 0. A score of 0.00 reflects that ladder alone; it is a structural governance indicator based on firm tenure, not a judgment on the technical quality of any audit work performed.

Microsoft’s Company Risk Rating is 3.30, Good. Until Microsoft’s directors complete the independence questionnaire, the four pillars are weighted equally and Director Independence carries its 5.00 default, so the rating is their average: (5.00 + 3.21 + 0.00 + 5.00) ÷ 4 = 3.30. Once the questionnaires are in, Director Independence takes its full 40% weight. What is not in question is the direction of the finding: on every pillar but one, Microsoft scores at or near the top of the scale. On Auditor Independence, a forty-three-year, uninterrupted relationship, predating the company’s own IPO, puts it at the bottom.

5. What to watch

FY27 Q1, reporting late October 2026. Guided capital expenditure of more than $50 billion, including the lease reclassification effect. This will be the first period showing the new classification in practice, and the first opportunity to see the gap between reported capital expenditure and total commitment including operating leases.

Operating lease disclosure in the FY27 10-Q filings. The $15 billion does not vanish from the accounts, it should appear as operating lease commitments. The disclosure quality there, and whether analysts adjust for it, is the practical test of whether the reclassification changed the optics or the underlying understanding.

The FY2026 proxy statement, expected October 2026. Two items. First, the Deloitte audit and non-audit fee split, the single most useful available indicator of whether the relationship remains predominantly an assurance relationship. Second, whether the audit committee’s report addresses the useful life change and the lease classification decision, and in what terms. A committee that discusses a significant estimate change in its own report is doing something visible. One that does not, is not.

Fair value inputs on the OpenAI and Anthropic holdings. Two of the last three quarters had GAAP earnings materially moved by revaluations of unlisted holdings. As those positions grow, an increasing share of reported GAAP profit depends on management estimates of private-company value rather than recurring operating performance, a trend distinct from, but adjacent to, the audit-tenure question above.

Board leadership structure. Satya Nadella holds the roles of Chairman and Chief Executive Officer. The board’s counterweight is a designated lead independent director. Our four pillars do not currently score combined chair and chief executive roles, which means this does not appear in the rating. It is still the arrangement under which every judgment described above is overseen.

Confirmed and contested

Confirmed. All financial figures are from Microsoft’s own Q4 FY26 earnings release of 29 July 2026, the accompanying earnings call, and the Q2 and Q3 FY26 releases, cross-checked in this pass. The accounting changes and their effect on guided capital expenditure are management’s own disclosure, in management’s own words, including confirmation that the useful-life extension applies to data centre and office building assets specifically, not servers or GPUs. Deloitte’s tenure since 1983 is drawn directly from the firm’s own signed audit report; the stated benefits of that tenure, the partner rotation cycle, and the pre-approval policy are from Microsoft’s proxy statements. Nothing in this article rests on an allegation, an unnamed source, or a claim Microsoft disputes.

Contested or unresolved. The Q4 Intelligent Cloud segment figure could not be reconciled between two sources in this pass and has been removed from the table pending a direct check against the 10-K. Third-party FY27 capital expenditure projections, which range above $250 billion, are analyst estimates and not company guidance. The platform’s displayed overall CRR of 3.30 does not exactly match a 40/20/20/20 weighting of the four individually-cited pillar scores, which computes to 3.64; this gap is noted rather than resolved in this pass.

No wrongdoing is alleged. The accounting changes described here appear to comply with applicable standards and were disclosed proactively. The question this article raises is not whether Microsoft’s accounting complies with U.S. GAAP. It is whether investors can be confident that the most consequential judgments behind these numbers were subjected to demonstrably independent challenge.

The quarter answered the question investors were asking, which was whether the spending converts to revenue. Azure at 43 percent and a $678 billion contracted backlog say it is converting.

The question nobody asked is the one the accounts just changed the answer to. Investors spent six months arguing about how much Microsoft is spending. Starting in FY27 they will be arguing about a smaller reported number, and the difference will not be a single data centre.

About the InsidEntity Company Risk Rating

The InsidEntity Company Risk Rating (CRR) scores companies on a 1 to 5 scale across four governance pillars, weighted 40% Director Independence, 20% Director Capacity, 20% Auditor Independence, and 20% Shareholder Influence. A 5 is Benchmark, a 1 is Risk, and because Director Independence carries the heaviest weight, a weakness there moves the headline number more than a weakness in any other single pillar. The CRR measures governance structure and is not a measure of capital adequacy, liquidity, or solvency. Ratings update regularly as corporate changes occur, so the platform is always the current view. Explore ratings for companies across 145 stock exchanges at InsidEntity.

Know your entity. This is independent research, not financial advice.

Related reading

• Amazon Crossed $200 Billion In A Quarter. $53.4 Billion Of Its Profit Came From A Company It Does Not Consolidate. The Same Audit Firm Signs Off On The Other Half Of That Asset At Alphabet.
• Apple’s New CEO Has Never Sat on Another Board. The Man Checking Him Just Stopped Being Independent.
• Walmart Has one of the Strongest Governance Architecture We Have Profiled. It Scores 3.23 Anyway.

Correction, 27 September 2026: This article first gave Microsoft’s rating as 3.80 and could not reconcile it with the pillar scores shown here. The platform had dated Deloitte’s appointment to 2017; Deloitte’s own audit report says 1983. With that corrected, Auditor Independence is 0.00 on the platform as well as in this article, and the rating is 3.30, Good.