Infosys Integrated Annual Report 2025-26: A Review

Official report: Get the Infosys Integrated Annual Report 2025-26 from the company's website. This page is The Footnotes' independent review of it.

Verdict: Infosys prints the most honest highlights page in Indian IT, then lets the rest of the report run on habits from 2018. The numbers are labelled and nothing flattering is cherry-picked. But the strategy diagram, the outcomes column and the sign-off page look carried forward rather than rethought. Score: 25 of 35.

Page numbers below are the printed numbers on the report's pages, not PDF page counts. The integrated section is pages 14 to 61, laid out as two-page spreads. The Board's report starts at page 64, the BRSR at page 171 and the financial statements at page 225. Screenshots are small excerpts reproduced for review and remain the copyright of Infosys Limited.

Clients come before the contents page

Most Indian annual reports open with a chairman, a theme line or a montage. Infosys opens with twelve pages of client stories before you reach the contents (pp. 2-13): Liberty Global, Hertz, Mondelēz, Microsoft and ABN AMRO.

They are marketing, but at least one has numbers a reader can check against the claim. The Microsoft page (pp. 10-11) says Infosys's agents cut the time to produce a root cause analysis from five days to one, and critical-incident engagement from 15 minutes to under 10. That is more evidence for "AI-led services" than most IT companies give in their whole report. The Hertz page is about rewriting COBOL, which says something honest about where AI revenue actually comes from: old code, not new products.

The cost is that the case studies push the integrated report to page 14, and they are not linked to anything later. None of the client numbers reappears in the intellectual capital page or the value creation model.

The highlights page every IT company should copy (pp. 30-31)

This page is why the report scores 5 for financial storytelling.

• The adjustment is labelled where it appears. The tile reads "Adjusted operating margin(3) 21.0%", and footnote 3 on the same spread says it excludes the ₹1,289 crore Labour Codes charge. You don't have to hunt for it in the notes to accounts.
• The table keeps the statutory figure. Net profit of ₹29,440 crore against ₹26,713 crore is marked as attributable to owners. It is not an adjusted number.
• Growth is shown both ways. 9.6% reported and 3.1% in constant currency, side by side on one tile.
• The bad number is printed. Market capitalisation is shown for five years: ₹5,07,192 crore at March 2026, down from ₹6,52,332 crore a year earlier and ₹8,02,162 crore in FY 2022. In dollars, US$54.8 billion against US$75.8 billion.

Very few companies put a 22% fall in their own market value in a highlights table. It tells the reader the numbers weren't picked for the photograph.

Where the cash went: a buyback, and a payout below policy

Two details are easy to miss and worth reading together.

The ₹18,000 crore buyback is set out in full on page 67. The Board approved it on 11 September 2025 and shareholders approved it by postal ballot on 4 November. Infosys bought 10,00,00,000 shares, about 2.41% of paid-up capital, at ₹1,800 each through a tender offer open from 20 to 26 November. It closed on 4 December 2025. That is the level of detail a shareholder deciding whether to tender actually needs.

The Financial Capital page (pp. 50-51) says over ₹37,500 crore was returned to shareholders for fiscal 2026. It then says Infosys has returned "approximately 82.1%" of free cash flow for fiscals 2025 and 2026. The policy on page 67 is to return about 85% over a five-year period. So even a record buyback leaves the company slightly behind its own target two years into the five. The report states both numbers but never puts them next to each other or explains the gap. It should.

A chairman's letter with an argument, and a diagram stuck in 2018

Nandan Nilekani's message (pp. 32-33) is one of the best-written letters in this year's Indian reports, because it makes an argument rather than a list. It names the question investors are asking: if coding becomes automated, why is Infosys needed at all? It answers it with testing, validation, architecture, security, data governance and the "deployment gap" in large enterprises. It contains no financial figures, which is right when the CEO's letter that follows has them.

Salil Parekh's letter gives the one AI number that matters: revenue from the six AI-first areas was 5.5% of total revenue in the third quarter. The strategy pages (pp. 43-47) then set out the six pillars (Orchestrate, Insight, Transform, Modernize, Innovate, Assure), split into "AI First" and "AI Augmented" services.

The problem is on the next spread. The strategy section still opens with "In 2018, we embraced a four-pronged strategy" (p. 42). And it is that 2018 diamond, not the six AI pillars, that sits at the centre of the value creation model (pp. 48-49). The words have moved on. The diagram hasn't.

The capitals: numbers in, adjectives out

The inputs and outputs are real numbers: ₹92,852 crore of net assets and ₹43,075 crore of cash in; 3.1% constant-currency growth and 31.6% ROE out. 3,28,594 employees in; 39.5% women, 155 nationalities and 84% "AI-aware" employees out.

The outcomes column is where it thins out. Financial capital's outcomes are "Profitable growth". Human capital's are "Best-in-class employee experience and learning". An outcome should be a measured change, such as attrition, client satisfaction or an emissions-intensity trend. One output, "#1 Rating in AI", has no source at all. (Our guide to value creation models in Indian integrated reports shows what a measured outcomes column looks like.)

Each capital then gets one page (pp. 50-61). That keeps the report short. It also means nothing connects the 84% of AI-aware staff to the 5.5% of AI revenue, or the patents to Topaz sales. The links are implied by sitting on neighbouring pages, not shown.

"Carbon neutral", with Scope 3 in the back

The reduction work is genuine: Scope 1 and 2 emissions down 71.22%, 81.8% renewable electricity in India, and 84% of office space at the highest green-building rating. The neutrality claim, though, covers Scope 1, 2 and 3 and depends on ten offset projects (p. 56).

The Scope 3 figure itself is not in the integrated section. It sits in the BRSR at page 211: 2,07,374 tCO2e against 2,08,659 the year before, a 0.6% fall. There is no science-based reduction target for it. TCS, by contrast, reports a 35% absolute Scope 3 reduction target and shows progress against it. Offsets deliver neutrality whatever the trajectory, so a report that leads with "carbon neutral" should print the trajectory next to it.

Assurance is strong: Deloitte Haskins & Sells LLP gives reasonable assurance on BRSR Core (report from p. 167, engagement letter dated 16 January 2026). Materiality is weaker. The BRSR says the ESG Committee approves "the continued relevance" of material topics each year, which is not the same as reassessing them, and no assessment date is given.

Signs of a report on autopilot (p. 16)

Two lines on the about-this-report page:

• The framework is credited to "the International Integrated Reporting Council (IIRC)". The IIRC became part of the IFRS Foundation in 2022.
• The report "has been reviewed and approved, for publication, by the Management of the Company". The Board's report covers the statutory content. But the integrated narrative is exactly where Board ownership matters, and HDFC Bank's report is approved by its Board.

Neither is serious alone. Together with the 2018 diagram, they suggest pages that are copied forward each year rather than re-read.

On the digital side, the contents page (p. 15) links to an online version of the report, and the PDF's bookmarks work. But the PDF is built as spreads, which is awkward on a phone, and the SDG mapping lives in a separate ESG report.

What other companies can copy

• Label every adjusted measure where it appears (p. 31). A footnote marker on the tile and the explanation on the same spread. No reader should have to dig through the notes to accounts to find it.
• Print market capitalisation in the five-year table, even in a bad year (p. 30). It costs nothing and buys credibility for every other number on the page.
• Explain a buyback like a prospectus, in one paragraph (p. 67): size, share count, percentage of capital, price, method, dates.
• Give the chairman's letter an argument (pp. 32-33). One question investors are asking, answered.
• Put a hard number on a client story (pp. 10-11). "Five days to one day" is evidence. "Transformational partnership" is not.

And what not to copy: a value creation model whose centre is an old strategy, an outcomes column without numbers, and a neutrality claim without the emissions figure next to it. For how the cover itself compares with other reports this year, see annual report covers compared.

The score

How it compares

Every review on The Footnotes uses the same seven-point card, so scores can be compared. Scores are editorial opinion of the published document.