ReportOperationsAnnika Sorensen11 min read
ReportOperations

| Revenue line | Sequential growth |
|---|---|
| AI revenue | +29.6% |
| Total revenue | +1.2% |
| Revenue excluding AI | -1.3% |
Infosys Exhibit 99.1 and Exhibit 99.3 to Form 6-K filed 28 October 2026; Mosaic Singularity analysis, November 2026. AI growth is derived from the stated 8.2% and 10.5% shares and the stated 1.2% sequential growth; revenue excluding AI is the residual.
Infosys sold $5,145m of services in the quarter to 30 September 2026 and reported that 10.5% of its revenue was AI revenue, against 8.2% in the June quarter. The AI line grew 29% sequentially. Total revenue grew 1.2% in constant currency. Put those rates together on the company's own shares and revenue excluding AI fell about 1.3% over the three months, with the AI line taking 2.3 points of the mix. The AI book is not expanding the services market. It is being funded out of the account beside it, and Infosys is the party reporting both numbers.
For a CIO closing a technology budget by 31 December 2026, the question is no longer whether AI gets its own line. It is which line pays for it, and the supplier has already assumed the answer.
The derivation needs only what Infosys publishes. Take the June quarter's total revenue as 100 and its AI revenue as 8.2. A total that grows 1.2% becomes 101.2, of which 10.5% is 10.626 of AI revenue, so the AI line grew 29.6%, which is the 29% the company stated. The remaining 90.574 compares with 91.8 a quarter earlier. Infosys does not disclose what it counts as AI revenue, so the arithmetic cannot separate new AI work from traditional work relabelled as AI, and the chief executive was asked exactly that on 23 October 2026. What the arithmetic does settle is the total: 1.2% for the whole business, on a line inside it growing at 29%.
The guidance moved the other way. Infosys revised full-year revenue growth to 1%-2% from 1.5%-3% in the same release, kept the operating margin range at 20%-22%, and had grown 1.9% in constant currency across the half year. A services firm that cuts its growth range in the quarter its AI line grew 29% is not describing a market that got bigger. It is describing work moving from one column to another inside a budget that did not move.
The pricing conversation moved first, and it moved toward the buyer. On the earnings call of 23 October 2026, Salil Parekh, Infosys's chief executive, said requests for AI productivity benefits to be passed on in price run across the client base, and that some go beyond the return the company is prepared to accept. The rate increase Infosys had assumed for the year did not arrive. Its guidance for the second half names continued sharing of those benefits as a headwind, alongside wage increases and large-deal ramp-ups, which is a supplier telling its shareholders that the client keeps part of the saving.
So on the pricing, what we are seeing is that across the client base, there is demand or request requirement, which is focused on productivity benefits from the AI that are getting passed on.
The segment commentary puts it in the language of individual industries. In communications, Infosys reports clients maintaining tight control over discretionary spend and AI driven productivity expectations putting pressure on pricing and the delivery model. In retail, expectations for AI-led productivity commitments are said to be increasing competitive intensity and reshaping the commercial model. The thing being negotiated is not a discount. It is a productivity commitment written into the contract before the productivity exists, which is why a supplier calls it a reshaping of the commercial model rather than a price cut.
Infosys Exhibit 99.1 to Form 6-K filed 28 October 2026; Accenture Form 10-K for the fiscal year ended 31 August 2026, filed 16 October 2026; Cognizant Form 10-Q for the quarter ended 30 September 2026, filed 29 October 2026
Accenture makes the trade legible because it reports the charge separately. Operating margin rose to 15.4% from 14.7%, adjusted margin to 15.8% from 15.6%, revenue grew 5% in local currency and diluted earnings per share reached $13.56, up 12%. Underneath, the company recorded $308m of business optimization costs in fiscal 2026 and $615m in fiscal 2025, primarily for employee severance, and excluded both from the adjusted margin. The charge is excluded. The saving it buys is not, and it lands in the same year.
| Item | Accenture | Cognizant |
|---|---|---|
| Period | Fiscal year ended 31 August 2026 | Quarter ended 30 September 2026 |
| Operating margin, reported | 15.4%, against 14.7% | 15.4%, against 16.0% |
| Operating margin, adjusted | 15.8%, against 15.6% | 16.3%, against 16.0% |
| Programme and status | Business optimization program, completed in the first quarter of fiscal 2026 | Project Leap, initiated in the second quarter of 2026, running through the first quarter of 2027 |
| Charges recorded | $308m in fiscal 2026, $615m in fiscal 2025, primarily employee severance | $51m in the quarter, $135m since inception, of which $102m employee separation |
| Total expected | No total stated; the programme is complete | $180m to $250m, of which $150m to $200m severance and personnel |
Accenture Form 10-K for the fiscal year ended 31 August 2026, filed 16 October 2026; Cognizant Form 10-Q for the quarter ended 30 September 2026, filed 29 October 2026, Note 4
Cognizant is the only one of the three that prices the mechanism and names AI in the reason. Project Leap was initiated in the second quarter of 2026, and the filing's glossary defines it as a program aimed at streamlining operations and enhancing productivity through AI-led efficiencies. The expected cost is $180m to $250m, of which $150m to $200m is employee severance and other personnel costs. By 30 September 2026 it had recorded $51m in the quarter and $135m since inception, $102m of that employee separation, and it expects $140m to $160m in 2026 with the remainder through the first quarter of 2027. Adjusted operating margin rose 30 basis points while the reported margin fell 60, because the charge is excluded and the saving is not. The filing's own risk factor concedes the trade: the programme could disrupt the business and may not result in anticipated savings.
The saving is not a smaller firm. Accenture ended fiscal 2026 with approximately 814,000 people against 779,000 a year earlier, utilisation up to 93% from 92% and voluntary attrition down to 13% from 14%. Infosys added 4,000 net in the quarter, hired 8,000 graduates, plans approximately 20,000 for the year against attrition it puts near 13%, which it translates as roughly 45,000 people leaving annually, and runs at 85% utilisation, which its management calls optimum. None of these documents describes a workforce being replaced by software. What the severance buys is a workforce reshaped rather than reduced: total headcount rose at both firms that report it, while the charge was taken once and the benefit runs through later quarters.
Accenture's bookings table is the leading indicator, and it is weaker than the revenue line. Total new bookings rose 3% in local currency against revenue up 5%. By type of work, consulting bookings rose 6% in local currency while consulting revenue rose 3%, and managed services bookings were flat in local currency while managed services revenue rose 6%. Managed services is the multi-year operation, maintenance and infrastructure work that a client buys as a run-rate rather than as a project. In fiscal 2026 the book that funds it stopped growing while the revenue from it grew fastest.
One year is one year. New bookings include renewals, extensions and changes to existing contracts, and Accenture's own filing says they are not comparable to, nor should they be substituted for, an analysis of revenues over time, and that the timing of a small number of large managed services contracts can move the number. The signal is worth testing at the next release, not worth a forecast on its own.
The case against this reading is strong and it comes from the same company. Infosys signed $5.2bn of large-deal total contract value in the quarter, 62% of it net new, including one deal over $2bn, and Parekh rejected the substitution reading directly. Asked whether the AI revenue was incremental or traditional work relabelled, he said: “It is not something which is replacing what we have. There is a lot of new work that we are doing.” Both can be true. Total contract value is the whole of a multi-year deal, not a client's annual spend, and a client can sign new AI scope while holding the annual number flat by retiring work it no longer wants. The revenue line is the budget, and it grew 1.2% in the quarter and 1.9% in the half against guidance cut to 1%-2%. What survives is narrower than the claim it started from, and it is still the decision in front of a buyer: on these three issuers, AI work is being funded inside a flat annual budget, and the deal book is not yet converting into a larger one.
The budget closes in December and the seats do not owe the same thing. The supplier has already assumed the re-cut; the buyer is the only party still deciding.
The question is not whether the company is buying AI. It is whether the same outcome is being bought twice, once through the run-rate and once through the programme, and the budget is the last document that can answer it before the money is committed.
The number to hold a supplier to is not its AI growth rate. It is the share of the client's annual spend the AI line represents, and whether the client chose to move it or had it moved for them. Infosys has published the arithmetic for one quarter. Every buyer carrying a flat run-rate and a rising AI line is looking at the same subtraction, and a budget signed in December is where it stops being a trend and becomes a decision.
Share this
Related