Summary

  • ServiceNow's Q2 2026 GAAP subscription gross margin was 73.5%, down 650 basis points from 80.0% a year earlier, even as subscription revenue grew 24.5% to US$3.877 billion.
  • Subscription cost of revenue increased US$405 million. Four disclosed drivers—personnel, infrastructure and software, third-party cloud commitments and acquired-intangible amortisation—sum to US$394 million, or 97.3% of that increase.
  • The remaining US$11 million is not separately explained, and the filing does not support attributing the whole margin change to AI, hyperscalers or any named acquisition.

Growth did not preserve the old delivery margin

ServiceNow added US$764 million of subscription revenue in the second quarter, moving from US$3.113 billion in 2025 to US$3.877 billion in 2026. GAAP subscription gross profit rose too, from US$2.488 billion to US$2.847 billion. The important asymmetry is that gross profit gained US$359 million while the implied cost of delivering subscription services gained US$405 million.

The arithmetic is visible without inventing a metric. Revenue less gross profit puts subscription cost of revenue at US$625 million in Q2 2025 and US$1.030 billion in Q2 2026. The difference is the same US$405 million increase described in the 10-Q. Because cost expanded faster than revenue, the earnings exhibit shows the GAAP subscription margin falling from 80.0% to 73.5%, a decline of 650 basis points.

The 10-Q narrative prints the same comparison at whole-percentage resolution: 73% against 80% for the quarter and 75% against 81% for the first half. That is not a second economic result. It is a coarser presentation of the exhibit's decimal margin. The 650bp measure belongs to the exhibit; calling the rounded prose a separate 700bp event would manufacture precision from rounding.

Four disclosed lines almost, but not completely, foot

ServiceNow identifies four cost movements in the paragraph explaining subscription cost of revenue. Personnel-related costs, including stock-based compensation and overhead, increased US$115 million. Infrastructure-hardware depreciation plus software, maintenance and other costs supporting data-centre capacity rose US$63 million. Expenses tied to contractual commitments with third-party cloud providers rose another US$63 million. Amortisation of acquired intangible assets increased US$153 million.

Together, those figures equal US$394 million. They account for about 97.3% of the US$405 million increase, leaving US$11 million between the itemised amounts and the reported total. The filing does not provide another line for that remainder. It could reflect other costs, rounding or both; choosing one would go beyond the disclosure.

Nor does ServiceNow label the paragraph an exhaustive attribution bridge. The categories are disclosed drivers, but their boundaries need not be perfectly independent. “Personnel” includes overhead, while the infrastructure line includes software, maintenance and other costs. A responsible reading therefore uses the four figures to understand scale and composition without pretending every dollar occupies a unique box.

The largest item, US$153 million of incremental acquired-intangible amortisation, is non-cash in the quarter. It reduces GAAP gross profit, but it is not a US$153 million supplier payment made during Q2. The filing also does not assign it to Armis, Moveworks, Veza or another named acquisition. Purchase accounting creates a time path for expense; it does not publish a transaction-by-transaction margin bridge here.

The pressure was visible before the second quarter

The first quarter provides a useful sequence. ServiceNow reported a 78% subscription gross profit percentage in Q1 2026, down from 81% a year earlier. Subscription cost increased US$259 million. The company then identified the same broad set of drivers: US$114 million of personnel costs, US$60 million of infrastructure and software costs, US$41 million of third-party cloud commitments and US$41 million of acquired-intangible amortisation.

This does not make Q2 a mechanical continuation; the mix changed sharply, particularly amortisation. It does show that the cost question did not begin with one July headline. Full-year 2025 subscription margin had already moved to 80% from 82% in 2024, and the 2025 annual filing expected a further decline in 2026, primarily because of growing third-party-cloud use and incremental acquisition amortisation.

The second-quarter filing now expects subscription cost of revenue to rise both in dollars and as a share of revenue for the full year. Its list is wider than AI. It includes serving more customers, increased usage inside customer instances, public-cloud offerings for regulated markets, changing data-residency requirements and continuing acquired-intangible amortisation. Those are distinct economic mechanisms: labour and support intensity, owned or leased infrastructure, external cloud consumption and accounting expense from past capital allocation.

AI is part of the forward context, not the historical bridge

ServiceNow said AI annual contract value crossed US$1 billion in Q2. Its earnings exhibit also says full-year gross-margin guidance reflects more customers using hyperscaler partnerships and faster customer AI adoption. Those statements make AI-related delivery economics a legitimate monitoring question. They do not establish that AI caused the measured 650bp year-on-year Q2 decline.

The company publishes no standalone AI revenue, gross profit or cost-of-revenue schedule in these documents. Nor does it name a hyperscaler in the US$63 million increase tied to third-party-cloud commitments. AI adoption can increase consumption, strengthen demand, improve product value and create new delivery expense at the same time. Without a product-level bridge, separating those effects is not possible.

Demand is clearly not absent. Current remaining performance obligations reached US$13.20 billion, up 21%, while total remaining performance obligations reached US$29.0 billion, also up 21%. These measures support a strong contracted-demand picture. They are not cash, recognised revenue or proof that a particular product offsets a particular cost. A durable margin assessment needs the conversion path from contract to revenue, from usage to delivery cost, and from acquisition accounting to eventual cash returns.

The next result should be read as a mix test

The question for ServiceNow is not whether subscription growth can coexist with a lower GAAP gross margin; Q2 shows that it can. The question is which component becomes structural. Personnel and data-centre investment may create operating capacity. Third-party cloud use may buy geographic reach, regulated-market compliance and elasticity while giving part of the unit-cost curve to outside providers. Amortisation records the cost of acquired technology and relationships even when the quarter's cash has already moved at another time.

If revenue continues to outgrow the operating delivery lines while amortisation peaks, the Q2 decline could prove transitional. If third-party-cloud expense and support intensity keep growing as a share of subscription revenue, the lower margin may reflect a more durable delivery architecture. The public record cannot choose between those paths yet. It can, however, define the next evidence: a stable or improving GAAP margin, a narrower gap between revenue and recurring delivery costs, and enough product-level disclosure to connect AI adoption with economics rather than slogans.

Sources