The SaaSpocalypse That Wasn't: Why the ServiceNow Split Is About Pricing, Not AI Disruption

John Ian Lau — Technology and AI infrastructure desk

On Friday, UBS analyst Karl Keirstead cut ServiceNow to Neutral and slashed his price target from $170 to $100. Two trading days earlier, Guggenheim's John DiFucci had upgraded the same stock with a $125 target. BTIG's Allan Verkhovski is at $150 with a Buy. Of the 54 analysts covering the name, 43 still rate it a Buy. Median target is $140. Range from bottom to top is now $50, roughly half the entire share price. That is the widest coverage dispersion on any large-cap U.S. software name of the past decade, and it opened up in the span of five sessions.

The temptation is to read the split as the market finally admitting that generative AI eats enterprise software — the so-called SaaSpocalypse thesis that has been circulating on the sell-side since late 2024. That framing misses what the desks are actually fighting about. The debate is not whether AI displaces SaaS. It is which pricing model wins the transition, and whether it happens fast enough to bridge the seat-license cliff before the AI-monetization ramp fills in. The July 22 second-quarter print is now the cleanest read on that question the market will get before year-end.

What the bears see

The UBS downgrade is the most interesting piece of research on the sector this quarter, because it does not rely on a thesis about AI capability. Keirstead's argument is a demand-side observation from the buyer side of the enterprise software transaction.

In a note dated June 23 that has since been widely circulated, Karl Keirstead, Timothy Arcuri and Taylor McGinnis at UBS wrote that, based on conversations with roughly a dozen enterprise IT executives, about 60 percent of enterprises were "in some manner throttling AI spend" by adding guardrails such as token pooling, model downgrading, waste alerts and per-user usage limits. The analysts described the pattern as a "modest emerging headwind" and noted that "token spend optimization has become a key issue in most organizations." They characterized the shift as "a healthy problem" rather than an alarm.

The healthy-problem framing matters. Keirstead is not calling a top on AI adoption. He is calling a plateau on AI-linked incremental software spend at the enterprise level, and cutting his 2026 remaining-performance-obligation growth estimate for ServiceNow from roughly 20 percent to 16 percent to reflect it. That is the mathematical link between the UBS macro survey and the single-name downgrade. If the enterprise buyer is starting to instrument its AI budget the way it instruments its cloud budget, the incremental dollar available for the AI overlay that ServiceNow, SAP, Salesforce, Workday and Palantir all now depend on is smaller than the Q1 booking data suggested.

Karl Keirstead's downgrade note this Friday was more pointed. Reducing his fair-value estimate to $100, he flagged not just enterprise budget pressure but a specific new competitive vector: enterprises are increasingly evaluating AI agents for workflow automation instead of layering AI features onto existing ServiceNow modules. That is the SaaSpocalypse bear case in its most current form. It is a fragmentation risk, not an obsolescence risk, and it is more difficult to price.

What the bulls see

Guggenheim's DiFucci and BTIG's Verkhovski are looking at the same enterprise-buyer data and reading it in a different direction. The bull observation is that within the group of enterprises that ServiceNow already sells to — the Global 2000 — the AI overlay is not a discretionary line item, and it is not being throttled.

Allan Verkhovski at BTIG wrote in a June 10 note to clients that "while the transition to a three-tier pricing model by ServiceNow is not a forced migration, our survey suggests that the company is benefiting significantly in two ways: early renewal by existing customers, and widespread acceptance of double-digit price increases when upgrading to new tier packages that include AI capabilities." Verkhovski's checks found the company was "shifting more focus to the high-end market rather than the mid-low end market." Target price: $150.

Verkhovski's channel checks and Keirstead's channel checks are, on the face of it, in the same field with the same data. The interpretive gap is over how much of the throttling signal in the broader enterprise IT budget hits the specific dollar that gets renewed to ServiceNow. The BTIG read is that at the high end of the customer base, where seat counts are not the marginal renewal driver, the pricing story is running ahead of any budget-throttling drag. The UBS read is that within twelve to eighteen months, that gap closes.

The company's own answer

ServiceNow's management has been more direct than most CEOs of trillion-dollar-adjacent software franchises usually allow themselves to be. On the Q1 2026 earnings call in April, CFO Gina Mastantuono said, in an exchange with the covering analysts, that the annual contract value target for Now Assist was being raised from $1 billion to $1.5 billion — a fifty-percent revision one quarter into the fiscal year.

"We had a goal to be $1 billion on our AI commit this year, as you know, and I think we might have understated that a little bit," CEO Bill McDermott said on the same call, per the Fortune transcript. "We're already talking about $1.5 billion now, and it's on a run." Elsewhere in the call, McDermott framed the strategic pivot: "There has never been a tailwind for ServiceNow like AI." Mastantuono added that "50 percent of net new business now comes from a non-seat-based pricing model, including tokens and other assets such as infrastructure, hardware, and connectors."

That last number is the operative fact. The whole SaaSpocalypse debate hinges on the assumption that AI shrinks the seat count that drives SaaS revenue. When management can credibly show that half of net new bookings are already priced against consumption rather than seats, the debate shifts from an existential one to a timing one — a question of whether the consumption ramp compounds fast enough to offset the seat compression, and whether the customer base tolerates a second layer of usage-based pricing on top of the cloud-compute usage-based pricing they are already absorbing.

McDermott's most quoted line from the past month is the one that most cleanly captures management's read on the bear case. Asked in a June interview about AI models displacing traditional software, he called the concern "parlor tricks," arguing that direct-to-model AI implementations in many enterprises had turned out roughly ten times more expensive than the equivalent workflow overlay running through ServiceNow. The claim is unverifiable in aggregate, but it lines up with the UBS survey finding that enterprises are actively guardrailing token spend. If the direct-model economics are worse than the orchestrated-workflow economics at scale, the SaaS orchestrator's price power holds.

The wider read across the software complex

ServiceNow is the highest-signal name, but the same dispersion exists across the AI-application software cohort. SAP is heading into its July 23 earnings with UBS at EUR 205, Berenberg at EUR 215, Jefferies at EUR 210 and Bernstein Research at EUR 276 — a range of thirty-five percent from top to bottom. CEO Christian Klein has committed the company to a full rebrand of its cloud ERP as an "Autonomous Suite" priced in "AI Units" rather than seats, and has frozen hiring and travel to fund the pivot. Salesforce, Workday and Palantir all face variations of the same question. Morgan Stanley kept Palantir at Equal Weight with a $205 target last week, balancing what it called ten straight quarters of accelerating revenue against valuation risk.

The overlay is the same in each case. A pre-AI SaaS moat with a seat-license revenue base is being repriced against an AI moat with a consumption-based revenue base, and the market is trying to solve for the transition path in real time. When the incumbents move first on pricing, they preserve the moat. When they hesitate, the AI-native entrants build a parallel category around the workflow layer that would have belonged to them.

Our view

The July 22 ServiceNow print is the read that matters. Three numbers will decide the direction: the Q2 CRPO growth rate against Keirstead's revised 16 percent full-year path; the share of net new bookings now coming from non-seat pricing, currently 50 percent; and any narrative shift on the number of customers spending more than $1 million on Now Assist, which grew 130 percent year-on-year last quarter. If two of the three land above the current bear-case assumption, the UBS downgrade will be faded by the peers within a week, and the sector will re-rate. If two of the three miss, the SaaSpocalypse debate acquires a data anchor it does not currently have, and the entire AI-application cohort — SAP, Salesforce, Workday, Palantir — reprices lower into the second-half earnings cycle.

The right expression here is not a single-stock trade. It is a paired position that isolates the pricing-transition variable from the market beta. We would be long the AI-application software basket — an equal-weight mix of ServiceNow, SAP, Palantir and Salesforce — against short exposure to the traditional per-seat SaaS cohort where the AI overlay is still bolted on rather than built in. The pair captures the belief that the winners of the transition will re-rate positively as the consumption model proves out, while the losers face the compression the UBS bear case describes. On individual sizing, ServiceNow is the highest information ratio expression into July 22. It is the name where the dispersion is widest, the earnings catalyst is nearest, and the pricing-model shift is most measurable.

The SaaSpocalypse debate has been running for eighteen months. In two weeks it stops being a debate and starts being an earnings print.

This note reflects the views of Solomon Grey Capital's Technology and AI infrastructure desk as of the date of publication and is provided for informational purposes only. It does not constitute investment advice, an offer, or a solicitation to buy or sell any security. Past performance is not indicative of future results.