European software: rising AI token budget constraints, but consumption expansion remains the core validation signal
AI summary card
European software: rising AI token budget constraints, but consumption expansion remains the core validation signal
Morgan Stanley believes that companies such as Uber and Walmart beginning to limit budgets for AI tokens or AI coding tools will reinforce discussions around ROI and spending efficiency, but the more important signal is still that token consumption is scaling rapidly.
- Corporate AI spending is moving from the experimentation and rapid deployment phase into a phase that places greater emphasis on budgets, efficiency, and ROI; Uber is reportedly limiting monthly usage of each AI coding tool to $1,500 per employee.
- OpenRouter data shows that the annualized run rate of token consumption this week is expected to be 34.7 trillion tokens, significantly above 2.1 trillion tokens in early June 2025, indicating that demand is still expanding rapidly.
- The core of SAP's cloud growth remains the migration of on-premise customers to S/4HANA; the report believes the FY25 cloud conversion pace is about three times the run rate of the previous four years, while the remaining opportunity has only penetrated about 20%-25%.
- Legal AI expert interviews show that Harvey has high usage for general AI purposes, but Westlaw and Lexis still retain critical advantages in credibility and legal infrastructure for complex legal research.
- Spanish private software company Factorial completed a $150mn Series D financing at a $2.5bn valuation, reflecting continued financing interest in the European HR and business management software space.
Report interpretation
Overview
This report is Morgan Stanley's weekly snippets research on European software and services, focusing on AI token budget restrictions, SAP cloud migration, Microsoft and NVIDIA driving edge-side Agentic AI, Legal AI expert views related to RELX and Thomson Reuters, and Factorial's financing event. Overall, the report is not a deep single-stock rating report, but rather a tracking report on industry events and themes.
Core views
The core view is that the rapid growth in corporate AI token usage is beginning to trigger budget controls, which may spark short-term debates about ROI and spending efficiency, but the rapid consumption itself also validates technology adoption and the strength of demand for AI tools. Medium-term investment themes in the software sector still include SAP cloud ERP migration, edge-side AI inference, the data and workflow moats of legal information service providers, and activity in European private software financing.
Analysis framework
The report uses an event-driven weekly tracking approach, combining news reports, Morgan Stanley's existing research, company disclosures, expert interviews, and private financing information to assess changes in demand, budgets, product form factors, and the competitive landscape in the European software and services industry.
Methodology notes
Use corporate budget restrictions and OpenRouter token run rates together to observe AI adoption intensity and spending pressure.
Budget caps indicate that companies are beginning to manage AI costs more precisely, but rapid expansion in token consumption can still be seen as validation of demand for AI tools and stickiness of usage.
Assess the sustainability of SAP's cloud growth through the maintenance support base, S/4HANA migration pace, cloud revenue leading indicators, and breadth of the cloud portfolio.
The report believes SAP is still in the early stage of converting its €10.5bn maintenance support base into cloud subscriptions, and migration typically brings more than 2x revenue uplift.
Compare Harvey, Westlaw, and Lexis in general AI and complex legal research through interviews with heads of innovation at U.S. law firms.
The interviews emphasize that general AI interfaces and ease of use improve adoption, but relationships among legal precedents, citation systems, and the credibility of underlying corpora remain important advantages of professional platforms.
Use Factorial's financing size, valuation, ARR, and customer count to observe heat in the European HR software market.
Factorial raised financing at a $2.5bn valuation, about 2.5x higher than its $1bn Series C in 2022, showing that select software subsectors can still attract capital support.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- European software and services sectorCore industry asset pool covered by the report
- Strengths
- AI adoption, cloud migration, and vertical software workflows continue to provide structural growth themes.
- Weaknesses
- Corporate IT budget growth is only about +3.7% y/y, and AI spending may face budget constraints and ROI scrutiny.
- Comparison
- An In-Line industry view implies neither a clear bullish nor bearish stance relative to the European benchmark.
- Risks
- Wider token budget restrictions, AI spending crowding out other software budgets, and valuations embedding overly high expectations for AI commercialization.
- SAP SECloud ERP migration and European software fundamentals theme
- Strengths
- S/4HANA migration is still in the early stages, and converting the maintenance support base into cloud subscriptions could bring more than 2x revenue uplift.
- Weaknesses
- Recent current cloud backlog is soft, and AI and data monetization are not yet sufficient on their own to support the investment thesis.
- Comparison
- Compared with mature on-premise maintenance revenue, cloud subscription migration still has a longer revenue release cycle.
- Risks
- Slower cloud migration pace, delayed customer conversion, and insufficient visibility on AI monetization.
- Microsoft / NVIDIA RTX Spark Windows PCsEdge-side Agentic AI and edge inference theme
- Strengths
- Extending AI agents from the cloud to the PC can help deliver faster and more persistent personal agent experiences.
- Weaknesses
- The products are still in the early stage of promoting a new form factor, and actual enterprise adoption and supporting software ecosystems still need validation.
- Comparison
- Relative to traditional application-centric computing, this path emphasizes an operating experience driven by natural language and AI workflows.
- Risks
- Hardware costs, enterprise deployment pace, and edge inference experience and security governance falling short of expectations.
- RELX / Thomson ReutersLegal AI and professional information services moat theme
- Strengths
- Westlaw and Lexis have long-standing strengths in legal corpora, citation systems, and integration with legal systems.
- Weaknesses
- General AI tools such as Harvey are attractive in interface, ease of use, and familiarity.
- Comparison
- Harvey is better suited for general AI use cases, while Westlaw and Lexis remain more trusted in complex legal research.
- Risks
- General AI outputs may appear comparable but may fill gaps rather than be based on complete legal corpora, potentially causing misuse or competitive disruption.
- FactorialEuropean private HR and business management software financing case study
- Strengths
- Completed a $150mn Series D financing at a $2.5bn valuation, serving more than 16,000 enterprises.
- Weaknesses
- Competition among software platforms serving SMEs is intense, and the company must continue proving ARR growth and unit economics.
- Comparison
- It operates in a broadly similar software submarket to Personio, BambooHR, and Rippling, with adjacent business overlap from Sage HR/payroll.
- Risks
- Valuation digestion, SMB demand volatility, and intensifying HR software competition.
- Uber / WalmartCorporate AI token budgeting and usage governance case studies
- Strengths
- High usage of AI tools shows that internal enterprise AI adoption is progressing quickly.
- Weaknesses
- Rapid budget consumption is forcing companies to set usage caps or restrictions on internal AI tools.
- Comparison
- These cases represent AI moving from the experimental deployment phase into the cost governance phase, rather than merely a stage of trying new technology.
- Risks
- If ROI is hard to quantify, companies may slow procurement of AI tools or reallocate budgets.
Key data
- Report date2026-06-07The cover time is June 7, 2026 03:00 PM GMT.
- European industry viewIn-LineThe report is labeled Europe Industry View In-Line.
- Corporate IT budget growthabout +3.7% y/yThe report says AI spending still mostly comes from relatively modestly growing IT budgets.
- Uber AI coding tool budget limit$1,500 per month per employee for each AI coding toolBloomberg reported that Uber implemented a standard limit; the report also mentions that Uber's CTO previously said the 2026 budget for AI coding tools had already been used up in the first four months.
- OpenRouter token consumption run rate34.7 trillion tokens/weekly run rateThe report says this week's expected run rate is 34.7 trillion tokens, above 2.1 trillion tokens in early June 2025.
- SAP maintenance support base€10.5bnThe report believes SAP is still relatively early in the stage of converting its maintenance support base into cloud subscriptions.
- SAP migration revenue uplift2x+Migration of on-premise customers to S/4HANA Cloud typically brings more than 2x revenue uplift.
- SAP cloud conversion penetration20%-25%The report estimates that the remaining conversion opportunity is still large, with current penetration only around 20%-25%.
- Harvey weekly law firm usage rate82%The expert said that 82% of lawyers at their law firm use Harvey each week.
- Acceptable output ratio for Harvey on the same promptabout 50%The expert estimates that when Harvey, Westlaw, and Lexis handle the same query, Harvey is good enough or better about half the time.
- Factorial financing$150mn Series D, valuation $2.5bnThe valuation is about 2.5x the $1bn Series C valuation in 2022.
- Factorial scale metricsmore than 16,000 enterprise customers, reaching $100mn ARR in September 2025The company provides an HR and business management platform for small and medium-sized organizations.
Impact & implications
For investment, AI token budget restrictions should not be interpreted simply as weakening demand; they more likely mean that corporate AI spending is entering a stage where it can be measured, allocated, and governed. The opportunity for software companies lies in converting AI usage into demonstrable productivity, cloud migration revenue, and highly trusted workflows; the risks are unclear ROI, budget pressure, or AI feature monetization progressing more slowly than valuation expectations.
Risks
- AI token budget restrictions in enterprises may spread from isolated cases into broader spending discipline, suppressing growth expectations for some AI software revenues.
- ROI for generative AI tools remains difficult to quantify precisely; if evidence of productivity gains is insufficient, budget approvals may become more stringent.
- Most AI spending still comes from modestly growing IT budgets and may compete with traditional software, cloud services, or new hiring budgets.
- Although SAP cloud migration has long-term runway, soft current cloud backlog and limited visibility on AI monetization may affect market sentiment.
- Legal AI carries risks of incorrect citations, insufficient understanding of precedent relationships, and users placing excessive trust in outputs.
- Morgan Stanley discloses investment banking, shareholding, or other service relationships with multiple covered companies, and investors should be mindful of potential conflicts of interest.
What to watch
- Whether more companies in 2H26 disclose budget limits for AI tokens, AI coding tools, or internal AI platforms.
- Whether token consumption run rates on platforms such as OpenRouter continue to rise rapidly.
- SAP's current cloud backlog, S/4HANA migration pace, and the progress of converting the maintenance support base into cloud subscriptions.
- Whether companies begin directly comparing token spending with new hiring and traditional software budgets.
- The usage share, reliability, and competitive dynamics in professional research scenarios among Harvey, Westlaw, and Lexis in Legal AI.
- Whether edge-side Agentic AI devices promoted by Microsoft and NVIDIA can build enterprise applications and developer ecosystems.
- Financing, ARR, and competitive dynamics of HR software companies such as Factorial, Personio, BambooHR, and Rippling.