IBM’s AI Power Grab
The Stack Weekly: IBM’s move to buy Confluent signals that real-time data infrastructure is now the strategic chokepoint for AI
The Stack: Weekly Strategic Signals for Leaders Building What’s Next in AI and Software.
Capital & KPIs: IBM’s bid for Confluent shows that in 2025, the fastest way to win AI is to buy the data layer outright.
Enterprise Buyer Behavior: Stalled NRR and rising CAC confirm that expansion is no longer automatic and only value-aligned pricing models still grow.
Product & AI Bets: The stampede to usage-based pricing signals the end of flat-rate economics as AI workloads crush seat-based margins.
Moats & Models: The 25x spread in AI agent pricing exposes a market with no shared value standard and a moat that now belongs to whoever proves outcomes.
Each section also includes ‘other signals on our radar.’
Write back and let us know if you’d like to see more details on any of those.
1. Capital & KPIs
IBM Nears 11 Billion Confluent Acquisition
What Happened
On December 7, 2025, the Wall Street Journal reported that IBM is in advanced talks to acquire Confluent, a real-time data-streaming company founded by the creators of Apache Kafka, in a deal valued at about $11 billion. A formal announcement could come as early as Monday, December 8 or 9, though sources warned the negotiations could still collapse. Confluent’s market capitalization closed near $8 billion before the report, while IBM’s valuation was about $290 billion. Confluent’s technology handles high-volume, real-time data flows used in applications such as retail and financial-services AI systems, which fits IBM’s push into higher-growth cloud and AI infrastructure. The potential acquisition would be IBM’s largest in years and follows its $6.4 billion purchase of HashiCorp in 2024 as part of its shift toward hybrid cloud and AI platforms.
Why It Matters
IBM is about to overpay on purpose for the data layer that feeds AI, not another application logo. That tells you where strategic value is concentrating in 2025, in real time streaming, event driven architectures, and governed data that lets agents and models act on live signals instead of stale warehouses. It also says the largest buyers have decided it is faster to buy Kafka scale distribution than to build it, which will compress outcomes for everyone else in the data infra stack. If this clears, every board conversation about exits, partnerships, and valuations for infra heavy SaaS will be benchmarked against Confluent, not generic ARR multiples.
Implications for You
Data layer control now dictates pricing power. Application-only vendors without native streaming or event integrations will lose leverage at renewal.
Neutral data platforms are disappearing. With Confluent likely absorbed, independent SaaS will be forced to pick ecosystems instead of claiming cloud-agnostic positioning.
AI value shifts toward ingestion and orchestration. Vendors sitting only at the workflow layer will be pressured to prove deeper ROI or risk commoditization.
M&A windows tighten. Any SaaS company owning proprietary data movement, lineage, or governed event streams now holds a scarce asset and should expect inbound interest.
Other Signals on our Radar:
Salesforce Completes Eight Billion Dollar Informatica Acquisition to Anchor Its Agentic AI Data Strategy
Salesforce closed its eight billion dollar all-cash acquisition of Informatica, integrating the company’s data cataloging, integration, governance, quality, privacy, metadata management, and master data management capabilities into the Salesforce platform. Informatica will continue operating its AI-driven data management products and supporting its ecosystem of more than two hundred native connectors and five thousand customers. The deal follows Salesforce’s earlier platform-scale acquisitions of Mulesoft, Tableau, and Slack, positioning Informatica as the data foundation for Salesforce’s emerging agentic AI strategy.
2. Enterprise Buyer Behavior
2025 SaaS Benchmarks Show Median NRR Stalls at 101-102%; Usage-Based Pricing Delivers 10% NRR Lift
What Happened
Benchmarkit’s 2025 SaaS benchmark analysis of 936 companies found that median NRR for private B2B SaaS firms held steady at 101–102%, with the 75th percentile at 110%; public SaaS companies maintained a higher 108–110% median, and best-in-class performers exceeded 120%. NRR performance varied sharply by ARPA: companies with monthly ARPA below $25 posted just 94% median NRR with only 2% surpassing 100%, while those above $500 a month reached a 105% median with 47% exceeding 100%. CAC efficiency deteriorated industry-wide, with median New CAC rising to $2.00 in 2025 (up 14% from 2023) and blended CAC at $1.59. Usage-based pricing emerged as a major performance separator, with companies using consumption-driven models reporting roughly 10% higher NRR, about 22% lower churn, and nearly 2× faster growth than flat-rate subscription peers. Early-stage companies in the $1M–$5M ARR band recorded NRR in the low-100s with upper-quartile performers near 110%, frequently enabled by hybrid subscription-plus-usage pricing.
Why It Matters
Flat NRR in a year of rising CAC tells you expansion revenue is no longer a given. Buyers are not growing seat counts the way they used to, and procurement is blocking upsells unless there is direct value proof. The gap between low-ARPA and high-ARPA performance exposes a structural divide in the market: cheap tools cannot fund the customer success or product depth required to expand, while premium tools can. Usage-based pricing emerges as the only model consistently improving both retention and expansion, not because buyers love it, but because it aligns spend with realized value when budgets are tight.
Implications for You
Expansion is now earned, not assumed. Product depth and real adoption, not sales motions, determine whether NRR gets above 105 percent.
Low-ARPA SaaS is structurally disadvantaged. Sub-25-dollar tools cannot support the CS, onboarding, or integration work that drives expansion.
Consumption models outperform because they track value. When procurement scrutinizes every renewal, usage alignment removes friction and cuts churn.
CAC pressure forces PLG and hybrid pricing. Teams relying on paid acquisition alone will see payback windows break unless expansion drives lifetime value.
Other Signals on our Radar:
Enterprise AI ROI Surges as Seventy-Four Percent Report Positive Returns
A Wharton study shows that seventy-four percent of enterprises measuring AI ROI are already seeing positive returns, with tech and telecom at eighty-eight percent and finance and professional services at eighty-three percent. Weekly AI usage has reached eighty-two percent, nearly half of leaders use AI daily, and seventy-two percent now track formal ROI metrics. Core value drivers include data analysis, summarization, and document generation. Parallel research highlights strong financial outcomes, with AI customer service delivering returns of three dollars and fifty cents for every one dollar invested. The enterprise AI market is projected to grow from ninety-seven point two billion dollars in 2025 to more than two hundred twenty-nine billion dollars by 2030, while global AI spending is forecast to exceed two trillion dollars in 2026 as budgets shift from pilots to sustained, performance-based investments.
3. Product & AI Bets
Usage-Based Pricing Adoption Surges to 67 Percent of SaaS Companies
What Happened
Maxio’s 2025 pricing trends analysis shows that 67% of SaaS companies now use usage-based or consumption-based pricing, up from about 52% in 2022, with broader industry benchmarks indicating that roughly 85% of companies had either adopted or were testing usage-based models in 2024 and 59% expect consumption-driven revenue to expand in 2025. These benchmarks also show that usage-based and hybrid pricing models outperform flat-rate subscriptions, delivering about a 10% lift in NRR, roughly 22% lower churn, and nearly 2x faster growth. The shift is driven in part by AI-related cost structures—GPU cycles, API calls, and data-processing loads—that cannot be supported by fixed seat-based pricing. Common usage metrics include API calls, storage, transactions, and compute consumption, with companies like AWS, Snowflake, and Twilio illustrating the model through resource-based billing. Emerging AI-agent pricing layers in per-interaction, token-based, and outcome-based fees, while enterprise deployments increasingly use hybrid subscription-plus-usage structures to balance predictability with cost alignment.
Why It Matters
Two thirds of SaaS companies now run consumption models because flat-rate pricing simply breaks under AI cost curves. GPU, inference, and data processing expenses scale with usage, not seats, and vendors that ignore this end up subsidizing their heaviest users. The shift is no longer philosophical; it is survival math. Customers also prefer spend that tracks realized value, which is why consumption models show meaningfully higher NRR and lower churn. This is the closest thing to a settled pricing direction the industry has seen in years.
Implications for You
Metering is now table stakes. If you cannot measure usage in real time, you cannot bill sustainably or defend margin as AI features scale.
Hybrid pricing becomes the default. A base subscription plus usage captures predictable revenue while still monetizing power users who drive real cost.
Procurement friction drops when spend tracks value. Consumption alignment removes renewal battles and shifts attention to product performance, not seat counts.
Infrastructure costs dictate product strategy. Teams must design features with cost visibility from day one or risk negative unit economics as usage grows.
4. Moats & Models
AI Agent Pricing Chaos Emerges with Enterprise Models Ranging from 50 to 500 Dollars per Month
What Happened
Benchmarkit’s 2025 SaaS benchmark analysis of more than 936 companies found median Net Revenue Retention holding at about 101–102% among private B2B SaaS firms, with the 75th percentile at roughly 110%, while public SaaS companies maintained higher medians of about 108–110% and best-in-class performers exceeded 120%. The data confirms strong ARPA-driven divergence: companies with monthly ARPA below $25 posted median NRR of about 94%, with only around 2% reaching 100%+, whereas firms above $500 ARPA achieved about 105% median NRR, with roughly 47% surpassing 100%. Benchmarkit also reports rising acquisition inefficiency, with median New CAC increasing to $2.00 in 2025 (up about 14% from 2023) and blended CAC at about $1.59. Usage-based and hybrid pricing models materially outperformed flat-rate subscriptions, producing roughly 10% higher NRR, about 22% lower churn, and nearly 2x faster growth. Early-stage companies in the $1M–$5M ARR band also showed low-100s NRR with top-quartile results near 110%, typically driven by subscription-plus-usage pricing.
Why It Matters
The 25x spread in agent pricing shows the market still has no shared definition of value, cost, or performance for AI agents. Enterprises don’t trust outputs yet, which is why quality outranks price as the top adoption barrier. Vendors are improvising with seat, usage, and outcome-based pricing in parallel, creating confusion that slows enterprise rollouts and inflates integration costs. This volatility is a signal that the real bottleneck isn’t the model but the operational plumbing around it: data pipelines, attribution, workflow reliability, and governance.
Implications for You
Pricing will consolidate around measurable outcomes. Vendors that cannot prove attributable savings or throughput gains will get pushed into low-margin seat pricing.
Integration, not inference, drives deal value. The 50–200k deployment cost shows buyers pay most for data prep and workflow wiring, not model access.
Attribution becomes a moat. If you can quantify the agent’s impact, you can charge for it; if you cannot, procurement will cap spend fast.
Performance, not price, wins renewals. Enterprises will pay premium rates for agents that work reliably every time, and churn instantly when outputs are inconsistent.
Other Signals on our Radar:
Ramp Data Shows Google’s Breakout and ServiceNow’s Spend Momentum in Enterprise SaaS
Ramp’s December 2025 rankings, drawn from more than forty-five thousand customers, show major shifts in SaaS adoption. Replit led in new customer count, followed by OpenAI, Canva, Intuit, and Anthropic. By new spend, Google ranked first, ahead of ServiceNow, Gong, Bird, and Applied Intuition. Replit, Lucid, Atlassian, Clerk, and Google posted the fastest growth in new customers, while ServiceNow, Google, Fusionware, Datadog, and Gong led in new spend growth. Google’s surge follows its late-November rollout of Gemini 3 and the Nano Banana Pro image model and its earlier decision to make pro AI features free across Workspace, which masks actual adoption because in-product usage is not captured by external spend data. Anthropic gained two point one percentage points of adoption in October, while OpenAI gained only zero point three. Consumer app download data from the Financial Times shows Gemini gaining ground on ChatGPT, signaling Google’s recovery as a leading force in both enterprise and consumer AI.
The Stack is a weekly intelligence brief for leaders building what’s next in AI and software. We deliver high-impact developments shaping the U.S. market: what happened, why it matters, and what to do about it. Each issue distills complex shifts into decision-grade insight.
Thanks for reading B2B SaaS Executive Intelligence! Subscribe for free to receive new posts and support our work.
About The Intelligence Council
The Intelligence Council publishes sharp, judgment-forward intelligence for decision-makers in complex industries. Our weekly briefs, monthly deep dives, and quarterly sentiment indexes are built to help you grow your top-line and bottom-line, manage risk, and gain a competitive edge. No puff pieces. No b.s. Just the clearest signal in a noisy, complex world.