End of Season Sale in Data Land
What $10B in M&A Tells Us About the Future of AI
We have seen 4 major acquisitions in 4 weeks as major players went shopping for strategic leverage in the AI stack. And like most things in AI, all roads led back to data.
2 June 2025, Snowflake buys Crunchy Data for $250m
27 May 2025, Salesforce buys Informatica for $8b
14 May 2025, Databricks buys Neon for $1b
7 May 2025, ServiceNow buys Data.World
Let’s unpack who’s buying what - and why.
🛠 Two Postgres Deals, One Strategic Shift
Databricks + Neon | Snowflake + Crunchy
Two of the biggest names in data just acquired two Postgres startups within weeks of each other. At a glance, it looks like both are just adding transactional muscle to their platforms. But look closer, and you’ll see two distinct philosophies - converging on the same reality: AI agents have changed the nature of what data infrastructure needs to be.
And Postgres - mature, pluggable, ubiquitous - is the bridge both are choosing to cross that gap.
💡 The Shared Insight
Both Snowflake and Databricks are reacting to the same signal: The world of dashboards and batch jobs is giving way to autonomous, real-time, read/write systems. Agents don’t just ask questions. They act. They create state. They need memory.
Both companies were built for different generations of data problems:
Snowflake for clean, centralized analytics - elegant separation of compute and storage, zero operational chaos.
Databricks for massive, messy data science - collaborative notebooks, lakehouses, batch-driven ML training.
Now, they’re both moving into operational, agent-compatible infrastructure - and Postgres, long the Swiss Army knife of databases, is suddenly strategic again.
🔍 But Their Moves Reveal Different Playbooks
Neon is Postgres reimagined for AI-native development. It’s ephemeral, serverless, API-driven. Databases spin up instantly - often by agents, not humans (Neon says 80% of theirs are agent-created).
This is agent-grade infra, meant for micro-apps, short-lived memory, autonomous orchestration.
By buying Neon, Databricks is chasing the future of software creation - where agents scaffold apps on the fly, not just humans writing code.
Crunchy is Postgres fortified for real-world enterprise deployment. It’s stable, secure, and hardened for compliance-heavy environments. Crunchy has traction in defense, healthcare, and government - sectors where write access must come with guardrails, not just speed. It gives Snowflake a foothold in hybrid OLAP/OLTP use cases, operational agent memory, and the credibility to serve customers who must audit every field.
By buying Crunchy, Snowflake is acknowledging that enterprise AI won’t just live in clean schemas. It will need to interact with messy, live systems - and those systems already run on Postgres.
⚖️ What This Tells Us About Their Worldviews
These deals are a sign of an underlying trend where AI shifts the locus of value from “insight” to “interaction.”
Batch gives way to real-time.
Read-only gives way to read/write.
Centralization gives way to composability.
And in that world, the players who win will be the ones who can power not just analytics, but action.
🧠 Two Governance Deals, One Clear Power Play
Salesforce + Informatica | ServiceNow + data.world
In the application layer of the enterprise stack, two giants just made big moves that may seem unglamorous at first glance - metadata, master data, governance.
But under the hood, they’re making a power grab for the most important asset AI agents lack today: context. Because AI agents are getting smarter - but not more aware.
These are bets on semantic control - knowing what your data means, where it came from, and who can trust it to make decisions autonomously.
💡 The Shared Insight
The model layer is getting commoditized. The real chokepoint is data alignment: across systems, teams, workflows, and formats. AI agents can retrieve information just fine. What they can’t do yet is answer questions like:
“Is this the same customer as that one?”
“Which invoice matters right now?”
“What is the correct price, region, owner, or SLA?”
That’s not a model problem. That’s a metadata, lineage, and governance problem. And it’s exactly what these acquisitions solve.
Salesforce and ServiceNow aren’t building the smartest copilots. They’re building the safest and most reliable execution environments, because in enterprise AI, hallucination = lawsuit.
🔍 But Again, The Philosophies Diverge
Informatica is the old guard of enterprise data plumbing - and that’s exactly the point. It’s the MDM, governance, privacy, and lineage layer trusted by Fortune 500s - and it gives Salesforce the backbone to support AI agents that can execute real business actions, not just summarize call notes.
Agentforce (Salesforce’s AI platform) only works if agents can trust what they’re acting on. Informatica ensures that trust. This is about making Salesforce the source of truth across the org, not just the front-end for it.
Data.World isn’t about plumbing - it’s about perspective. It uses knowledge graphs and semantic models to overlay meaning across data and workflows. Not just what the data says but how it relates. So when an AI agent is managing an onboarding request or resolving an IT incident, it understands not just the task - but the why behind it.
With this, ServiceNow’s Workflow Data Fabric becomes more than a queue. It becomes a knowledge graph by giving agents a mental model of the org.
Salesforce and ServiceNow are building semantic supervisors - systems of record for what matters, what’s allowed, and what connects to what.
Informatica makes Salesforce the enforcer of trust.
data.world makes ServiceNow the interpreter of meaning.
Together, they show us where enterprise AI is really headed: Not just toward smarter software, but toward context-aware systems that act with intention, not just intelligence.