Databricks' $7B Round: AI Infrastructure's $134B Watershed Moment

So, What Actually Happened?

So I was digging through 190,000 articles this week and here's what kept popping up: the data infrastructure war just got its most expensive week yet. Databricks closed a $7 billion raise at a $134 billion valuation, while Nebius dropped $275 million to acquire Tavily for agentic search capabilities. Meanwhile, Cisco quietly unveiled a new networking chip purpose-built for AI data centers — because all that AI magic needs actual plumbing. And while the money flows upward, regulators are moving laterally: California's CCPA regulations are finally crashing to shore, the EU's Digital Omnibus package is rewriting GDPR for the AI age, and your CISO is wondering how to stop employees from feeding your crown jewels into unauthorized AI tools.

The Bottom Line: The infrastructure is being built, the money is flowing, and the regulation is catching up — all at the same time. If you're not paying attention to all three simultaneously, you're only seeing a third of the picture.

The Tracks That Matter

1. Databricks Hits $134 Billion: The Data Platform That Ate the World

Remember when Databricks was ”just” a Spark company? Those days are so far gone they need a time machine. The company just closed a $7 billion funding round at a $134 billion valuation — making it the third-most valuable private tech company in the world. But the valuation isn't the story. The story is the revenue: Databricks surpassed a $5.4 billion run rate, growing more than 65% year-over-year.

What's driving it? AI products are the growth engine. The company's pivot from ”analytics platform” to ”AI and data intelligence platform” isn't just marketing — it's showing up in revenue. Toyota Motor, JPMorgan Chase, and Goldman Sachs Alternatives all participated in the round. When banks invest in your data platform, they're not speculating — they're buying infrastructure.

Here's what I notice as a DJ who's watched music formats come and go: Databricks did what Spotify did to music. They didn't just build a better record player — they changed how the entire industry thinks about distribution. Your data lake isn't storage anymore. It's the stage.

Here's what works: If you're still running a multi-vendor data stack with separate compute, storage, and AI layers, this is your pricing signal. Consolidation is happening, and the platforms with the largest ecosystems will set the terms. Evaluate your stack before the market evaluates it for you.

2. Nebius Buys Tavily for $275M: The Agentic Search Land Grab

Most people haven't heard of Nebius. Even fewer have heard of Tavily. But this $275 million acquisition tells you exactly where the AI infrastructure race is heading: full-stack AI clouds need search built in, not bolted on.

Nebius — the AI cloud platform spun out of Yandex — is adding Tavily's agentic search capabilities to its platform. Tavily built an API purpose-designed for AI agents to search the web, retrieve data, and synthesize information. The bet: as AI agents become the primary interface for enterprise work, they need a search layer that speaks ”agent” — not ”human typing into a search bar.”

This is the infrastructure equivalent of what happened in DJing when CDJs replaced vinyl. The old interface (search box, ten blue links) doesn't work for the new user (autonomous agents making thousands of queries per minute). Someone has to build the new interface. Nebius just bought it.

Here's what works: If you're building AI agent workflows, evaluate whether your search infrastructure is designed for agent consumption or human consumption. The latency requirements, output formats, and query patterns are fundamentally different. Tavily was built agent-first — most enterprise search wasn't.

3. Cisco's Silicon One G300: The Chip No One's Talking About That Everything Depends On

While the headlines chase GPU wars, Cisco quietly launched the Silicon One G300 — a networking chip specifically designed to handle the traffic patterns of AI data centers. The chip powers new systems and optics for what Cisco calls the ”Agentic AI era”, where the real bottleneck isn't compute — it's moving data between compute nodes fast enough.

At the Cisco AI Summit, the company positioned 2026 as the year AI moves from breakthrough to deployment — and networking is the infrastructure that makes deployment possible. Think about it: you can have all the GPUs in the world, but if the network connecting them can't handle the traffic, you've built a sports car with bicycle wheels.

This is a foundational technology play. The G300 appeared across multiple articles this week but made zero mainstream headlines. That's how you spot real infrastructure: engineers are using it, marketing hasn't caught up yet.

Here's what works: When evaluating AI infrastructure vendors, don't just look at compute. Ask about network architecture. The companies that solve the data movement problem — not just the data processing problem — will define the next generation of AI deployment.

4. The CCPA Wave Finally Hits: California's Privacy Enforcement Gets Real

For two years, everyone said ”the CCPA regulations are coming.” Well, they're here. A long-delayed wave of CCPA regulations is finally crashing to shore, and the California Attorney General isn't waiting for companies to catch up. Even more pointedly, the AG just announced an investigative sweep around surveillance pricing practices — using AI to personalize pricing based on individual data profiles.

Meanwhile across the Atlantic, the EU's Digital Omnibus package is rewriting the rules too. The package proposes to raise the breach notification threshold from ”risk” to ”high risk,” clarify the definition of personal data for AI processing, and introduce EU-level AI regulatory sandboxes. It's expected to be finalised during 2026.

What nobody's connecting: the US and EU are converging on the same enforcement patterns — AI-powered personalization and automated decision-making — but from opposite regulatory traditions. California leads with enforcement actions; the EU leads with framework legislation. Both are arriving at the same destination.

Here's what works: Audit your AI-driven pricing, personalization, and automated decision-making systems now. California's surveillance pricing sweep is the canary. If you're using AI to set individual prices based on personal data, you need legal review this quarter — not next year.

5. Shadow AI Gets Its Own Security Stack: Reco Raises $30M

Here's a number that should wake up every CISO: Reco reports 500% growth between 2023 and 2024, and an additional 400% growth in 2025. The Israeli startup just raised $30 million in a Series B to secure AI SaaS adoption — because employees are adopting AI tools faster than security teams can monitor them.

And Reco isn't alone. CIO just published a comprehensive wakeup call on Shadow AI practices, noting that AI agents are ”quietly rewriting risk profiles faster than policies can keep up.” The article is blunt: enterprises that don't get visibility into unauthorized AI usage are sitting on data breaches they don't know about yet.

Reco's platform monitors everything from AI-powered apps to autonomous agents to Model Context Protocols (MCPs). That last one tells you where the market is heading: the security layer has to understand not just which AI tools employees use, but how those tools connect to your data.

Here's what works: Don't wait for a breach to discover your Shadow AI problem. Start with an audit: which AI tools are employees using that IT didn't provision? What data are they feeding into those tools? Reco and competitors like WitnessAI exist because the answer to both questions is usually ”way more than you think.”

6. Bretton AI Raises $75M: Financial Crime Automation Goes Mainstream

While most AI funding goes to flashy chatbots and image generators, Bretton AI quietly raised $75 million in a Series B led by Sapphire Ventures to automate financial crime compliance. This is the kind of AI application that doesn't make TechCrunch headlines but saves banks millions in compliance costs.

Financial crime compliance is one of those industries where the humans can't keep up with the criminals. Anti-money laundering teams at major banks process thousands of alerts daily, most of them false positives. The cognitive load is crushing, the regulatory penalties for missing real threats are massive, and the talent shortage is getting worse. Bretton AI automates the triage, investigation, and reporting — the boring, essential work that nobody wants to do.

This is vertical AI at its finest. Like I always say: the DJ who masters one genre gets booked every weekend. The one who plays everything gets booked once.

Here's what works: If you're in financial services, evaluate your AML and KYC workflows for AI automation potential. The ROI on reducing false positive investigation time alone can justify the investment. But start with the compliance team, not IT — they know where the pain is.

7. Chile Launches Latam-GPT: AI Sovereignty Goes South

Something happened this week that got zero Silicon Valley attention: Chile launched Latam-GPT, an open-source AI model designed specifically for Latin America. It's not competing with GPT-5 on benchmarks. It's competing on something more fundamental: linguistic and cultural representation for 650 million people.

Most LLMs are trained predominantly on English text and reflect Anglo-American cultural norms. Latam-GPT is trained on Spanish and Portuguese language data that captures regional dialects, legal systems, and cultural context. For a government trying to deploy AI for citizen services in Bogota or a hospital using AI for patient intake in Sao Paulo, a model that actually understands Latin American Spanish isn't a nice-to-have — it's the difference between useful and useless.

This is part of a broader AI sovereignty trend emerging across the globe. Korea's Trillion Labs is building next-gen AI to challenge US dominance. The EU is funding its own models. The Global South is saying: we don't just want to consume AI, we want to build it.

Here's what works: If you operate in Latin American markets, watch Latam-GPT closely. Open-source, culturally-aligned models may outperform general-purpose LLMs for region-specific use cases — especially anything involving legal text, customer service, or government communications. The best model isn't always the biggest.

Signal vs. Noise

🟢 Signal: Vertical AI specialization is winning the funding wars. Bretton AI ($75M for financial crime), Reco ($30M for AI SaaS security), and Databricks' AI-product-driven revenue growth all point to the same thing: investors are rewarding companies that picked a lane and mastered it. The era of ”general-purpose AI startup” raising on vibes is ending.

🟢 Signal: AI sovereignty is a real geopolitical force. Chile's Latam-GPT and Korea's Trillion Labs aren't footnotes — they're signals that the next phase of AI development won't be US-only. Enterprise leaders need geographic AI strategy, not just vendor strategy.

🔴 Noise: ”AI will destroy all software” narratives. Fortune reported this week that ”AI agents aren't eating SaaS software, they're using it.” The pendulum swung too far toward panic. SaaS isn't dying — it's being augmented. The incumbents can't sleep easy, but they're not dead either. Investors swinging between ”all good” and ”all bad” are creating noise, not insight.

🔴 Noise: Sam Altman mention growth (+433%) with declining influence (-2.7% PageRank). More noise about the person, less substance about the technology. When mentions spike but influence drops, it's PR, not progress.

From the 190K

We scanned 190,000 articles this week. Here's what no one's talking about:

The Regulatory Compliance Convergence

Something strange is happening across four completely different domains. Regulatory compliance showed up as a bridge concept connecting Third-Party Risk Management, Fundraising, Data Quality, and Data Privacy/Cybersecurity — four areas that used to be managed by four different teams with four different budgets.

The CCPA regulations are hitting simultaneously with the EU's Digital Omnibus rewrite, while Reco's $30M raise for AI SaaS security and Bretton AI's $75M for financial crime compliance tell the same story from the vendor side. Compliance isn't a cost center anymore — it's converging into the core operational stack.

The pattern that only emerges at 190,000-article scale: every enterprise function is becoming a compliance function. Your marketing team's AI personalization? CCPA territory. Your data team's model training? GDPR scope. Your finance team's AML workflows? Bretton AI territory. The organizations that see compliance as a horizontal layer — not a departmental silo — will spend less and sleep better.

🔍 Below the surface: Data Integration appeared in 97 articles this week but made zero headlines. Here's how you spot real infrastructure: when something shows up everywhere but headlines nowhere, it means engineers are building on it and marketing hasn't caught up. Data Integration's Katz centrality (foundational importance) remains high even as its PageRank (trending visibility) declines. The plumbing is invisible — until it breaks.

By The Numbers

• $134 billion — Databricks' new valuation after its $7B round, making it the third-most valuable private tech company globally
• $275 million — what Nebius paid for Tavily, betting that agentic search is the next infrastructure layer
• 500% + 400% — Reco's growth rate across 2023-2024 and 2025 respectively, proving Shadow AI security isn't a niche
• 88% — the AI project failure rate that enterprises are still ignoring when planning strategy
• 97 articles — how often Data Integration appeared this week without making a single headline
• $75 million — Bretton AI's Series B for financial crime automation — the AI that's boring but essential
• 65%+ YoY — Databricks' revenue growth rate, driven primarily by AI product adoption
• 750 million — Gemini's monthly active users, up from 400 million nine months ago

Deep Dive: The Data Readiness Paradox

There's an article that landed this week with a title that should be on every CDO's wall: ”Why Companies Spend Millions Cleaning Data That Doesn't Matter.”

I've seen this movie before. I once worked with a client who spent eight months and €1.2 million cleaning their entire data estate to ”get ready for AI.” When we finally looked at what the AI actually needed, it was three tables. Three. The rest was expensive busywork.

The Paradox Itself

The data readiness paradox works like this: organizations know they need clean data for AI. So they launch massive data quality initiatives across the entire estate. But AI use cases are specific — they don't need perfect data everywhere, they need relevant data somewhere. The result? Companies spend millions getting data ”AI-ready” that no AI model will ever touch.

Why It Keeps Happening

Two forces collide. Data teams think in platforms (”let's clean everything”). AI teams think in models (”we need these specific features”). Without a bridge between the two, data quality becomes a cost center instead of an enabler. It's like a DJ who re-organizes their entire record collection alphabetically when they really just need to find five tracks for tonight's set.

The Revenue Guardian Flip

The flip side is emerging too. Strategic data management is moving from cost center to revenue guardian. Organizations that tie data quality to specific business outcomes — rather than abstract ”readiness” scores — are seeing data teams contribute directly to revenue.

What Actually Works

1. Start with the use case, not the data: Identify your top 3 AI use cases, then clean only the data those use cases need
2. Measure data quality by business impact: ”99% accuracy” means nothing if the 1% error is in your revenue calculations
3. Kill the boil-the-ocean projects: Incremental data quality tied to specific deployments beats comprehensive cleaning that never finishes
4. Make data teams accountable to revenue: When data quality is measured by ”AI models deployed” instead of ”fields cleaned,” priorities shift fast

The festival doesn't need every instrument tuned to perfection. It needs the right instruments tuned for tonight's setlist.

What's Coming

The Agentic Infrastructure Build-Out Accelerates

Nebius-Tavily ($275M) and Salesforce's acquisition of Cimulate are early signals. Expect more acquisitions targeting the infrastructure layer that AI agents need: search, orchestration, security, and monitoring. The ”agentic” thesis is moving from slide decks to M&A strategy.

Privacy Enforcement Enters AI Territory

California's surveillance pricing investigation and the EU Digital Omnibus aren't isolated events. AI-driven personalization is becoming a privacy battleground. Companies using AI to set individual prices, target ads, or make automated decisions about people should expect regulatory attention in 2026 — not 2027.

AI Sovereignty Becomes a Budget Line

Chile's Latam-GPT, Korea's Trillion Labs, and the EU's investment in regional models are creating a new category of AI infrastructure spend. Multinationals will need to evaluate whether US-built models meet regulatory, linguistic, and cultural requirements in every market they serve. ”One model fits all” is ending.

For Your Team

Thursday's meeting prompt: ”We're cleaning data across our entire estate to 'get ready for AI.' But what if we're spending 80% of our data quality budget on data no AI model will ever use? How do we identify the 20% that actually matters?”

The Data-AI Alignment Framework:

1. Map use cases first — Document the top 5 AI use cases your business actually plans to deploy in the next 12 months
2. Trace the data lineage — For each use case, identify exactly which data sources, tables, and fields are required
3. Score quality by impact — Assign data quality priority based on business value of the dependent use case, not abstract completeness
4. Kill the zombie projects — Any data cleaning initiative that can't be linked to a deployed AI use case within 6 months gets paused
5. Measure in revenue, not records — Track ”AI models deployed” and ”revenue influenced by clean data,” not ”rows cleaned”

Share-worthy stat: Databricks is now worth $134 billion on a $5.4 billion revenue run rate — a 25x revenue multiple that tells you exactly how the market values AI-ready data infrastructure.

Go deeper: Track AI infrastructure trends in real-time →

The Track of the Day

”AI agents aren't eating SaaS software, they're using it.”
— Fortune, Feb 10, 2026

The hype cycle loves binary narratives: AI kills SaaS, AI kills jobs, AI kills everything. Reality is always more nuanced — and more interesting. The real story isn't replacement. It's integration. The data platforms that survive are the ones that become the stage, not the act.

We scanned 190,000 articles this week so you don't have to. Data Pains → Business Gains.

Published: February 11, 2026 | Curated by Yves Mulkers @ Ins7ghts