Oracle Cuts 30,000 Jobs to Fund $Billions in AI Infrastructure

Your weekly signal boost from 190,000+ articles, served with a DJ's ear for what actually matters.

So, What Actually Happened?

We scanned 190,000 articles this week so you don't have to. And the signal that punched through everything else? Oracle sent 30,000 employees a 6AM email telling them their jobs no longer exist, freeing up billions for AI data centers. Meanwhile, pharmacy data is quietly consolidating under private equity ownership in ways that should concern every healthcare CIO. On the bright side, nCino posted a 17% jump in annual contract value, proving that banking AI works when it is the system of record, not a feature bolted on top. And in a story nobody headlined, seven identity and API security tools launched specifically for the AI agent era, because the security layer for autonomous agents simply does not exist yet.

The Bottom Line: The AI investment thesis split into two camps this week: companies burning headcount to fund infrastructure, and companies proving AI works when it is embedded in the workflow. One camp is making headlines. The other is making money.

The Tracks That Matter

1. Oracle Just Sent 30,000 People a 6AM Email. The AI Pivot Has a Human Price Tag Now.

Oracle laid off approximately 30,000 employees, roughly 19% of its global workforce, to redirect billions of dollars toward AI data center investments. The layoffs arrived via a 6AM email. No warning. No phased transition. A senior operations manager, Michael Shepherd, posted that ”the individuals affected were not let go because of anything they did or didn't do.” That sentence tells you everything about the current state of enterprise AI strategy: the people are fine, the strategy just changed, and the people are collateral.

This is no longer the quiet reshuffling that companies like to frame as ”efficiency optimization.” Oracle has annual revenue of $55.4 billion. They are not struggling. They are making a conscious, massive capital reallocation decision: fewer humans, more GPU clusters. And they are doing it at a scale that makes the previous rounds of tech layoffs look like rounding errors.

The structural pattern is what matters here. Last week we covered CEOs dropping euphemisms and openly naming AI as the reason for layoffs. Oracle just turned that trend into a spreadsheet. Thirty thousand positions eliminated. Billions redirected to infrastructure. The conversion ratio between human labor and compute investment is becoming explicit, and Oracle just published theirs for the world to see.

Here's what works: If you work at a company with a large traditional workforce and an aggressive AI roadmap, map the gap. How much of the current payroll funds roles that AI infrastructure could replace? That is not a theoretical exercise anymore. It is Oracle's actual financial strategy. If your leadership team has not done this math openly, someone is doing it quietly. Ask the question before the 6AM email answers it for you.

2. The Quiet Consolidation of Pharmacy Data Should Worry Every Healthcare CIO.

Private equity firms are systematically acquiring pharmacy data companies, and the implications reach far beyond the pharmacy counter. Sycamore Partners acquired Shields Health Solutions, a specialty pharmacy platform with deep clinical and operational data. New players like Veridex Lab are entering the space with data analytics offerings aimed at independent pharmacies. The pattern is familiar to anyone who watched the electronic health records consolidation of the early 2010s: whoever owns the data infrastructure ends up controlling the information flows.

Pharmacy data is uniquely sensitive. It sits at the intersection of HIPAA, GDPR (for global operations), and state-level regulations. It contains prescribing patterns, patient adherence data, insurance billing information, and increasingly, genomic and biomarker data linked to precision medicine. When private equity consolidates this data under a few platforms, the regulatory surface area multiplies while the accountability structure narrows. Independent pharmacies lose visibility into how their patient data gets aggregated and resold.

This story connects directly to the acquisition pattern we tracked last week. Five data infrastructure acquisitions closed in 48 hours across unrelated industries. Pharmacy data is following the exact same playbook: AI models are commoditizing, so the value concentrates in whoever owns the domain-specific data. Private equity understood this before most healthcare CIOs did. The question is not whether pharmacy data consolidation is happening. It is whether your organization's data is part of someone else's asset portfolio.

Here's what works: If your organization touches pharmacy data (and if you are in healthcare, insurance, or benefits, you almost certainly do), audit your data agreements. Specifically: who owns the aggregated insights derived from your operational data? If the answer involves a private equity-backed intermediary, understand what they can and cannot do with that data. The consolidation is already underway. Your leverage is highest before the contract renewal, not after.

3. A Banking AI Platform Just Posted 17 Percent Growth While Everyone Else Is Still Running Pilots.

nCino reported $149.7 million in Q4 revenue and $594.8 million for fiscal year 2026, reflecting year-over-year growth of 6% and 10% respectively. The real number is the one investors care about: annual contract value grew 17%, with a net retention rate of 112%. CEO Sean Desmond put it bluntly: ”AI is moving quickly from help me write and help me search to help me complete meaningful productive tasks.”

What makes nCino structurally different from the flood of AI vendors claiming enterprise traction is the system-of-record position. They are not a feature. They are, as Desmond described it, ”the system of record and user experience for many of the most important processes in a financial institution.” More than 170 banking customers, from global enterprises to community credit unions, run lending, onboarding, account opening, and portfolio management through nCino. When you are the system of record, AI is not a bolt-on. It is a multiplier.

This matters because of the contrast. We reported last week that 83% of AI pilots fail. nCino's 17% ACV growth says the other 17% that succeed share something in common: the AI is embedded in the workflow, not sitting in a separate application that people have to remember to open. The same lesson emerged from the Tempus/Medtronic clinical trial results earlier this week. The pattern is consistent: embedded AI works. Bolted-on AI becomes shelfware.

Here's what works: Ask your team the nCino question: ”Is our AI embedded in the system of record, or is it a separate tool that requires users to change their workflow?” If users have to leave their primary workspace to interact with AI, adoption will plateau. The companies posting real growth numbers are the ones where AI lives inside the workflow users already have. Embed or stall.

4. Seven Security Tools Just Launched for the AI Agent Era. Most Companies Have Zero of Them.

Security Boulevard published an evaluation of seven identity and API security tools designed specifically for the emerging AI agent landscape, and the gap between what agents can do and what security infrastructure can monitor is alarming. Among the launches: MojoAuth is pushing passwordless authentication that eliminates database breach risk entirely. Gopher Security is shipping a unified MCP gateway with zero-trust networking and post-quantum cryptography that provides deep inspection of tool calls made by AI agents.

The timing is not accidental. Fidelity's institutional analysis last week described AI agents that operate persistently across time, tools, and environments. Those agents need to authenticate, access APIs, and execute workflows. But the security infrastructure for that world barely exists. Most enterprise security stacks were designed for human users clicking through interfaces, not autonomous agents making API calls at machine speed.

The post-quantum cryptography element deserves specific attention. Gopher Security's decision to ship post-quantum crypto in a tool gateway (not just in transit encryption) signals that the security vendors closest to AI infrastructure are already planning for a world where current encryption standards are breakable. If your AI agents are handling sensitive data today, the encryption protecting their API calls has a shelf life. The vendors building the next security layer are not waiting for standards bodies to finalize. They are shipping now.

Here's what works: Inventory every API call your AI agents make. Not the ones in your architecture diagram. The actual calls, in production, right now. How many of them go through authenticated, monitored, rate-limited endpoints? If the honest answer is ”we don't know,” you have an agent security gap. These seven tools exist because the market knows the gap is real. Close it before your agents become your largest attack surface.

5. Physical AI Just Got a Platform. Wiliot Wants Every Product on the Planet to Think.

Wiliot and Databricks announced a partnership to run Wiliot's Physical AI platform on the Databricks infrastructure, enabling supply chain automation at enterprise scale. Adi Applebaum, Wiliot's VP of Product, stated that ”running the Wiliot Physical AI platform on Databricks gives our customers the ability to operationalise physical-world data at scale.” Roberto Robles from Databricks went further: ”Physical AI represents the next frontier of data intelligence.”

Most AI conversations happen in the digital world. Text, images, code, spreadsheets. Wiliot operates in the physical world: tiny IoT sensors that attach to products and generate data about location, temperature, handling, and condition throughout the supply chain. The partnership with Databricks means that physical-world data (where is this product, what happened to it, when did it arrive) can now feed into the same AI infrastructure that processes digital data. The bridge between the digital twin and the physical object just got a major platform upgrade.

The business implication is in the phrase ”operationalise physical-world data.” Today, most supply chain AI works on historical data: what happened last quarter, what the forecast predicts. Wiliot's sensor layer generates real-time data about individual items. Combined with Databricks' processing and AI capabilities, this means predictive actions (reroute this shipment, flag this temperature excursion, reorder this product) based on what is happening now, not what happened before.

Here's what works: If your business involves physical products (manufacturing, logistics, retail, healthcare), ask whether your AI strategy includes real-time physical-world data or just historical digital records. The companies that close the gap between the physical item and its digital representation first will have a structural advantage in supply chain optimization. Wiliot's bet is that the sensor layer is the missing piece. If your supply chain AI runs on batch data, you are optimizing for yesterday.

Signal vs. Noise

🟢 Signal: Banking AI is posting real growth numbers while the rest of the market debates adoption frameworks. nCino's 17% ACV growth and 112% net retention across 170+ financial institutions is the clearest proof point this week that embedded AI (AI inside the system of record, not beside it) converts into revenue growth. The pattern echoes across healthcare (Tempus/Medtronic) and supply chain (Wiliot/Databricks). When the conversation shifts from ”should we adopt AI?” to ”our AI platform just posted double-digit growth,” the early movers are pulling away.

🟢 Signal: Security vendors are building for the post-agent, post-quantum world before enterprises know they need it. Seven identity and API security tools launched with features designed for autonomous AI agents, including post-quantum cryptography for tool gateways. The security industry is building for a threat model most enterprises have not even mapped yet. When the tooling arrives before the demand, it means the builders see something the buyers have not.

🔴 Noise: The ”AI will take your job” narrative keeps recycling the same fear without new evidence. Oracle's 30,000 layoffs are real and consequential, but the media response is the same template that has run after every tech layoff since 2022. The noise is not that layoffs are happening (they are). The noise is that the coverage treats each round as if it were the first, ignoring that Oracle's reallocation is a capital strategy, not a robot apocalypse. The story that matters is where the freed capital goes, not the fear cycle it generates.

From the 190K

The System of Record Is Quietly Becoming the AI Moat. No One Wrote That Headline.

We scanned 190,000 articles this week. Here is what only emerges at scale:

Three completely unrelated stories, in three different industries, published the same structural insight within 48 hours. nCino posted 17% ACV growth in banking because it is the system of record. Pharmacy data is consolidating under private equity because whoever owns the data system of record controls the AI training data. Wiliot partnered with Databricks to become the system of record for physical-world product data. The banking press covered nCino. The healthcare press covered pharmacy consolidation. The IoT press covered Wiliot. Nobody connected them.

The pattern: the application layer is commoditizing (exactly as Fidelity's institutional analysis predicted last week), but the system-of-record layer is where value concentrates. AI models can be swapped. Data platforms cannot. The companies embedding themselves as the default data backbone for their industry are building moats that no foundation model can replicate. The acquisitions, the partnerships, and the earnings reports all point the same direction: own the record, own the future.

🔍 Below the surface: AI is showing up in packaging, police body cameras, and mortgage loan origination systems this week. Zero headlines for any of them. Here is how you spot real adoption: when AI stops being the story and becomes a feature buried in the fourth paragraph of an industry trade publication. The hype machine cannot make packaging AI sexy, which usually means it actually works.

By The Numbers

• 30,000 jobs: Oracle employees laid off in a single round, approximately 19% of its global workforce, to fund AI data center investments.
• $594.8 million: nCino's fiscal year 2026 revenue, with annual contract value growing 17% year-over-year and net retention hitting 112%.
• 170+ customers: Financial institutions running on nCino's banking AI platform, from global enterprises to community credit unions.
• 7 security tools: Launched specifically for the AI agent era, including passwordless authentication and post-quantum cryptography for tool gateways.
• 3 GDPR references: In a single day's articles, with HIPAA at 2 and CCPA at 1. Regulatory density continues compounding across every sector AI touches.
• $55.4 billion: Oracle's annual revenue. They are not cutting because they are struggling. They are cutting because they are reallocating.
• 1,200+ restaurants: GiftAMeal's partner network across 46 states, acquired by Swipe Savvy. Guests using the platform visit 39% more often and spend 20% more per order.

Deep Dive: The System of Record Is the Only Moat That Survives the AI Flood

You know the difference between a DJ who owns the master recordings and one who just has a playlist? When Spotify changes its algorithm tomorrow, the playlist DJ is scrambling. The one with the masters still gets paid. That distinction is playing out across every industry right now, and this week's data made the pattern impossible to ignore.

The Proof Points Are Piling Up

nCino did not grow 17% because it has better AI than its competitors. It grew because it is the system of record for banking workflows. When 170 financial institutions run their lending, onboarding, and portfolio management through your platform, adding AI features is a retention play, not a science project. The data is already there. The workflows are already there. The users are already there. AI becomes the upgrade, not the product. That is why their net retention hit 112%: existing customers are spending more, not because nCino's AI is magical, but because it is embedded in the place they already work.

The Consolidation Play

Private equity understood this before most technology executives did. The pharmacy data consolidation story is not about pharmacy. It is about the template. Acquire the system of record for a regulated industry. Layer analytics on top. Use the proprietary dataset to train AI models that no competitor can replicate because no competitor has the data. Then sell insights back to the industry that generated them. It is the data flywheel that every AI pitch deck promises, except the private equity firms are building it through acquisition instead of organic growth. Faster, messier, and already happening in healthcare, real estate lending, and now pharmacy.

Why Application Layer Companies Should Be Worried

Fidelity's institutional analysis last week argued that AI agents will erode the value of applications as user interfaces. This week's evidence supports that thesis from a different angle. The companies posting growth (nCino, Wiliot) are infrastructure companies, not application companies. They do not compete on interface design. They compete on being the data layer that everything else depends on. When AI agents can navigate any interface (Fidelity's prediction), the interface becomes a commodity. But the data underneath? The system of record that captures outcomes, stores history, and trains the next generation of models? That is the moat. That is the master recording.

What Actually Works

1. Audit your system-of-record dependencies. For every critical workflow, identify which platform holds the authoritative data. If that platform belongs to a vendor, understand what they can do with your data. You are either building the moat or filling someone else's.
2. Prioritize data capture over AI features. The companies winning right now are not the ones with the best models. They are the ones with the most comprehensive outcome data. Every interaction, every decision, every result: captured and structured for future AI training.
3. Watch the private equity playbook. PE firms are buying domain-specific data companies because the math works: proprietary datasets in regulated industries have compounding value. If your industry has fragmented data ownership, consolidation is coming. Position accordingly.
4. Stop evaluating AI tools. Start evaluating data moats. The question is not ”which AI vendor has the best model?” Models are commoditizing quarterly. The question is ”which vendor captures and retains the data that makes future models better?” That is the defensible position.

When I started organizing my vinyl collection, I built the database myself. Not because I was smarter than the record shops. Because I knew that whoever controls the catalogue controls the collection. Every rare record I logged, every BPM I tagged, every cross-reference I built became an asset that no one else could replicate. The companies that understand this about their data will own the next decade. Everyone else will be renting.

What's Coming

AI Agent Security Standards Will Emerge From Financial Services First

The combination of seven new security tools for the AI agent era and nCino's banking AI growth points to a near-term collision: financial institutions deploying AI agents at scale will need security standards that do not exist yet. Banking regulators move slowly, but they move first. Expect draft frameworks for AI agent authentication, monitoring, and audit trails from at least one major financial regulator before Q4 2026.

The Pharmacy Data Consolidation Will Accelerate After AI Drug Discovery Deals

Private equity's pharmacy data play will accelerate as AI drug discovery matures. Last week's Eli Lilly $2.75 billion bet on AI-powered drug discovery increases the value of real-world prescribing data. When drug companies need real-world evidence to validate AI-discovered compounds, the companies that own structured pharmacy data become essential partners. Watch for at least two more pharmacy data acquisitions in Q2 2026.

Enterprise AI Budgets Will Shift From Models to Data Infrastructure

Oracle's decision to eliminate 30,000 positions to fund AI data centers is a leading indicator, not an outlier. Data challenges continue to persist as companies rush AI adoption, and the gap between AI ambition and data readiness is widening. Expect Q2 and Q3 earnings calls to show a measurable shift in enterprise AI spending from model licensing to data infrastructure, data quality, and system-of-record investments.

For Your Team

Monday's meeting prompt: ”Oracle just laid off 30,000 people to fund AI infrastructure. nCino just posted 17% growth because it IS the infrastructure. And private equity is buying pharmacy data companies because data is the asset, not the AI model. Here is the question: in our AI strategy, are we building the system of record, or are we building on top of someone else's?”

The System of Record Audit:

1. Map your data dependencies. For every AI initiative, identify which platform holds the authoritative data. If that platform is a vendor, review the data rights clauses in your contract.
2. Measure your capture rate. What percentage of valuable interactions, decisions, and outcomes in your business are being captured in a structured, queryable format? The gap between ”things that happen” and ”things that are recorded” is your AI training data deficit.
3. Identify your moat candidates. Which of your internal datasets could not be replicated by a competitor starting from scratch? Those datasets are your defensible position. Invest in making them larger, cleaner, and more connected.
4. Audit your agent security posture. Inventory every API call your AI tools make. How many go through authenticated, monitored endpoints? If you cannot answer that question, you have an attack surface you have not measured.

Share-worthy stat: Oracle laid off 30,000 employees, approximately 19% of its workforce, to redirect billions toward AI data centers. That is not a tech layoff. That is a capital reallocation at a scale that makes the human cost of the AI pivot explicit.

Go deeper: Track AI infrastructure and data consolidation signals in real-time →

The Track of the Day

”AI is moving quickly from help me write and help me search to help me complete meaningful productive tasks so I can focus on other work to grow my business more efficiently and profitably.”
— Sean Desmond, CEO, nCino

Today's set: ”Blue Monday” by New Order. In 1983, a band lost its frontman and decided the only way forward was to rebuild everything from the ground up. They traded guitars for synthesizers, structure for experimentation, and released a track that became the best-selling 12-inch single of all time. The factory that pressed the records (Factory Records) actually lost money on every copy sold because they spent too much on the sleeve design. The point is not the record. The point is that New Order understood something Oracle is learning the hard way: when you rebuild, the old team does not automatically come with you. But the ones who adapt to the new instruments? They create something the old approach never could. The companies that survive this AI transition will not be the ones that cut the deepest. They will be the ones that rebuild the fastest, with the people who learned the new instruments first.

Yves Mulkers, your data DJ, mixing 190,000 articles into the tracks that actually matter.

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

Published: April 3, 2026 | Curated by Yves Mulkers @ Ins7ghts