Nerd Out on Business is back!

After taking a few months off to focus on a new AI education and community business that I started for people in the agriculture industry called Fullstack Ag, I missed writing these producing these reports! So, I'm back.

Here are a few things to note as I restart this:

• I'm going to start producing these reports three days a week: Tuesday, Thursday, and Saturday.

• The goal of this newsletter is to give you insight into a lot of different companies by looking at how they operate and how they make money. This isn't meant to be any sort of investment newsletter. With that said, I'm going to start including some stock market performance data, as that's obviously an important part of any public company's story.

• For those that don't know, I founded a farm management software company called Harvest Profit that was acquired by John Deere in 2020. A lot of the previous reports I wrote focused on my lessons with Harvest Profit and how each day's report applied to those lessons. Now I have Fullstack Ag to share more about, as that's a current venture. I'll be sharing more insight into how I'm thinking about that business.

• The team behind Fullstack Ag is going to start to produce workshops and products that are applicable outside of agriculture. Things like how we're using the Claude set of tools or how we're building and hosting AI-generated software applications. In early June, we'll be hosting our first Nerd Out paid workshop, so keep your eye open for that.

With that, today, I'm digging into Oracle (ORCL).

They build the databases, business applications, and cloud infrastructure that run the back office of much of corporate America. Over the past 18 months, they've also become one of the biggest landlords for AI training workloads, with deep ties to OpenAI and the Stargate buildout.

Historically, when I heard commentary about Oracle, it revolved around legacy software and good but expensive databases. But recently, its founder, Larry Ellison, decided to make a big bet on AI.

A few things that stood out to me:

• Capex has gone from boring to bet-the-company. Capital expenditures jumped from $8.7B in FY2023 to $21.2B in FY2025, and TTM capex now runs at roughly $47B. That is more than 75% of revenue going into the ground as data centers and servers.

• The cash machine and the AI build sit inside the same company. Cloud and License still throws off a ton of cash (operating margins above 30%), and management is funneling all of it, plus a wall of new debt, into AI infrastructure. Net debt has climbed from roughly $81B at the end of FY2023 to about $124B today.

• The stock has decoupled from the rest of enterprise software. Oracle is up 20% over the past year while SAP and Salesforce are both down more than 40%. The market is paying for the AI story, but the 52-week range of $135 to $346 tells you it has not been a smooth ride.

But it's not all clean and perfect:

• Free cash flow turned negative in FY2025 (negative $0.4B) for the first time in living memory, and TTM free cash flow is deeply negative as capex outruns operating cash flow.

• Similar to a lot of companies that I've profiled, customer concentration is an issue here. A huge slice of the AI cloud backlog reportedly traces back to a handful of frontier model labs, and those workloads can move.

Oracle spent 40 years building a slow-moving, ridiculously profitable software business with sticky customers.. Now they are using that cash flow as the down payment on a hyperscaler-style infrastructure bet, compressing a decade of investment into 18 months.

this is going to be an interesting one to follow, given the aggressive nature of their AI infrastructure build out.

One ancillary lesson here is how founder Larry Ellison has used his equity in Oracle to become one of the wealthiest people in the world.

As I think about the ventures that I am working on and going to work on, Larry's approach to keeping as much equity as possible in a business that he deeply understands really appeals to me.

Larry has rarely sold shares in any meaningful size, while the company has spent more than $100B on buybacks over the past fifteen years, including stretches where Oracle bought back around 10% of its shares outstanding in a single year. His share count stayed roughly flat while the denominator shrank, so his stake quietly compounded from the low-20s as a percentage of shares outstanding to north of 40% today.

Most founders dilute themselves slowly into the single digits through secondaries, stock comp, and the gravitational pull of "taking some off the table." While I think it's prudent to take some off the table, Ellison did the opposite. I’m a fan.

His shares in Oracle are worth approximately $215 billion today.

With that, I'll see you on Thursday!

Nick

TL;DR

• Oracle sells enterprise databases, ERP, HCM, and industry applications, plus a fast-growing cloud infrastructure (OCI) business that hosts large AI workloads.

• Cloud and License is about 86% of FY2025 revenue at $49.2B; Services adds $5.2B and Hardware $2.9B.

• Recent results show revenue growth accelerating into the teens and operating margins above 30%, with trailing capex of roughly $47B funded by $44B of new net debt over two years.

• The stock has compounded 156% over five years and 443% over ten, but the 1Y total return of +20% sits inside a wild trading range as AI infrastructure hopes and fears whipsaw the multiple.

• Operator takeaway: a legacy cash cow can fund an aggressive new bet, but only if the new business serves real, paying demand and not a single concentrated customer.

The 30,000-Foot View

Oracle sells the software that companies use to run themselves. The flagship product is still the Oracle Database, the system of record for finance, supply chain, and operations at a huge number of large enterprises. Around that core, they have layered on Fusion Cloud applications (ERP, HCM, SCM), NetSuite (an SMB-focused ERP suite they bought in 2016), and Cerner (healthcare records, acquired in 2022). Margins on this legacy software stack are high because customers cannot easily rip it out without disrupting the business.

The new chapter is Oracle Cloud Infrastructure (OCI). For years it was a distant fourth or fifth in the cloud race, well behind AWS, Azure, and Google Cloud. The pivot started by building a network optimized for high-bandwidth, low-latency workloads, which turned out to be exactly what the AI training boom needed. Over the past two years, Oracle has signed massive multi-year capacity deals with frontier AI labs, most notably OpenAI through the Stargate partnership, and is in the middle of a brutal capex cycle to deliver that capacity.

The business model is a mix of recurring software subscriptions, multi-year license support contracts, and cloud consumption. Customers get locked in by data gravity (their data lives in Oracle), by integration depth (workflows are wired through Oracle apps), and by the regulatory burden of switching off a system of record.

Revenue mix (FY2025):

• Cloud and License Business: ~85.8% ($49.2B)

• Services Business: ~9.1% ($5.2B)

• Hardware Business: ~5.1% ($2.9B)

Geographic mix (FY2025):

• Americas: ~63%

• EMEA: ~24%

• Asia Pacific: ~12%

Key Stats

• Market cap: $537.2B

• TTM revenue: ~$62.6B

• TTM gross margin: 66.4%

• TTM operating margin: 30.8%

• Employees: ~162,000

• 1Y total return: +20.0%

• Industry: Software, Infrastructure

Company History

• 1977: Larry Ellison, Bob Miner, and Ed Oates found Software Development Laboratories, the company that would become Oracle.

• 1979: Oracle launches the first commercial relational database, beating IBM to market with the very technology IBM Research invented.

• 1986: IPO on the NASDAQ at $15 per share.

• 1990s: Becomes the dominant relational database vendor as client-server computing takes over the enterprise.

• 2005: Acquires PeopleSoft for $10.3B after a long hostile battle, kicking off a decade of aggressive M&A in enterprise apps.

• 2010: Closes the Sun Microsystems acquisition for $7.4B, bringing Java, Solaris, and a hardware business in-house.

• 2016: Buys NetSuite for $9.3B, adding a true cloud ERP suite for mid-market customers.

• 2019: Hires former Tesla and Microsoft cloud talent to seriously rebuild OCI as a workload-optimized cloud.

• 2022: Closes the Cerner acquisition for $28.3B, the largest deal in company history, bringing in electronic health records.

• 2023: Headquarters officially moves from California to Austin, Texas.

• 2025: Announces a central role in the Stargate AI infrastructure venture with OpenAI and partners; capex roughly doubles year over year as data center construction accelerates.

• 2025–2026: Annual capex runs at the highest level in company history (over $21B in FY2025, with TTM capex pushing toward $47B), funded by a steadily growing debt stack.

Show Me the Money

Standout Financial Features:

• Revenue growth has accelerated, from $50.0B in FY2023 to $57.4B in FY2025, with TTM tracking around $62.6B. That is a noticeably faster ramp than the low-single-digit growth Oracle ran on for most of the 2010s.

• Operating margin expanded from 26.2% in FY2023 to 30.8% in FY2025 even as the cloud build accelerated, a remarkable result if you have ever watched a software company try to add infrastructure margin to an apps business.

• The bottom of the income statement is great but the middle of the cash flow statement is where the story changes. Operating cash flow of $20.8B in FY2025 was overwhelmed by $21.2B of capex, flipping free cash flow negative for the year.

• Net debt has increased $43B in the last two years to fund the infrastructure buildout.

Financial Data

Stock Performance

1-Year peer comparison:

Oracle outperformed two of its most direct enterprise application peers by over 60 percentage points over the past year, almost entirely on AI infrastructure expectations. The longer-run picture is even more flattering: a 443% ten-year return for a "boring" database company is not what most people would have guessed in 2016. Worth noting that the 1Y return masks a huge intra-year swing, with the stock trading between roughly $135 and $346 over the past 52 weeks.

(Peer rationale: SAP and Salesforce were chosen because both compete directly with Oracle's application businesses, ERP and CRM respectively, and both sit at roughly Oracle-comparable enterprise scale rather than the hyperscaler peers that would dwarf the comparison.)

The N.O.O.B. Nine: Competitive Powers

The Nerd Out on Business Nine is made up of Hamliton Helmer's famous "7 Powers" of competitive advantage (Scale Economies, Network Economies, Counter-Positioning, Switching Costs, Branding, Cornered Resource, and Process Power) combined with two of my own (Data Flywheel and Distribution Advantage).

Average Score: 3.6/5 - The moat is anchored by switching costs and brand, supported by scale and distribution, with the cloud business adding optionality rather than yet adding durable advantage.

Memorable Marketing

Oracle’s marketing has never felt separate from its sales machine. The point was not to make people feel something about Oracle. The point was to make sure Oracle was already in the room when a big software contract came up. The brand worked because it was everywhere a CIO looked, and because the sales team knew how to turn that visibility into a signed deal.

Notable campaigns and tactics:

• Oracle CloudWorld (annual): Flagship customer and partner conference that anchors product launches, keynote announcements, and big partnership reveals. Used aggressively in 2024 and 2025 to surface AI customer wins.

• The "We Don't Compete With Our Customers" positioning (ongoing): A subtle but consistent message aimed at enterprises and AI labs nervous about hosting workloads with Amazon, Microsoft, or Google, each of whom also runs competing businesses or models.

• Ellison-as-spokesperson moments (recurring): Larry Ellison still anchors big keynote moments, often dropping unscripted shots at AWS and Microsoft. Polarizing, memorable, and earns coverage that paid media cannot.

• Stargate and OpenAI co-branding (2025): Oracle's inclusion in the headline AI infrastructure venture functioned as the most expensive piece of marketing in company history, instantly reframing OCI from also-ran to top-tier.

• Sports sponsorships (Red Bull Racing F1, America's Cup): Title sponsorships of premium global properties keep the Oracle brand in front of C-suite buyers in markets where billboards do not reach.

Tactical takeaways:

1. Pick a positioning your competitors cannot copy without breaking their own business model. "We don't compete with our customers" is hard to imitate if you have a competing business.

2. Use a single flagship event as your marketing tentpole, then build the rest of the year around it. This can drive both marketing and product cadence and come to be something your customers look forward to.

3. A polarizing founder voice is a marketing asset if you let them be themselves.

4. Headline customer wins double as the most credible marketing you can buy.

AI Uses & Opportunities

Current exposure:

• OCI is a primary training and inference platform for several frontier AI labs, including OpenAI through the Stargate venture, which has made AI infrastructure the fastest-growing piece of Oracle's revenue mix.

• Autonomous Database uses machine learning for tuning, patching, and security, reducing the need for human DBAs and making the product stickier.

• Oracle has embedded generative AI features across Fusion Cloud Applications and NetSuite, with focus on automating finance and HR workflows where the data already lives in Oracle.

• The Cerner healthcare unit is being rebuilt around AI-native clinical assistants, with the goal of compressing physician documentation time.

Future opportunities:

• Sovereign and regulated AI cloud, where customers in defense, healthcare, and financial services need a non-hyperscaler option with strong data residency.

• AI for ERP and HCM, where Oracle's structured data assets are arguably better suited to enterprise agents than the unstructured corpora powering consumer chatbots.

• Vertical AI products built on Cerner data, particularly around population health, clinical decision support, and revenue cycle management.

• Sales and support automation that finally takes some of the labor cost out of the famously expensive Oracle account model.

• Database-native vector and embeddings workloads, letting customers run retrieval-augmented workflows without moving their data elsewhere.

Bumps in the Road

• Capex is running at a level that requires the AI customer story to keep paying off. If frontier-lab demand softens or shifts to in-house silicon, Oracle is left holding a very expensive bag.

• A large slice of the AI cloud backlog is reportedly concentrated in a small number of customers, which is the opposite of the long-tail diversification that built the legacy business.

• Its Cerner acquisition has been slower to pay off than expected.

• Long-term debt has grown materially, and interest expense of roughly $3.6B per year is now a meaningful drag on net income.

• The legacy on-premise database business still funds the new bet, but it is in slow secular decline, which puts pressure on cloud growth to deliver on time.

Your Swipe File

• Keep an eye on Oracle and their AI bet. It’s going be a fun bet to watch play out in realtime!

• Use your most profitable legacy business as the funding source for your most aggressive new bet, rather than constantly trying to keep them in balance.

• Credibly saying "we don't compete with our customers" is powerful marketing.

• A polarizing founder voice and a single flagship customer conference can do a lot of brand work.

• When you take a customer-concentration bet, structure the contract so that capex risk is shared (if possible).

• Sticky, mission-critical software is still one of the best businesses in the world.