Let’s try that again! The exciting event for technology this week is NOT Nvidia’s GTC. We expect few surprises there. But the sector that has vastly outperformed the semiconductor sector in the last year is the optical sector. Yes, photonics are where it’s at, so we will be at OFC 2026 🔦, the annual confab gathering the brightest lights (ahem) in the optical sector, this year in Los Angeles. Look for companies like Applied Optoelectronics (AAOI: NASDAQ), Coherent (COHR: NYSE), Corning (GLW: NYSE), Lumentum (LITE: NASDAQ) and Fabrinet (FN: NYSE) to have potentially big moves on the news.

You can find prior notes and the full research archive at https://epistrophy.beehiiv.com.

As always, I’m focused on three things:
1) Technology-driven change;
2) the latest in innovation and startup trends, and;
3) stock fraud.

Companies Discussed

In This Note:

1. Companies Discussed

2. OCI: The Software Engine

   1. Oracle’s growth now runs through its data centers.

   2. OCI As A Leader

3. Tweet O’ The Week

4. 📆 of Epistrophy Events

5. Availability This Week

OCI: The Software Engine

Oracle’s growth now runs through its data centers.

Oracle (ORCL:NYSE)  reported fiscal third-quarter revenue of $17.2 billion on March 10, up 22%. Adjusted earnings per share reached $1.79, also up 22% – that’s the first time in more than a decade that revenue and EPS rose at better than 20%.

This software growth isn’t in spite of the efforts put forth towards the $20 billion annual Oracle Cloud Infrastructure business – it’s because of it. 

As we pour through the results of this quarter and listen to the litany of contract wins, it’s increasingly clear that the cloud business is driving the rest of Oracle’s business. Q3 Cloud revenue climbed 44% to $8.9 billion. Remaining Performance Obligations, Oracle’s measure of contracted future revenue, reached $553 billion.

Those numbers landed in the middle of a debate sweeping the software industry. Artificial intelligence can now generate code, design workflows and execute tasks once embedded inside applications. It’s not the woke virus, but the AI “Rack Virus” that is poisoning the minds of Wall Street – the belief that racks of AI models will obliviate the need for software businesses. Oracle’s quarter suggests the opposite. The future of enterprise software may depend on who owns the infrastructure where AI runs.

OCI As A Leader

Oracle’s growth now runs through its data centers.

Oracle Cloud Infrastructure generated $4.9 billion in revenue during the quarter, up 84% year over year. Modern AI training environments require dense computing architecture. Thousands of GPUs operate simultaneously across low-latency networking fabrics. High-bandwidth optical interconnects carry gradients and training data between nodes during distributed training. High-throughput storage pipelines feed the cluster continuously. Dense GPU racks draw large power loads and require advanced thermal management.

Oracle’s infrastructure integrates these components with its database platforms. Oracle Autonomous Database and related services operate inside the same cloud environment as the GPU clusters. The arrangement places enterprise data close to the compute resources analyzing it.

Artificial intelligence operates on data. Corporate databases store financial records, logistics metrics, customer histories and operational data. Training models on those datasets requires direct access to the databases where the information resides. Running AI workloads inside OCI allows those models to interact continuously with Oracle’s data platforms. Applications such as Fusion ERP and Fusion HCM operate above that data layer, executing business workflows that rely on both.

Demand for that architecture appears in Oracle’s contract backlog. Remaining Performance Obligations reached $553 billion at the end of the quarter. That’s not all data centers. The metric represents revenue already committed through signed agreements that will convert into future sales. Oracle highlighted several enterprise wins across applications and infrastructure during the quarter. FedEx (FDX: NYSE) selected Oracle Fusion Cloud ERP and Oracle Fusion Cloud SCM to modernize finance and supply chain systems, a competitive displacement of SAP (SAP: NYSE). Grocery chain Kroger (KR: NYSE) expanded deployments of Oracle Fusion Cloud HCM for workforce management, while Albertsons (ACI: NYSE) adopted Fusion HCM in a win against Workday (WDAY: NASDAQ). Healthcare provider Cleveland Clinic implemented Oracle Fusion Cloud ERP to modernize financial operations across its hospital network. On the infrastructure side, enterprises expanded AI training workloads on Oracle Cloud Infrastructure and deployed Oracle Database@Azure, allowing Oracle Autonomous Database and Oracle Database 23ai to run inside Microsoft (MSFT: NASDAQ) cloud environments while connecting those workloads back to OCI.

Training large models requires predictable computing resources. Organizations therefore commit to multi-year agreements to secure data center capacity. Oracle’s backlog suggests that many customers have already reserved infrastructure for future AI workloads.

Enterprise Software Matters

Oracle’s traditional software business remains intact alongside the infrastructure expansion.

The company reported roughly 4% growth in software license revenue during the quarter. License sales include database technology and enterprise applications deployed inside corporate environments. That segment accounts for roughly 29% of Oracle’s total revenue.

Oracle’s application portfolio is clearly gaining new traction. Fusion ERP manages financial accounting and procurement workflows. Fusion HCM handles payroll and workforce management. Industry-specific applications support healthcare providers, retailers and financial institutions. Each system maintains the operational records organizations rely on.

Artificial intelligence increasingly interacts with those records. Oracle has begun embedding AI agents within parts of its application suite. These agents analyze enterprise data and automate tasks inside existing workflows. Finance applications can forecast cash flows or reconcile transactions. HR systems can analyze workforce patterns or automate administrative processes.

The applications remain the system of record while AI functions as a computational layer above them. That structure preserves the role of enterprise software while expanding the capabilities operating on its data.

Infrastructure Strengthens Software

AI workloads require continuous access to structured enterprise data. That data resides inside Oracle databases and applications. Running models within OCI therefore increases the number of database queries and application interactions occurring across the system.

A supply chain model analyzing inventory levels repeatedly queries the database storing logistics records. An AI agent automating procurement triggers transactions inside the enterprise application managing purchasing workflows. Each AI process generates new activity inside the underlying software platform.

Oracle has also extended its database technology into rival cloud environments. The company now deploys Oracle Database inside cloud platforms operated by Amazon (AMZN: NASDAQ), Microsoft and Alphabet’s Google (GOOG: NASDAQ). Customers can run Oracle databases within those environments while connecting workloads to Oracle services.

The strategy reflects the fragmented nature of modern enterprise computing. Large organizations often operate across several cloud providers while maintaining internal systems. Oracle’s database technology operates across those environments while still connecting back to Oracle infrastructure and applications.

Co-CEO Clay Magouyrk, Oracle’s OCI chief, said new AI infrastructure deals often require customers to fund the hardware directly. “A combination of bring your own hardware and upfront customer payments enables us to continue expanding without any negative cash flow from Oracle,” he said on the earnings call.

Higher utilization improves the economics of each facility. As GPU clusters install and workloads begin running, revenue from those contracts converts into operating income. The infrastructure layer generates additional activity across Oracle’s databases and applications at the same time.

The predictive question emerging from the quarter concerns how deeply artificial intelligence becomes embedded in enterprise operations. Early deployments focus on analytics and workflow automation. As models improve, AI agents may assume broader roles within finance systems, supply chains and workforce management platforms.

Those agents will require constant access to enterprise data. Databases storing that information will remain central to the computing architecture. Infrastructure capable of hosting large AI workloads near those systems will therefore gain strategic importance.

Oracle’s strategy places the company at that intersection. Enterprise data resides inside its databases. AI infrastructure operates inside OCI data centers. Enterprise applications execute the workflows connecting the two.

Artificial intelligence may reinforce the companies controlling enterprise data rather than displacing them. Oracle’s results suggest that the future of enterprise software could depend as much on power capacity and GPU clusters as on the applications themselves.

Limp Software Loans

A look at TPVG’s loan book.

We got some customer questions about the specific loans help by the topic of last week’s newsletter, private credit lender TriplePoint Venture Group (TPVG: NASDAQ).

Yes, it’s the same well known Sand Hill Road venture firm TriplePoint Capital that once backed widely recognized and industry-leading companies like YouTube (acquired by Alphabet (GOOG: NASDAQ)), AppNexus (acquired by Microsoft (MSFT: NASDAQ)), Beyond Meat (BYND: NASDAQ), Chegg (CHGG: NYSE), Etsy (ETSY: NASDAQ), OncoMed (OMED: NASDAQ) (later acquired by Mereo BioPharma), Proteolix (acquired by Onyx Pharmaceuticals, later acquired by Amgen (AMGN: NASDAQ)), RingCentral (RNG: NYSE), Ruckus Wireless (acquired by CommScope (COMM: NASDAQ)), Segway (acquired by Ninebot), Shazam (acquired by Apple (AAPL: NASDAQ), Splunk (acquired by Cisco Systems (CSCO: NASDAQ)), Square (XYZ: NYSE), and Workday.

But the publicly-traded vehicle it is now full no-name software companies that dominate. An examination of the most recent TPVG 10-K shows that more than one-third of the loan book is to troubled software sector.

The company is unclear how it selects which loans the advisors select for the public shareholder. Software was just $125.8 million, or 18.6%, of the loan book a year ago. Today it is $250.2 million, or 31.9%.

This may reflect desperate companies seeking more loans. But it clearly reflects the TriplePoint parent company dumping software loans into the public vehicle.

Either could be seen as a troubling sign.

Tweet O’ The Week

📆 of Epistrophy Events

Availability This Week

I’ll be all over this week, covering NVIDIA’s GTC in San Jose, then off to the Optical Communications Fiber 🔦 conference in Los Angeles, then in San Francisco office later in the week. Reach out!

Written reports are available to clients, with video summaries on YouTube, and of course our popular summaries of the summaries on Instagram, TikTok, and YouTube Shorts.