We track the fifty most compelling companies in technology. Last week, we analyzed key reports from Adobe (ADBE: NASDAQ), Oracle (ORCL: NYSE), and Rubrik (RBRK: NASDAQ).

This week, attention turns to big conferences with Nvidia GTC in San Jose, CA, followed by the Elastic Public Policy Summit in Washington DC – we’ll be covering both in size. And look for an important quarterly earnings report from Micron (MU: NYSE).

You’ll find a full calendar of events that could matter to you at the end of this note.

Our research notes are available for paying clients. And of course for rich video summaries of our research subscribe to our YouTube channel—less reading, more watching!

Our focus remains:
1) Technology-driven change
2) Emerging innovation and startup trends
3) Stock fraud

Companies Discussed

In This Note:

• Companies Discussed

• GTC 2025 Preview: ♟️Jensen’s Next Move

• The $130B Elephant In The Room

  • Top Ten Largest Order Books, Ever

• Amazon: The Optical $tandard?

• 📆 of Epistrophy Events

• Availability This Week

GTC 2025 Preview: ♟️Jensen’s Next Move

Nvidia’s biggest event of the year isn’t just about GPUs anymore. At GTC 2025, the company will push deeper into AI infrastructure, networking, and quantum computing, reinforcing its ambitions beyond hardware.

We expect Nvidia (NVDA: NASDAQ) to use GTC 2025 to reinforce its position in AI hardware while addressing concerns about competition, supply constraints, and regulatory challenges. While Nvidia holds an estimated 82% share of the AI GPU market, hyperscalers such as Amazon, Microsoft, and Google are investing in custom AI chips. U.S. export restrictions, particularly on sales to China, further complicate Nvidia’s outlook.

The most watched moments will be and should be the address from the leather-jacketed CEO Jensen Huang’s on March 18 (10 am PT) where he will likely introduce the GB300 AI chip, an evolution of the Blackwell architecture, with shipments expected in May. Nvidia may also outline early details of the Rubin GPU, due in 2026, as part of its strategy to maintain architectural leadership (we expect that it is named for producer Rick Rubin, not the sandwich or the Fifth Avenue NYU dormitory of the same name) . GTC has historically set the AI industry’s technological roadmap, and we expect this year’s event to follow that pattern.

Beyond GPUs, Nvidia is expected to emphasize its AI computing infrastructure. Its networking portfolio—including InfiniBand and Ethernet solutions—has become integral to AI data centers, and we anticipate updates in this area. Nvidia will likely expand its messaging on sovereign AI, robotics, and quantum computing. The introduction of a dedicated Quantum Day signals its interest in post-GPU architectures, possibly as a hedge against long-term reliance on existing computing paradigms.

Expected product announcements at GTC 2025 include:

• GB300 AI Chip – The latest AI accelerator, expected to succeed Blackwell with improved efficiency and performance.

• Rubin GPU Architecture – The next-generation GPU series, planned for 2026, promising a significant leap in computing power.

• Grace Hopper Superchip – A hybrid CPU-GPU designed for AI and high-performance computing workloads.

• BlueField DPU – A data processing unit aimed at improving networking, security, and storage acceleration in AI data centers.

• Spectrum-X Networking Platform – Nvidia’s latest high-performance Ethernet networking solution optimized for AI workloads.

This is an interesting time for the now-dominant Nvidia. Nvidia’s ecosystem advantage is also its greatest challenge. CUDA has historically locked AI developers into its hardware, but in-house silicon from cloud providers and alternative architectures are eroding that lock-in. GTC 2025 will reveal whether Nvidia can strengthen its software and services to maintain its hold on AI development.

ASICs and FPGA-based accelerators are emerging as viable alternatives to Nvidia’s GPUs, with companies seeking more power-efficient and cost-effective solutions. AMD (AMD: NASDAQ) is pushing its MI300 series, designed for AI workloads that demand high memory bandwidth and computational efficiency. Meanwhile, Positron, a rising player in AI acceleration, is developing custom ASICs optimized for large-scale model inference (full disclosure, I’ve backed that company). As AI workloads diversify, these alternative architectures threaten to erode Nvidia’s dominance, particularly among cloud providers seeking to optimize their infrastructure.

We expect GTC 2025 to feature over 1,000 sessions, 2,000 speakers, and 400 exhibitors, including AWS (AMZN: NASDAQ), Cadence (CDNS: NASDAQ), Cisco (CSCO: NASDAQ), Google Cloud (GOOGL: NASDAQ), Microsoft Azure (MSFT: NASDAQ), Snowflake (SNOW: NYSE), and Synopsys (SNPS: NASDAQ). The conference will likely showcase practical AI applications across industries, from healthcare and finance to automotive and manufacturing. Nvidia must demonstrate its continued relevance as alternative architectures emerge.

Looking ahead, GTC 2025 should provide insight into Nvidia’s strategic direction. Will it push deeper into software and AI services to diversify beyond hardware? Will it accelerate product releases to stay ahead of competitors? And how will it respond to growing demand for cost-effective AI solutions? The answers to these questions will shape Nvidia’s trajectory, with GTC offering the first signals of its next move.

The $130B Elephant In The Room

The number is so large it almost seems like a typo: $130 billion. Oracle (ORCL: NASDAQ) announced this staggering figure on March 11, 2025, during its Q3 earnings report. Oracle’s remaining performance obligation (RPO) as of Q3 2025 represents revenue that customers have committed to pay but that Oracle hasn’t yet delivered. And it may be one of the three largest forward sales backlog in the history of enterprise technology—if not business itself.

Somehow, this monumental achievement barely registered. The stock actually sold off on the news as the Street seemed to focus  on the 6% year-over-year revenue growth (to $14.1 billion) and the 22% surge in net income ($2.9 billion). But those numbers, impressive as they are, miss the point. (Bias report: I own Oracle).

Of that $130 billion, $117 billion comes from cloud services. Oracle has been dismissed for years as a cloud also-ran, a legacy player that moved too late to challenge Amazon Web Services (AWS), Microsoft Azure, and Google Cloud. But AWS, the market leader, reported a backlog of $156 billion in its most recent quarter—only about 20% more than Oracle. Microsoft’s RPO, by comparison, stood at around $212 billion last quarter, but that includes long-term commitments across its entire software business, not just cloud. Salesforce, the dominant SaaS provider, finished the fiscal year with an RPO of $63 billion—less than half of Oracle’s.

Oracle’s backlog isn’t just large—it’s growing faster than anyone else’s.

Top Ten Largest Order Books, Ever

1. Microsoft – $298 B: Reported Jan. 29, 2025: dominant in software and cloud contracts.

2. Amazon Web Services – $177 B: Reported Feb. 6, 2025, leading cloud computing backlog.

3. Oracle – $130 B: Reported Mar. 11, 2025, fastest-growing RPO, driven by cloud demand.

4. Alphabet – $93.2 B: Reported Dec. 31, 2024, Google Cloud and advertising-related commitments.

5. SAP (SAP: NYSE) – $68 B: Reported Jan. 25, 2025, strong cloud ERP contract base.

6. Salesforce (CRM: NYSE) – $64 B: Reported Jan. 31, 2025, largest SaaS RPO in enterprise software.

7. Cisco – $41 B: Reported Jan. 25, 2025, software and services backlog growth.

8. Workday (WDAY: Nasdaq) – $25 B: Reported Jan. 31, 2025, growing enterprise SaaS subscription commitments.

9. ServiceNow (NOW: NYSE) – $22 B: Reported January 29, 2025, dominant in enterprise IT service contracts.

10. Adobe (ADBE: Nasdaq) – $20 B: Reported April 22, 2025, major digital media and experience commitments.

These figures redefine the scale of enterprise software sales and highlight Oracle’s emergence as a major force in cloud computing. The RPO increased 63% year-over-year in Q3. AWS’s backlog grew just 19%. Salesforce’s? 11%.

These are not the growth numbers of a company playing catch-up. This is a company that has already caught up and is now pulling away.

The core of Oracle’s success is a simple but brutal strategy: sign bigger, longer, more locked-in cloud deals than anyone else in the industry. Oracle isn’t just selling cloud capacity or SaaS subscriptions. It’s locking down entire IT infrastructures for the long haul, using a combination of technological lock-in, aggressive pricing, and sheer execution power.

And it’s paying off. CEO Safra Catz put it plainly on the earnings call: “Oracle signed sales contracts for more than $48 billion in Q3, pushing our RPO up 63% to over $130 billion.” That’s nearly 3.5 times quarterly revenue in new, locked-in commitments in a single three-month period.

This is unprecedented. It is, quite simply, one of the most staggering sales achievements in the history of enterprise technology.

A huge part of this backlog is AI-driven demand. OpenAI, xAI, Meta, NVIDIA, AMD—all have signed on to Oracle Cloud Infrastructure (OCI) to power their next-generation AI workloads. GPU consumption on Oracle’s cloud skyrocketed 244% year-over-year.

And Oracle is racing to keep up. The company is on track to double its data center capacity this year. That’s not just growth—it’s an arms race. These aren’t vanity deals—they’re long-term strategic bets on Oracle’s infrastructure. They’re betting their future AI strategies on Oracle’s infrastructure.

For years, Oracle has been an afterthought in cloud computing. The market still thinks of it that way. But a $130 billion backlog rewrites the narrative. If this were Amazon, Microsoft, or Google posting these numbers, investors would be losing their minds. If Salesforce announced a 63% increase in backlog, its stock would be up 20% in a day.

Instead, Oracle books what may be one of the largest enterprise sales backlog in history—and the industry shrugs.

Amazon: The Optical $tandard?

The biggest one-day event in the history of the optical data center business belonged to two companies: Applied Optoelectronics (AAOI: NASDAQ and ) and Fabrinet (FN: NYSE). Their shares surged after Amazon backed investments in both, sending a clear signal that optical networking is no longer a niche concern but a foundational pillar of AI infrastructure. The industry took notice, as these two companies now sit at the center of the race to expand bandwidth for data-intensive computing.

For years, the limiting factor in cloud and AI computing has not been merely raw processing power but the bandwidth to move data at scale. The surge in demand for AI workloads has pushed hyperscalers like Amazon Web Services (AWS), Microsoft Azure, and Google Cloud to rethink infrastructure. Optical interconnects—particularly transceivers, cabling, and switches—have emerged as critical choke points. Whoever controls the supply of these components controls the pace of innovation. Applied Optoelectronics and Fabrinet are increasingly central to this equation.

Applied issued a warrant that allows Amazon to purchase up to 7,945,399 shares of Applied Optoelectronics' common stock at an exercise price of $23.6954 per share. This warrant is exercisable until March 13, 2035. Notably, 1,324,233 shares vested immediately upon issuance, with the remaining shares set to vest over time, contingent upon Amazon's discretionary purchases from Applied Optoelectronics, potentially totaling up to $4 billion. 

This strategic arrangement provides Applied Optoelectronics with a substantial potential revenue stream, aligning with Amazon's interests in securing a reliable supply of high-speed optical transceivers essential for its data centers. The immediate vesting of a portion of the warrant shares underscores Amazon's commitment to this partnership, while the performance-based vesting structure incentivizes Applied Optoelectronics to meet Amazon's procurement needs effectively.​

Fabrinet sold Amazon the right to acquire up to 381,922 shares at $208.4826 per share. We think that Amazon has committed to some $400 million in annual module purchases from Fabrinet as part of that deal. 

These are game changes – and at a pivotal time for both companies. Fabrinet and Applied Optoelectronics sit at opposite ends of the optical networking ecosystem. Fabrinet specializes in precision manufacturing of optical and electromechanical devices, serving as a key supplier for Cisco, Nvidia and others. Applied Optoelectronics, meanwhile, designs and manufactures high-speed fiber-optic networking products, primarily for data centers and broadband applications. Their expertise in producing high-speed, low-latency optical connections makes them indispensable to the AI-driven shift in data center architecture.

The significance of optical connectors extends beyond these two companies. Google Cloud, Meta, Microsoft and Oracle are all major buyers of high-speed interconnects, and any disruption in supply could ripple through the industry. The rise of AI-specific networking—such as Nvidia’s InfiniBand and Ethernet-based solutions—complicates the landscape further. If Applied Optoelectronics and Fabrinet can capitalize on this moment, they may find themselves not just securing their market position but redefining it.

📆 of Epistrophy Events

Availability This Week

I’ve got a packed travel schedule next week—Los Angeles to Silicon Valley, then Washington D.C. and New York, with another trip to New York not far behind. Nvidia’s conference could dominate the headlines, but with the news cycle as volatile as ever, surprises are a given.

You can follow our analysis across multiple platforms:

• Full written reports for clients

• Video breakdowns on YouTube

• Executive summaries on Instagram and TikTok (@drilldownpod)

If you’ve got sharp insights on economic indicators or enterprise tech spending, I’m all ears. And if you know someone who’d find this analysis valuable, send them my way.

Looking forward to the conversation.