CapEx Titans & AI Factories: Oracle, Nvidia & AMD’s Trillion-Dollar Compute Gamble
Welcome, AI & Semiconductor Investors,
In today’s AI arms race, capital outlays have become the new performance metric. Oracle’s $21.2B datacenter blitz, Nvidia’s $1.5T sovereign AI factory blueprint, and AMD’s 288GB MI355X unveiling prove that compute scale now trumps code. Yet sky-high CapEx and strained supply chains threaten margins, forcing investors to weigh which juggernauts can convert scale into sustainable returns.— Let’s Chip In
What The Chip Happened?
💡 Oracle’s Cloud Orders Go Parabolic, But CapEx Bills Arrive Just as Fast
🚀 Quantum-Classical or Bust: Jensen Maps a $1.5 Trillion AI Factory Boom
🔧 AMD Arms the MI355X to Storm Nvidia’s Blackwell
[Oracle's FY25 Q4: Cloud Surge Lifts Revenue, Sets Stage for FY26]
Read time: 7 minutes
Get 15% OFF Finchat — MY FAVORITE STOCK MARKET DATA PLATFORM —
Oracle (NYSE: ORCL)
💡 Cloud Orders Go Parabolic, But CapEx Bills Arrive Just as Fast
What The Chip: On June 11, 2025, Oracle capped FY 2025 with an 11 % YoY revenue jump to $15.9 billion and non-GAAP EPS of $1.70, beating guidance. Management says the real inflection point lies ahead: they project total cloud revenue growth accelerating from 24% in FY25 to over 40% in FY26 as Oracle scrambles to build enough data center capacity to meet the “insatiable” demand for AI.
The Situation Explained:
🚀 Cloud Flywheel Spinning Faster – Q4 cloud (IaaS + SaaS) hit $6.7 B (+27 % YoY); IaaS alone surged 52 % to $3.0 B as OCI consumption revenue leapt 62 %.
📈 Bookings Bonanza – Remaining performance obligations (RPO) soared 41 % YoY to $138 B, with Safra Catz guiding to “> 100 % RPO growth” in FY 26. Roughly one-third of today’s RPO will convert to revenue within 12 months.
🛠️ CapEx: The Cost of Hyper-Growth – FY 25 capital spending hit $21.2 B (+79 % YoY) and is now pegged “at over $25 B” for FY 26 as Oracle rushes to open 47 additional multicloud regions and strap in GPUs for colossal AI clusters (project “Stargate” still forming).
📊 Guidance Ratchets Up – FY 26 total revenue outlook raised to ≥ $67 B (+16 %), driven by > 70 % IaaS growth. Q1 FY 26 revenue guide sits at +12–14 % YoY, but EPS growth is a tamer +5–7 % as tax rate and CapEx bite.
💾 Database Moat Widens – Ellison touted Oracle 23AI (vector + relational in one engine) and said multicloud database revenue from AWS, Azure & Google jumped 115 % QoQ; 23 regions live, 47 more coming.
🌐 New Logos & Workloads – ByteDance, Temu, Uber, and security up-starts are migrating performance-sensitive stacks to OCI, validating Oracle’s lower-latency network fabric and AI-centric silicon design.
⚠️ Watch the Margins – Non-GAAP operating margin ticked down 70 bps YoY to 44 % in Q4; free cash flow was –$2.9 B this quarter as datacenter build-out consumed cash.
🦾 Management Soundbite – “Demand is astronomical…customers ask us to take all the capacity you have—wherever it is,” Ellison said; Catz added, “I still waive customers off or schedule them out because supply can’t yet meet demand.”
Why AI/Semiconductor Investors Should Care: Oracle is morphing from a software-heavy vendor into a capital-intensive cloud and AI compute utility. If management executes, OCI’s scale-driven operating leverage could unlock a second growth act similar in size to Oracle’s legacy license business. Yet that upside comes with risks: ballooning CapEx, near-term margin pressure, and the need to keep GPU pipelines full. For investors, the bet is that Oracle’s differentiated autonomous stack and multicloud partnerships will translate demand into durable, high-return cash flows once the build-out phase peaks.
Get 15% OFF Finchat — MY FAVORITE STOCK MARKET DATA PLATFORM —
Nvidia (NASDAQ: NVDA)
🚀 Quantum-Classical or Bust: Jensen Maps a $1.5 Trillion AI Factory Boom
What The Chip: On June 11 2025, CEO Jensen Huang used NVIDIA’s Paris analyst call to lay out a sweeping vision: quantum-classical hybrid computers, gigawatt-scale “AI factories,” and a new RTX Pro server to drag traditional enterprise IT into the AI era, while bluntly admitting China revenue is now zero and Europe is just warming up.
The Situation Explained:
🌐 Quantum-Classical Pivot – “Every supercomputing center is going quantum-classical…CUDA-Q is the revolution.” Huang sees 20-100 logical qubits in ~5 years and plans to feed GPU clusters synthetic quantum data for AI training.
🏭 Sovereign AI Surge – Europe alone is lining up ~20 AI factories (several gigawatt-scale); globally NVIDIA pegs the next-few-years build-out at $1.5 trillion as telcos, regional clouds, and governments chase data-sovereignty.
🖥️ RTX Pro Ships Now – First “universal” AI server that runs Windows, VMware, Red Hat & NetApp—plug-and-play for legacy data centers starved of GPUs. Huang calls enterprise AI a “hundreds-of-billions” opportunity.
🔌 Supply-Chain Confidence – No current bottleneck in CoWoS or HBM: NVIDIA locks 12-month wafer-to-supercomputer cycles and places “hundreds of billions” in long-lead orders across TSMC, Samsung, Micron & partners.
🇺🇸 Made-in-USA Ramp – Plans to build $0.5 trillion of AI systems on U.S. soil with TSMC, Amkor, Foxconn, Wistron et al. Goal: diversify while “avoid conflict—job #1.”
🚫 China at $0 Revenue – Export controls cut China data-center sales from $30-40 B to zero in NVIDIA’s forecast; any policy reversal would be “a bonus.”
🔗 NVLink Everywhere – New NVLink chiplets offered to CPU/ASIC partners; Jensen expects most DIY AI ASIC projects “will get canceled,” but those that survive will buy full NVLink racks from NVIDIA.
⚡ GB300 on Track – Blackwell GB300 boards keep the same chassis as GB200; ramp begins this quarter despite last-minute GB200 bug, preserving roadmap cadence.
Why AI/Semiconductor Investors Should Care: NVIDIA’s message is clear: the spending wave is far from peaking. Sovereign AI demand, enterprise GPU starvation, and quantum-classical hybrids create fresh multi-year TAM—even with China at zero. Meanwhile, RTX Pro broadens the moat beyond hyperscalers, and a U.S. manufacturing push mitigates geopolitical risk.
Advanced Micro Devices (NASDAQ: AMD)
🛠️ AMD Arms the MI355X to Storm Nvidia’s Blackwell
What The Chip: Tomorrow, June 12, 2025 @ 12:30 p.m. ET—Dr. Lisa Su takes the “Advancing AI 2025” stage to prove AMD can hit a one-year cadence and muscle Instinct GPUs into hyperscale racks. Street chatter says the star of the show will be the MI355X, a memory-packed monster meant to blunt Nvidia’s Blackwell momentum.
The Situation Explained:
🧠 Flagship silicon – MI355X packs 288 GB HBM3E at 8 TB/s and ~120 TFLOPS FP8 inside a 1.4 kW OAM; an 8-way “OXM” sled aggregates ≈2.3 TB of HBM for big-context Llama-3/4 and GPT-5 training. Volume ramp guides to 2H 2025.
🗓️ Road-map check – Su is expected to reaffirm the annual Instinct cadence she promised at Computex 2024: MI350 family ships mid-25, while “CDNA Next” (MI400) stays on track for 2H 2026.
🛠️ ROCm 7.0 boost – New Triton kernels, Flash-Attention-3 and one-click PyTorch images headline AMD’s software push, shrinking the inertia gap vs. CUDA.
🤝 Hyperscale logos – Citi flags Amazon or OpenAI as likely on-stage partners; a big-name design win would validate MI300 traction and bolster the MI355X order book.
💰 FY-25 math – Street whispers $8-11 B Instinct revenue this year—double 2024. A $1 B guide bump could add roughly $0.15 EPS to consensus.
🏭 Capacity signal – TSMC’s CoWoS lines expand to 30-35 K wafers/month exiting 2025, easing fears that HBM and advanced-packaging shortages will choke supply.
⚠️ Bear traps – Watch for any slip beyond Q3 on MI355X production, a thin ROCm roadmap, or an empty hyperscale-customer slide; each would reignite concerns that AMD remains a distant No. 2.
Why AI/Semiconductor Investors Should Care: A proven one-year GPU cadence, a competitively priced 288 GB MI355X, and marquee cloud wins would show AMD can turn today’s MI300 foothold into a double-digit-billion AI run-rate and grab share as data-center budgets pivot to GPT-scale training. Miss those proof-points, and bulls will have to wait for MI400 while Nvidia continues compounding its lead.
Youtube Channel - Jose Najarro Stocks
X Account - @_Josenajarro
Get 15% OFF Finchat — MY FAVORITE STOCK MARKET DATA PLATFORM
Disclaimer: This article is intended for educational and informational purposes only and should not be construed as investment advice. Always conduct your own research and consult with a qualified financial advisor before making any investment decisions.
The overview above provides key insights every investor should know, but subscribing to the premium tier unlocks deeper analysis to support your Semiconductor, AI, and Software journey. Behind the paywall, you’ll gain access to in-depth breakdowns of earnings reports, keynotes, and investor conferences across semiconductor, AI, and software companies. With multiple deep dives published weekly, it’s the ultimate resource for staying ahead in the market. Support the newsletter and elevate your investing expertise—subscribe today!
[Paid Subscribers] Oracle's FY25 Q4: Cloud Surge Lifts Revenue, Sets Stage for FY26
Date of Event: June 11, 2025
Executive Summary
*Reminder: We do not talk about valuations, just an analysis of the earnings/conferences
Oracle Corporation closed fiscal 2025 with double‑digit top‑line momentum and a decisive tilt toward cloud subscription revenue. Fourth‑quarter (Q4) total revenue reached $15.9 billion, up 11 percent year over year, while full‑year (FY25) revenue climbed 8 percent to $57.4 billion. Cloud services and license support generated 77 percent of FY25 revenue and grew 12 percent to $44.0 billion. Management highlighted the scale of future demand: remaining performance obligations (RPO) expanded 41 percent to $138 billion, and operating cash flow improved 12 percent to $20.8 billion.
Chief Executive Officer Safra Catz credited broad‑based adoption of both infrastructure and applications offerings, noting, “We expect our total cloud growth rate—applications plus infrastructure—will increase from 24% in FY25 to over 40% in FY26.” Chairman and Chief Technology Officer Larry Ellison added that multicloud database revenue “grew 115% from Q3 to Q4,” underscoring traction with hyperscale partners. Together, these results position Oracle to enter FY26 with record backlog, accelerating cloud mix, and clear visibility into demand for additional capacity.
Growth Opportunities
Multicloud database services represent the most conspicuous near‑term catalyst. Oracle is already live with 23 multicloud datacenters—deployed inside Amazon Web Services (AWS), Microsoft Azure, and Google Cloud—with 47 additional sites under construction. Q4 multicloud database revenue accelerated 115 percent sequentially, and management expects “triple‑digit” growth to continue in FY26 as more enterprises adopt OCI Database Services on their preferred cloud.