Bubble or the Next AI Investment Cycle? What We Learned From Megacap Earnings

It was a blowout week for tech — NVIDIA’s GTC set the tone, and the megacaps followed through with massive capex increases that signal this AI build-out is only accelerating. While many investors came into earnings worried about a bubble, the data suggests we may be just getting started.

“AWS is growing at a pace we haven’t seen since 2022 — reaccelerating to 20% YoY.” — Andy Jassy, Amazon CEO (3Q earnings call)

In this week’s edition, we break down the developments driving the next leg of AI infrastructure growth:

• NVIDIA: Setting the Stage for Tech Earnings
• Capex Boom: Cloud Titans Double Down
• Results Deep Dive: Amazon, Microsoft, Google

NVIDIA: Setting the Stage for Tech Earnings

NVIDIA’s GTC updates were nothing short of explosive. Management guided toward $500B in revenues from Blackwell + Rubin through 2026, and we think even that might be conservative.

Key takeaways:

• Blackwell = 5x Hopper — Total Hopper sales were ~$100B while Nvidia’s current generation chip is expected to deliver $500B in sales, based on orders as of right now.
• Cost per GW and per token continue to decline. The cheaper compute gets, the more demand scales.
• Our math implies $700B for Blackwell — more on this ahead of Nvidia earnings. 

Bottom line: We are far from a bubble. NVIDIA’s GTC comments essentially set the tone for this entire earnings season — and likely for tech into 2026.

CAPEX BOOM: Cloud Titans Double Down

Every hyperscaler (Azure, AWS, Google, Meta) reported accelerating capex growth this quarter. While market reactions were mixed — Amazon and Google outperformed; Microsoft and Meta lagged — the common denominator was clear: spending is up and demand still exceeds supply.

Highlights:

• Capacity shortages now extend into 1H26, versus expectations of balance by 2H25.
• Amazon’s Trainium chips are sold out, underscoring the sustained compute crunch.
• Based on equipment orders and permits, we see a 3-year visibility into continued capex acceleration.

Investor takeaway: The market is still building the infrastructure layer required to unlock the next wave of AI applications. Returns will follow — but only after this foundational build-out is complete.

Company Deep Dive

Amazon (AWS): Leading the Build-Out

• AWS growth: +20% YoY vs. 18% expected — a clear beat amid macro uncertainty.
• Infrastructure: +3.8 GW TTM, +1 GW expected in 4Q, and Project Rainier adding 2.2 GW once all 30 DCs go live.
• Backlog: October bookings eclipsed all of 3Q’s total deal volume — visibility is strong into 2025.
• Trainium2: +150% QoQ growth → now a multi-billion-dollar business ahead of the T3 launch.

Microsoft (Azure): Growth Meets Constraint

• Azure growth: +39% YoY, largely in line with expectations post recent OpenAI-related deals. Capex revised 
• Capacity constraints: Limiting short-term upside — outlook affected by supply constraints by ~200 bps. 
• Focus is shifting to returns on AI investments and growth in applications (e.g., Copilot) vs. raw compute capacity. Still early innings and more to come in ’26.

Google Cloud: Execution and Monetization

• Growth: +29% YoY — impressive given the tough 3Q comp.
• Profitability: Continued operating margin expansion despite elevated AI infrastructure spend.
• AI monetization: Early traction in Vertex AI and Gemini integration across Workspace; signs of AI-driven lift in ad targeting efficiency.
• Customer demand: Enterprise AI and data analytics deals scaling; notable momentum from startups and mid-market customers deploying GenAI use cases.

Big Picture

We’re entering the next phase of the AI infrastructure cycle — one defined by massive investment, declining unit costs, and widening adoption curves. With supply still chasing demand, the setup into 2026 remains structurally bullish for compute, capex, and cloud infrastructure.