🧠 Meta Spends $100B—And Still Needs More NVDA

🌞Good Monday Morning, Folks!

Last week? A classic case of market theater.

NVDA dips on a headline—“Meta’s testing TPUs!”—and suddenly, the AI bull case is crumbling. Brokers flash red. Reddit blows up. Everyone acts like NVIDIA just got dethroned.

Give me a break.

While the crowd’s playing panic whack-a-mole, trillion-dollar companies are still begging for more chips—even after spending $100 billion this year. They’re not walking away from NVIDIA. They’re stacking everything they can get, because demand is exploding faster than supply can exist.

That’s not a crack in the story. That’s the whole point.

In this week’s One Big Idea, we cut through the TPU noise and expose why this pullback isn’t a warning—it’s a rare moment of clarity in the most important infrastructure build of our lifetime. If you’re still thinking in headlines, you’re already behind.

⚡ Quick Hits

🎄 Global Week Ahead: December — Santa Rally or More Bah Humbug?

With markets heading into December, analysts are debating whether the seasonal “Santa Rally” will show up — or if macro headwinds and rate uncertainty will spoil the party. Historically, the Santa Claus Rally — the final five trading days of December through the first two of January — has delivered gains for major indices most years. For investors, that means a possible short-term pop could be in store, but only if economic and policy risks don’t derail the usual year-end lift.

🚀 This Stock-Split Stock Is Up 88,600% Since Its IPO

One firm that executed a stock split and skyrocketed nearly 88,600% from its IPO price shows how powerful compounding can be when strong fundamentals meet timing and execution. This kind of return isn’t luck; it’s built on years of aggressive growth and market dominance. For newer investors, it’s both an inspiring reminder of what “moonshots” can look like — and a warning that the biggest gains often come early in the journey.

💡 The Overlooked Account That Could Supercharge Your Retirement

The Health Savings Account (HSA) is emerging as a stealth retirement tool thanks to its triple-tax advantage: tax-deductible contributions, tax-free growth, and tax-free withdrawals for qualified expenses. Once reserved mainly for medical costs, HSAs are now viewed as a hidden weapon for long-term wealth building. For anyone serious about retirement planning, it’s worth exploring how an HSA can complement a 401(k) or IRA.

TOGETHER WITH OUR PARTNER

Run ads IRL with AdQuick

With AdQuick, you can now easily plan, deploy and measure campaigns just as easily as digital ads, making them a no-brainer to add to your team’s toolbox.

You can learn more at www.AdQuick.com

💡One Big Idea: The “TPU Panic” Is Actually NVDA’s Strongest Signal Yet

The TPU Panic Was a Mirage

Last week, headlines flared again: “Meta is scaling Google’s TPU v5 chips—bad news for NVIDIA.”

NVDA dipped toward $162, down from its October high of $178. Fear rippled through retail channels: “Competition is here. The monopoly is cracking.”

I felt that tug too. But then I looked at what Meta actually said—and what NVIDIA just reported.

In its November 20, 2025 earnings call, NVIDIA reported $35.2 billion in quarterly revenue—64% YoY growth—with Data Center alone at $32.1 billion.

Meanwhile, Meta confirmed it’s on track to spend over $100 billion in 2025 on AI infrastructure, and supply chain sources indicate $65–75 billion of that is flowing to NVIDIA for Blackwell systems.

And yet? “We’re still capacity constrained,” Meta’s CFO admitted.

That’s not weakness. That’s proof demand is outpacing supply by a historic margin—and no alternative changes that math.

📦 Stacking Chips, Not Switching Sides

Let’s be brutally clear: Meta isn’t replacing NVIDIA.

They’re layering every silicon source they can access—because their AI workloads are growing faster than even their $100B budget can satisfy.

Importantly, Meta’s TPU testing is limited to inference workloads for low-priority models—like recommendation engines or internal tooling. For frontier models (Llama 5, multimodal AI), they remain fully dependent on NVIDIA’s Hopper and Blackwell platforms.

That distinction matters: inference is commoditizable; training is strategic. And NVIDIA owns the training stack—end to end.

So they hedge at the edges… while signing multi-billion-dollar, multi-year deals for Blackwell.

This isn’t competition. It’s triage in a supply famine—and the market mistook diversification for defection.

⚙️ The Moat That Matters: CUDA Isn’t Going Anywhere

NVIDIA’s real edge isn’t Blackwell’s speed (though it’s unmatched). It’s CUDA—the software layer that underpins PyTorch, TensorFlow, Llama, and every major AI framework.

Google’s TPU is powerful—but closed. AMD’s MI300X is gaining traction—but lacks ecosystem depth.

Meta can test alternatives, but they can’t retrain Llama 5 or build custom tooling for every chip architecture. Rewiring their AI stack would cost billions and delay product launches by 12–18 months.

That’s why even Microsoft—deep in custom silicon—still runs 95%+ of its Azure AI training on NVIDIA.

CUDA isn’t just a toolkit. It’s the operating system of AI—and once you’re in, you don’t leave.

🏭 The Supply Crunch No One Can Solve

NVIDIA’s Blackwell ramp is historic—over 500,000 systems shipped or committed in 2025.

But global AI demand is even bigger:

• One major training cluster = 100,000+ GPUs

• TSMC’s CoWoS advanced packaging capacity = sold out through mid-2027

Blackwell isn’t just incremental—it’s 4x faster than Hopper on key LLM workloads, with 2x better energy efficiency. Microsoft recently called it “the backbone of Copilot+ and Azure AI.” Oracle is building 100,000+ GPU clusters exclusively on Blackwell.

This isn’t speculative demand—it’s deployed, paid-for infrastructure scaling through 2027.

Even with $72–76 billion in projected 2026 free cash flow, NVIDIA can’t magically create more supply.

So hyperscalers do the only rational thing: buy everything available—and test backups.

That’s not a threat to NVIDIA. It’s validation of its irreplaceability.

💰 The Real Math Behind the Headlines

Let’s cut through the noise:

NVDA Price: $162 (down 9% from highs)
2026 FCF Estimate: $74B
Forward P/FCF: ~44x
Forward P/E: ~24x (on $6.80 EPS)
Gross Margin: 76.5% (record high)

This isn’t the 2023 euphoria (80x P/E). It’s a high-quality compounder priced for strong—but not perfect—growth.

And with $18 billion in buybacks already executed in 2025, shareholders are getting paid to wait.

Compare that to the broader market:

• S&P 500 trades at 21x P/E

• But with <10% earnings growth

NVDA offers 60%+ top-line growth, 75%+ margins, and structural scarcity—at a modest premium.

That’s not a bubble. It’s a premium for dominance.

🧠 My Watchpoints: What Would Make Me Wrong?

I’m not ignoring risk. Here’s what would break this thesis:

1. CUDA erosion: If Meta, Microsoft, or OpenAI ship production-scale models on non-NVIDIA stacks by 2026, the moat cracks. So far, no sign.

2. Pricing pressure: If Blackwell discounts exceed 10%, margins fall. Current data shows ASP stability—hyperscalers are paying full price.

3. Export controls: New U.S. rules could limit sales to the Middle East or Southeast Asia—but those regions are <8% of Data Center revenue.

I monitor PyTorch GitHub commits, TSMC capacity reports, and hyperscaler procurement leaks weekly.

If the data shifts, I’ll shift with it. But today? The trend is accelerating—not stalling.

🎯 The Pragmatic Takeaway

NVDA at $162 isn’t “cheap.” But it’s not pricing in perfection either.

With 44x FCF, a 76.5% gross margin, and trillion-dollar customers still begging for more chips, this pullback looks less like a warning—and more like a rare moment of doubt in an enduring trend.

A move toward $150–155 would mark the most attractive entry since Q1 2025, aligning with ~40x 2026 FCF. But even at $162, the risk/reward favors holding or scaling in for long-term investors.

So while the crowd sells on TPU headlines, ask yourself:
If $100 billion in annual AI spend still leaves Meta short on chips… who’s really in trouble?

Spoiler: It’s not NVIDIA.

TOGETHER WITH OUR PARTNER

Pet insurance can help your dog (and your wallet)

Did you know 1 in 3 pets will need emergency treatment this year? Pet insurance helps cover those unexpected vet bills, so you can focus on care—not cost. View Money’s list of the Best Pet Insurance plans and protect your furry family member today.

Protect your pet and wallet

🧠 Final Thought

I keep coming back to this: in markets driven by headlines, real conviction looks like stillness. While everyone races to interpret the latest signal—TPU tests, chip shortages, stock dips—the companies building the future are doing so quietly, relentlessly, and without permission.

What feels like disruption is often just noise. What looks like risk is often just impatience. And the true edge? It’s not in predicting the next twist—it’s in staying anchored to what doesn’t change: scale, scarcity, and systems no one else can replicate.