A couple of weeks ago, NVIDIA reported one of the best quarters you’ll ever see from a company at this scale: $57 billion in revenue, up 62% year-over-year, crushing analyst expectations. Jensen Huang said Blackwell demand is “off the charts” with massive orders lined up.

The stock jumped 5% after hours... then dropped 3.2% the next day. Three weeks later, it’s down about 10% from its recent peak near $200. Today, NVIDIA trades around $182, well off its 52-week high of $212. Is the AI spending cycle that’s driving NVIDIA’s valuation sustainable, or are we in the late stages of a capex supercycle that’s about to slow down?

And here’s where Palantir comes in. Because Palantir’s UK NHS contract—a £330 million disaster where fewer than a third of hospitals are actually using the software—is the clearest preview we have of what happens when AI spending meets economic reality.

Let’s dig in.

Why the Market Didn’t Care About Great Earnings

NVIDIA’s valuation isn’t the problem. At ~25x forward earnings for 50%+ growth, it’s actually reasonable by historical tech standards. The dot-com bubble had companies trading at 40-45x forward P/E with no earnings. NVIDIA is growing 62% at scale with 73% gross margins.

But the stock dropped anyway because of two main concerns:

The Circular Spending Problem

The day before earnings, NVIDIA announced a $10 billion investment into Anthropic (alongside Microsoft’s $5 billion). Anthropic then committed to buying $30 billion in compute from Microsoft Azure... which runs on NVIDIA chips.

NVIDIA → Anthropic → Microsoft → NVIDIA.

It’s not a perfect circle, but it’s close enough to make investors nervous. How much of NVIDIA’s “demand” is real customer need versus financial engineering to create the appearance of demand?

The Hyperscaler Spending Cliff

Microsoft, Google, Amazon, and Meta are collectively planning massive AI infrastructure investments through 2028. But when do they start making money from this?

Media reports suggest that Big Tech is expected to generate enough AI revenue to cover only a fraction of their AI spending through 2028. These are mature companies with shareholders who eventually want ROI.

If in 2026-2027 cloud providers say “okay, we’ve built enough infrastructure, time to monetize,” what happens to NVIDIA’s 60% growth rates?

The bull case assumes spending accelerates indefinitely. History suggests every capex cycle eventually ends.

The Preview of What Comes Next

And this is why Palantir is so instructive.

Quick Palantir metrics (as of December 8, 2025):

• Stock price: ~$182 (up over 140% YTD)

• Market cap: ~$407 billion

• Revenue (Q3 2025): $1.18 billion (+63% YoY)

• 2025 guidance: $4.4 billion

• Trailing P/E: ~396x

• Forward P/E: ~185x

• US Commercial growth: +121% YoY

Palantir is THE defense AI company. They’ve got the major Army contracts, AI targeting systems, Space Force work, all the classified projects you can imagine.

The NHS Disaster: What Happens When There’s No Mandate

Palantir won a £330 million, 7-year contract to build a “Federated Data Platform” for NHS England in 2023. This was supposed to be the blueprint for healthcare AI.

What actually happened (as of mid-2025):

• Fewer than one-third of hospital trusts are actively using it

• NHS staff say their existing tools already exceed what Palantir offers

• The UK government had to hire KPMG for an additional £8 million just to “promote adoption”

• Critics are calling it underwhelming for £330 million

• 417 out of 586 pages of the contract were redacted

This is Palantir without a security clearance forcing people to use it. When users actually have a choice and can compare it objectively to alternatives, adoption stalls hard.

And this terrifies me about the broader AI narrative.

What If Most AI Spending Is Like the NHS Contract?

Think about it: What if a lot of AI spending is happening because:

• CEOs feel FOMO about not having an “AI strategy”

• Consultants are selling expensive AI transformations

• Nobody wants to be the one who “missed AI”

• Companies are locked in by government mandates or incumbent relationships

But when you actually measure “Did this deliver ROI better than what we already had?” the answer is often: Not really?

The NHS experience suggests that in commercial markets where customers can objectively assess value, Palantir struggles. Its revenue growth is real, but it’s concentrated in government/defense where switching costs and security clearances create moats.

US Commercial grew 121% YoY—impressive! But it’s off a smaller base. The real test is Year 2-3 retention when CFOs ask: “We paid $2M for this. What exactly did we get that we couldn’t have built ourselves for $500K?”

What This Means for Investors

On NVIDIA: Fairly valued at current levels if growth moderates to 25-35% over 2-3 years as the capex cycle matures. Not a short, but not a conviction buy either. The risk is asymmetric: 25% upside if spending continues, 40% downside if it slows.

On Palantir: Overvalued at $182. Fair value closer to $100-120 based on defensible government revenue. The NHS contract shows what happens in commercial markets without mandates—and it’s not pretty. Microsoft’s Azure bundle strategy is an existential threat within 18-24 months.

On the bubble question: This isn’t 1999. The technology is real, the profits are real, the moats are real. But AI infrastructure spending is elevated relative to current monetization, and every capex cycle eventually slows.

My base case: Strong 2025, digestion in late 2026/2027 as hyperscalers shift from buildout to monetization, then normalization. NVIDIA grows 15-25% instead of 50%+, multiple compresses from 25x to 18-20x. Stock stays flat to down 20-30% over next 18 months despite earnings growth.

If the economy weakens (and I think it will—the K-shaped recovery can’t continue forever), this timeline accelerates.

I’m not betting against AI. I’m just betting against the assumption that this level of spending is sustainable indefinitely.

VL

Disclaimer: This is not financial advice. These are my personal opinions based on public information. Do your own research before making any investment decisions.