Words: 1,925 Time: 8 Minutes
- Watching hyper-scaler capex growth and relative depreciation
- Notable investors sold large cap tech – others are buying
- S&P 500 continues to lose momentum
All eyes will be on the world’s most valuable company – Nvidia – when they report earnings later this week.
It goes without saying they will handily beat all revenue and earnings estimates.
However, what will their CEO – Jensen Huang – say about the expected AI investment boom over the next few years?
For example, Huang revealed in October that his company has $500 billion in orders, in 2025 and 2026 combined, for its chips that are at the heart of the AI boom (more on these figures soon when I show hyper scaler investments in Net PPE (Property Plant and Equipment)).
For a company that has seen its quarterly revenue grow nearly 600% over the past four years, Huang will be telling us that the AI boom still has room to run.
And that’s what investors hope to see…
But it raises a couple of big questions – especially for (some) investors who are paying sky-high multiples:
- What is the impact to free cash flow?
- What will be the hit to earnings and margins?
- And how much depreciation is currently being factored in?
After all, graphic process units (GPUs) have a limited life (like any form of computing hardware).
Generally that life is around 3-years (maybe 4 years at most).
But today – most of the Mag 7 are using 5 to 6-year deprecation schedules for their investments in these AI chips.
Why?
Simple: depreciation is an expense. If we declare a lower value – it will inflate your earnings.
Let’s take a look.
AI’s Unanswered Question
In accounting, depreciation is the act of allocating the cost of a hard asset over the course of its expected useful life.
Lets say you have a laptop or phone which you use to run your business.
The tax office will allow you to depreciate the cost of that hardware over 3 years.
For most people, after about 3 years that piece of hardware will need replacing.
This is the same for companies today spending “hundreds of billions” on AI chips (GPUs) and infrastructure.
However, there is a debate as to whether these chips should be written off after 3, 4, 5 or 6 years?
Using back of the envelope math – large cap tech has more than $500B to $750B of depreciable GPU (AI) chips on their books.
A Reluctance to Depreciate?
Just last month – most of the Mag 7 told us they were significantly increasing their capital expenditure spending on artificial intelligence infrastructure.
To help this discussion – consider the following figures (which I have derived from their quarterly filings from the past 10 years).
| Ticker | Net PPE 2015 | Net PPE TTM 2025 |
10Y CAGR | Avg Dep / PPE 2015–2024 |
TTM Dep / PPE | Under-depreciation Assessment |
|---|---|---|---|---|---|---|
| AAPL | $22.5B | $61.0B | 10.5% | 3.4% | 5.1% | Slight, improving |
| GOOGL | $29.0B | $238.0B | 23.5% | 3.2% | 2.6% | High |
| MSFT | $14.7B | $229.0B | 31.1% | 3.6% | 4.8% | Moderate, improving |
| META | $5.7B | $177.0B | 37.3% | 4.1% | 2.8% | High |
Three things to note:
- The relative under investment in Net PPE from Apple – with a Net PPE CAGR of just 10.5% and ‘respectable’ deprecation;
- The large CAGR in Net PPE over the past 10 years – in excess of 31% for each of MSFT and META;
- The modest levels of depreciation for GOOGL and META over the trailing twelve months.
Let’s now look at each of the respective last four quarters for any trends:
| QTR | Ticker. | Net PPE TTM. |
Avg Dep / PPE 2015–2024. |
TTM Dep / PPE |
Under Depreciation |
|---|---|---|---|---|---|
| LQ | AAPL | $61.0B | 3.4% | 5.1% | Slight, improving |
| LQ-1 | AAPL | $60.2B | 3.4% | 5.0% | Slight, improving |
| LQ-2 | AAPL | $59.0B | 3.4% | 4.9% | Slight, improving |
| LQ-3 | AAPL | $58.0B | 3.4% | 4.8% | Slight, improving |
| LQ | GOOGL | $238.0B | 3.2% | 2.6% | High |
| LQ-1 | GOOGL | $234.0B | 3.2% | 2.5% | High |
| LQ-2 | GOOGL | $230.0B | 3.2% | 2.4% | High |
| LQ-3 | GOOGL | $226.0B | 3.2% | 2.5% | High |
| LQ | MSFT | $229.0B | 3.6% | 4.8% | Moderate, improving |
| LQ-1 | MSFT | $225.0B | 3.6% | 4.7% | Moderate, improving |
| LQ-2 | MSFT | $220.0B | 3.6% | 4.6% | Moderate, improving |
| LQ-3 | MSFT | $218.0B | 3.6% | 4.5% | Moderate, improving |
| LQ | META | $177.0B | 4.1% | 2.8% | High |
| LQ-1 | META | $175.0B | 4.1% | 2.7% | High |
| LQ-2 | META | $172.0B | 4.1% | 2.6% | High |
| LQ-3 | META | $170.0B | 4.1% | 2.5% | High |
Hyperscalers’ rising capex (and Net PPE) reflects the “insatiable AI appetite” that Nvidia’s CEO keeps referring to.
His company has benefited more than most.
And whilst I don’t question the relative levels of demand – I do question:
- Whether the same (highly profitable) monetization models will be there with AI (as we’ve seen the past decade); and
- What impact could see with depreciation expenses if they are not?
This is important for both investors and lenders financing the giant AI buildouts.
For example, the longer equipment remains valuable, the more years a company can stretch out depreciation and the less it hurts profits.
The opposite holds true.
Now in fairness to large cap tech who are hesitant to depreciate their investments in GPUs – they are still relatively new to the market.
Nvidia’s first AI-focused processors for data centers came out around 2018.
The current AI boom started with the launch of ChatGPT in late 2022.
Since then, Nvidia’s annual data center revenue has jumped from $15 billion to $115 billion in the fiscal year that ended in January (a CAGR of ~97%)
You are likely to hear various opinions regarding how long AI chips will retain their value… it could be anywhere from 2 years to maybe 6?
For example, the company CoreWeave – which buys GPUs and rents them out to clients – has used six-year depreciation cycles for its infrastructure since 2023.
Their CEO – Michael Intrator – told CNBC that his company is being “data driven” about GPU shelf life.
From mine, and based on the past 50 years or computing history, I think a life of anything more than 3 years is ambitious.
And if that assumption is reasonable – then these companies are effective “on the clock” to start showing (very) strong returns on their investments.
So far, it’s not enough to justify the current multiples being asked.
Tread Carefully
From mine, the valuations (and expectations) on the Mag 7 are excessive.
The only stock I was willing to add to earlier this year (in Q2) was GOOG at a price of around $155 – where the valuation was reasonable around 16x forward earnings. That felt like a good risk reward bet for the long term (so far so good)
However, today that stock is more inline with its peers.
Consider the following 10-year average and real-time multiples for these popular names:
| Ticker. | P/FCF 10Y Avg |
P/FCF Real-Time. |
P/S 10Y Avg. |
P/S Real-Time. |
EV/EBIT 10Y Avg. |
EV/EBIT Real-Time. |
|---|---|---|---|---|---|---|
| AAPL | 20.5 | 41.3 | 5.3 | 9.8 | 19.9 | 31.2 |
| GOOG | 28.0 | 45.8 | 6.1 | 8.75 | 24.1 | 27.2 |
| MSFT | 28.8 | 52.2 | 8.8 | 13.28 | 25.5 | 29.3 |
| AMZN | 26.7 | 240.1 | 3.1 | 3.67 | 63.4 | 34.1 |
| META | 28.2 | 34.2 | 9.2 | 8.1 | 23.4 | 18.9 |
| TSLA | 82.5 | 189.8 | 8.9 | 13.56 | 158.5 | 285.3 |
As an aside, I maintain this level of valuation data (and history) for every S&P 500, Nasdaq and Russell 2000 company.
It’s a matrix of (a) quality (~30 metrics) vs (b) value (such as those above)
This table gives you a sense of the 10-year average valuations for these high quality stocks vs the multiple you’re paying today for “AI earnings growth”
Specifically, look at the free cash flow (FCF) multiples.
It shows how their existing cash flows are being hammered by the investments made in these chips and data centers (arguably resulting in large staff layoffs elsewhere).
And whilst the EV/EBIT ratios appear modest in some cases (e.g., GOOG trading about 3x higher than its 10-year average) or AMZN half its average)… it’s the free cash flow I’m watching closely (especially over the next 12 months).
But as I say – the “trillion” dollar bets being made by these players in AI may well pay off.
However, we will not know for maybe 3-to-5 years (likely case).
For my money – I’m happy waiting for a more attractive entry points (like we saw with GOOG during Q2)
Stocks Lose Momentum
Before I go – a quick look at the loss of momentum the market is experiencing.
For example, I’ve had a few people saying “what do I think about this dip” – asking me if it’s worth buying?
My answer… we’re yet to see any dip (as our weekly chart shows)
Nov 18 2025
As part of this post three weeks ago – I flagged the potential resistance around 7000 for the S&P 500.
Since then, the market has not made any ground – trading at the same levels we saw during the week of Sept 15th.
From a fundamental lens – investors are right to question valuations and earnings growth.
However, from a technical lens, we can see the strong negative divergence signal with the weekly RSI.
In short, as the price made new highs, the momentum indicator did not. Put another way, the new highs were not confirmed.
We saw something very similar in the lead up to the 20% correction in 2022.
And whilst stocks are slightly off the highs this year… I will get interested when we see the index correct to the tune of around 10 to 15%.
Putting It All Together
Later this week I will share some of the notable moves with leading fund managers and their 13F filings.
For example, I saw that Peter Thiel sold its entire stake in Nvidia (smart)
The other day I mentioned that hedge fund manager Michael Burry of The Big Short fame – disclose bearish wagers against both Nvidia and AI-powered software developer Palantir.
However, there was one exception.
Berkshire Hathaway disclosed a new stake in Google.
Berkshire purchased just over 17M units during Q3 – where GOOGL now represents ~4% of its total portfolio.
Here’s are all the movements for Berkshire based on their latest 13F (Note: I track ~50 funds back over 15 years)
Someone wrote to me and said “Wow!! Buffett bought ~17M units of Google at this valuation – it must be good value’ (aware that it’s my largest holding in my portfolio at around 12% weighting). Two things to note:
1. The purchase was made b/w July 1 and Sept 30 – where GOOG’s price ranged b/w $174 and $256 per share. That is ~40% and 12% cheaper than today’s asking price of around $290; and
2. We don’t know if it was Buffett buying. Buffett allows his underlings – Ted Weschler and Todd Combs – to buy stocks without his involvement. From mine, it’s more likely to be either Weschler or Combs.
If you’re trying to follow Berkshire into GOOG at the current price of ~$290 per share – you’re paying a large premium.
What will be interesting is whether Berkshire added to the position in Q4 at these multiples. We will learn about that Feb 15 2026.
Note: Bill Ackman’s Pershing Square reduced his stake in GOOG by ~10% last quarter (whilst not making any new stock purchases)
Now if GOOG were to correct in the realm of ~20% – yes – consider making it around 4% of your portfolio.
My general advice here – it pays to remain prudent.
Valuations matter.