Can $NVDA take less financing risk without losing profitable chip sales?
In this week's Equity Letter, I examine the business, valuation scenarios and what could change my investment case.
I also explain why I opened a position.
Read the free letter →
https://t.co/Jua3KsknY4
Gold is supposed to protect you in a crash.
This week it fell more than the S&P 500.
$SPY lost 8.6%, its 9th worst week since 1993.
$GLD lost 9.2%.
On Friday, bond yields fell and the dollar was flat. Neither gives gold a reason to drop. It still fell 3.7%.
That's forced selling. When investors need cash fast, they sell what they can. Gold is easy to sell.
Since 2004, gold has dropped 9% in a crash week only twice before: September 2011 and March 2020.
Both times, the S&P hit its low within 6 trading days. In 2020 it fell another 17% first.
A year later it was up 28% and 47%.
I'm holding my positions.
@saxena_puru Agree that we are nowhere near 2000 levels, but a lower P/E doesn't eliminate the risk of an earnings downturn.
Back then telecom equipment supplier had booming earnings until overcapacity forced customers to cut spending.
One thing worth considering is how much the S&P 500 Info Tech index has changed since 2020.
Visa, Mastercard and Paypal were removed in 2023 while NVDA's weight increased significantly.
That makes comparing today's aggregate forward P/E with historical levels less meaningful.
Doesn't mean tech isn't cheap, just not an apples to apples comparison.
@FrostyFinances Agree that appointing Collette is reassuring, but I wouldn't rule out accounting issues just because the CFO became interim CEO.
An SEC investigation is still a long way from proving fraud.
Yesterday's AMD story suggests that the issue might be much broader and not just specific to NVIDIA.
The problem is that even if chipmakers fully comply with export regulations, China might continue finding ways to access restricted chips through various channels.
Companies could end up paying the price for something they never did wrong.
@LongGameEquity Right now valuations have become meaningless.
The market just doesn't care how cheap something gets.
Everything gets sold indiscriminately.
At some point fundamentals will matter again but the question is how much damage gets done before then.
@DimitryNakhla The question is how much of that competitive advantage translates into value for shareholders.
A company can be essential to an entire industry while its customers capture most of the upside.
Being a great business and great investment isn't the same thing.
@DeepValueBagger Right now it feels like every negative headline is basically interpreted as confirmation of the worst-case scenario.
The market has become completely irrational.
@oguzerkan The problem happens when investors combine a long term investment horizon with short term leverage.
You can have all the conviction and emotional control in the world, but if your broker forces you to sell it doesn't matter anymore.
@KevInvestingYT There are legitimate risks but you're jumping from an SEC investigation, to artificially inflated demand to bankruptcies and a global recession.
A lot of assumptions being treated as inevitable.
@Mapago9 Memory supply for 2027:
Samsung: 2027 HBM sold out.
Micron: >75% of output committed.
Sandisk: ~50% of FY2027 production committed.
The market: “The memory cycle is over.”
2026 is the year of the AI buildout yet the best sector in the stock market is oil.
YTD total returns through Thursday:
• Energy sector (XLE) -> +39%
• Tech (XLK) -> +12%
Energy has been ahead on 197 of 223 trading days.
The easy explanation is that AI needs energy. But the energy sector is oil and gas companies, not power plants.
Its best stocks are oil refiners like Marathon Petroleum, up 130%.
That's geopolitics, not AI. Oil and fuel got scarce after the Strait of Hormuz disruptions and the damage to Russian refineries.
The part of energy that is connected to the AI trade is natural gas. It supplies more than 40% of the electricity US data centers use, based on IEA estimates.
The sector's two gas-focused producers are both down this year:
• EQT, -7%
• Expand Energy, -23%.
Two power producers with data center contracts from Meta, Amazon or Microsoft did no better. I ranked the S&P 500's stocks by this year's return.
Both are in the bottom tenth:
• Vistra $VST, down 23%
• Constellation $CEG, down 28%
On January 9 Vistra announced 20-year contracts to sell nuclear power to Meta. The stock hit its high for the year six days later. At Thursday's close it was more than 17% below its price the day before the deal.
Meta's contracts cover about 6% of Vistra's power plant capacity. Almost half of it is in Texas.
My read: Texas explains the fall. Power prices there are low, and the state froze data center approvals.
What I learn from this: 2026 is the year of the AI buildout, but the biggest story of the year doesn't decide where the money is made.
True that the backlog makes the next couple of quarters relatively safe.
But that's only a small fraction of NVDA's valuation, which reflects a much longer period of discounted cash flows.
I think that the current price action is driven much more by uncertainty about the long term perspective than by concerns over the next quarters.
Still believe that it's going to be fine in the long run, just risks worth keeping in mind.
@jrouldz I wonder if we're starting to see the next crowded trade forming.
Investors fleeing from AI stocks in supposedly "safe stocks" creating the next bubble.