Market On Open, Friday 5 September
DISCLAIMER: This note is intended for US recipients only and, in particular, is not directed at, nor intended to be relied upon by any UK recipients. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Cestrian Capital Research, Inc., its employees, agents or affiliates, including the author of this note, or related persons, may have a position in any stocks, security, or financial instrument referenced in this note. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice. Companies referenced in this note or their employees or affiliates may be customers of Cestrian Capital Research, Inc. Cestrian Capital Research, Inc. values both its independence and transparency and does not believe that this presents a material potential conflict of interest or impacts the content of its research or publications.
The Beginning Of The End
by Alex King, CEO, Cestrian Capital Research, Inc
For the avoidance of doubt:
- I am long $NVDA personally, and
- I believe the stock has meaningful upside from here.
But I believe the beginning of the end of Nvidia’s dominance is underway; shareholders and would-be shareholders should factor this into their investment decisions.
Who, you might ask, is responsible for upending the apple cart?
Is it $AMD, with cousin-nemesis Lisa Su at the wheel? It is not.
Is it $INTC, freshly armed with federal government backing? It is not.
It’s a pincer movement by two constituencies.
- Customers, who have long eyed Nvidia’s huge gross- and cashflow margins with annoyance (yes, they all sing Mr. Huang’s praises in public but I can tell you there is cursing in private), and;
- Power grid participants (meaning utility providers, state and federal planning authorities, and so forth).
If you run AI datacenters on GPU architecture - architecure originally designed for graphical functions in a world of abundant power vs. the compute tasks at hand - then you very quickly run into a host of problems arising from that core design. Everything runs hot, meaning everything has to be cooled. The compute plant eats power, because the calculation methods are brute-force in nature, and the cooling plant eats power, because the compute plant is wildly heat-inefficient and spends most of its day trying to demonstrate the various laws of thermodynamics. This is great news if your datacenter is near the Arctic Circle but if it’s in Ohio, New York State, Germany or any other part of the populated world, not so much.
There isn’t enough power that can be provisioned quickly enough in the right places to meet the demand for compute cycles. So something has to give. Either the demand goes unserved (unlikely) or a lower-power architecture has to be brought to bear.
The two silicon providers already in place to deliver this are $ARM, which is low power by design - it’s why there is a >90% chance that the cellphone in your pocket runs on ARM cores - and $AVGO, which has multiple application-specific type strategies available to hyperscalers already.
ARM and AVGO won’t kill NVDA tomorrow, next week or next year, and NVDA won’t fail. But v1.0 of the AI datacenter is now in decline, and v2.0 (where power management is a primary consideration driving everything from geographic placement to the applications processed in the datacenter to the silicon used to process them) is now upon us.
Don't be blindsided by this. Be ready.
Anyway, let’s get to work. As usual we cover equity indices, key sectors, bonds, volatility, crypto, oil and more below.