Key insights
- The author argues that Nvidia's future profitability is overstated due to the underestimation of long-term support and warranty costs associated with its data center GPU sales. They draw a parallel to EVGA's exit from the GPU market due to unsustainable support costs. The author suggests that increased competition and rising support expenses could hinder Nvidia's ability to maintain its current revenue trajectory, potentially leading to a significant drop in valuation.
Pretext:
I grew up a fan boy of NVDA products, and as such, became very familiar with the branding of their cards. One such brand was EVGA.
EVGA offered the same level of guarantees that NVDA itself is offering alongside these monster data center deals. Warranties, replacements, the whole package.
EVGA went out of buisness very quickly, despite its upstart into the most popular brand name for NVDA GPU's. As it turns out, offering this type of support for high value chips is hardly feasable.
Large scale context:
As sales ramp up, so does the capacity to support those sales. Replacements will be warrantied, along with a need to scale up capacity for repairs.
We are still in the celebratory period - sales have closed, with nearly a trillion worth of obligations to fullfill in the future. Nobody can predict if or when there will be widespread or isolated catastrophic failures as this ramps up.
Expense of the future:
It will take a whole year of data for us to understand the true cost of supporting these operations. Its worth stressing that none of these are open and shut deals - they all come with warranties, guarantees, on-site support ect. The big green numbers we see make it look like the greatest company to ever exist yet gloss over longer term implication of near term profits.
5 million earned per employee drops significantly as support teams double, tripple, ect in size. Travel expense, tooling, training - and at a time when blue collar is more expensive than ever. If 2025-2026 are the years of sales thanks to first-to-market advantage, then 2027-2035 are the expense years.
In other words, these huge figures will be covering the next decade of expense in support inferastructure. And it must - the first to market advantage has come and gone, competitors can make bids against nvda offering the same product or better.
End:
If nvda cant mantain record sales, let alone increase them as support infastructure expenses rise, then youre looking at a 4.5 trillion dollar company who may not even turn a net profit in 2028, so youre betting those people will think "profit isnt there becouse of those amazing sales 2 years ago, no big deal"
Yeah. Good luck with that.