The thought just passed through my head that I should just leave the rest of this post blank. That would underline what I actually know about DeepSeek, a Chinese-made AI construct that uses an earlier, cheaper generation of Nvidia chips and needs less electric power.

…but no!

Wall Street’s reaction to the revelation of the existence of DeepSeek over the weekend was to lop around 20% off the prices of the leading AI-related stocks.

That’s all in the rear view mirror now. The important question for us as investors, and for me as a holder of NVDA and AVGO, is where to from here.

I’ve had a big position in NVDA for a long time, thanks to my younger son beating me over the head with the idea. Over the past six months, I’d switched the bulk of that into AVGO. Two reasons: the position size, and my observation that overall employee + SG&A costs had become so small for NVDA in relation to sales that operating leverage had virtually disappeared. Therefore, for a while, profits would likely grow in line with sales but no longer significantly faster. My guess is has been that this deceleration has been unnoticed and would trigger a negative reaction once Wall Street worked this out.

My thoughts now:

–what has attracted me to NVDA is that the company seems to me to have the true growth company’s ability to reinvent itself. This is a rare quality. Typically, visionaries are gradually replaced by bureaucrats and the world begins to pass a firm by. INTC is a classic example. But NVDA has been the graphics card company, then the crypto mining company, then the AI company …and it is already laying plans for its next transformation. My main issue with NVDA has been valuation.

–we don’t know how much of the DeepSeek story is true and complete, although my guess is that at least the part about Chinese coders doing more with less is correct. I think of it as like racing car builders only having access to last year’s top of the line engine. But better tires, a more aerodynamic frame and faster gear=shifting might still let them still produce a faster car. The issue here may be how far better software can go in replacing better hardware. I choose to think that the NVDA chips in DeepSeek do maybe 90% of what the latest models can do–which fits with the timing of the Biden export ban. My guess is also that costs have been understated

–over the past year NVDA has doubled, even after yesterday’s decline, vs. the S&P up 25% and NASDAQ up 27%. So the stock was arguably overdue for profit takers to emerge. To the degree that this is correct, DeepSeek is the trigger, not the cause

Overall, I’m happy to hold what I have of both names. My guess, though, is that this whole area will tread water for a time.