A little over a month ago, I wrote that the worst was not yet behind nVidia. UBS was second in line, and they based their decision on competition and growth worries. My basis for the need to hold-off on NVDA’s stock was based on both research and intuition. For starters, they had a manufacturing issue that cost them almost $200 Million. A mistake of this magnitude usually takes at least a couple of quarters to correct – whether it is a chip redesign or a retooling in the manufacturing process. Since it was a manufacturing issue that involved chip packaging [thermal issues] – which the company thinks has solved using both software and hardware solutions [by working with the laptop manufacturers to switch on the cooling fans early and often, and by fixing their chip packaging issues].
NVDA is well capitalized, their employees work hard, but their employee morale is at an all-time low [mostly due to the stock]. Yet, the company soldiers on, and wants Operating Systems to target the GPU for computing power. But with Intel proliferating the number of cores per CPU and with the latest OS’s in development becoming smarter at parcelling out different tasks to different cores.
Defenders of the GPU faith look at sheer MFLOP numbers and claim that the GPU could crank out complex jobs 20 to 40x faster than multiple CPU’s, but in a not so recent test [protein-folding simulation at Stanford], an ATI Radeon had about a 50% advantage over a dual-Opteron configuration. Anyway, I expect this technology [Stream Computing] to be at first be used by the high-end gaming software developers [though there are serious issue - as GPU's are not designed to share a common pool of memory [unlike CPU's], so NVDA and AMD use ad-hoc methods in their drivers to track each GPU on the bus.
NVDA is expected to announce revenues of $900 Million [range is $887 M to $913 M] and profits of 13c/share [range is 12 to 14c/share]. My current estimate is between $850 to $875M for revenues, and net income in the 8c range [I will revise these numbers based on channel checks before earnings].
Despite their troubles, NVDA has $3/share in cash and no debt. At a market value of 1.3x sales, and with the stock at a 75% haircut from its peak, it looks attractive on a valuation basis, but I am not biting yet. AMD’s finances [or the tightness of AMD's finances rather] is good for NVDA.
1. $200 Million manufacturing problems take several quarters to resolve in the semiconductor business.
2. Channel checks indicate revenues below the low end of analysts’ estimates. My estimates for EPS is 25% lower than the lowest current estimate. [I will revise these numbers in a month].
3. Stream Computing is unlikely to become integrated into OS’s in the next year, though targeted applications will use GPU’s tremendous computing power.
4. I like NVDA – the stock looks cheap, but I am recommending a “wait and see” – as it could get cheaper still.
1. Analysts’ estimates from Zacks.
2. Some numbers extracted from the latest 10Q of NVDA’s.
Disclosures: No positions in NVDA, AMD.