,
25 tweets,
6 min read
Read on Twitter
1) $NVDA – a few thoughts ahead of analyst day around the RTX launch and my belief that the VC funded deep learning startups are not a significant threat to $NVDA’s franchise. Semiconductor VC is akin to biotech VC and the $1.25b invested just isn’t enough to be relevant.
2) This chart is critical to understanding the RTX launch. Frames per second drives a higher K/D b/c you see your opponent before they see you. Gamers who play shooters pay up for faster FPS in the same way golfers pay up for better clubs (don’t play but allegedly big advances) 
3) RTX gives less of a frames per second uplift than prior generations (35% vs. 70% previously) and DLSS has been disappointing so far (as pointed out by my fav journalist focused on tech stocks) in terms of mitigating this.
4) DLSS will steadily improve, but today RTX is less of a “value” for shooter focused gamers in terms of FPS uplift per $. Recent price reductions on the RTX 2080 ($650) mean that the “value” has improved but this chart still directionally correct.
5) However, frames per second are irrelevant above 60 FPS (movies run at 24 FPS) for non-shooter games. i.e. For games like Skyrim, Fallout, Assassins Creed, Red Dead Redemption, GTA, etc. And all of these games/their successors will eventually support ray tracing.
6) Defining GPU performance/value solely as frames per second is *absurd* when it comes to games ex shooters. Ray Tracing is the biggest advance in the visual beauty of graphics since programmable shaders. For gamers who don’t primarily play shooters, RTX is incredible value.
7) And for shooter focused gamers, $NVDA is still the only game in town. The RTX 2080 outperforms its $AMD comparable Vega VII despite being a node behind and devoting valuable silicon real estate to Ray Tracing. This is incredible engineering by $NVDA.
8) The Vega VII should outperform the RTX 2080 by at least 20-30% on a FPS basis in non ray-traced rendering given it’s a node ahead and all of its silicon is focused on traditional GPU tech. Suggests $NVDA has a 25 to 35% architecture advantage over $AMD.
9) These dynamics - being great value for non-shooter gamers and being the best GPU for shooters (and maybe less crypto exposure) - are why $NVDA share went up in Q4 despite the "disappointing" RTX. And $NVDA gets its best chance ever to create a proprietary graphics standard.
10) Semi customers want buffer inventories to equal lead times. When lead times go up, inventories go up and vice versa. These inventory cycles are what drive semi cyclicality today – not capacity cycles. Right now $NVDA is underearning as they are undershipping end demand.
11) GPU lead times went up in 2018 b/c of crypto, buffer inventories went up to match these lead times. $NVDA thus overshipped end demand. When crypto demand evaporated that cycle unwound. $NVDA will continue undershipping/underearning until inventories are worked down.
12) $NVDA ASPs are up significantly in last 4 years. This pricing power has been an important part of the story. I think the real LT risk to the gaming business is that the ROI on crypto mining may have driven much of $NVDA’s recent pricing power.
13) i.e. Gamers could afford to buy a better card because they could mine at night, in the winter, etc. when it was economical. (Winter being important b/c can turn off heat and rely on mining rig to heat house). Time will tell - and crypto could always recover - $FB stablecoin!
14) A lot is written about the threat to $NVDA from deep learning startups and the $1.25b that has gone into funding them. $1.25b is nothing at the leading edge of semiconductors. This is very different from SaaS, consumer internet and more like biotech.
15) The cost to design a leading edge semiconductor is $175m at 10 nanometer, $300m at 7 nanometer and $540m at 5 nanometer (roughly speaking). Mask costs alone are tens of millions. And you have to spend that steadily escalating amount every 2 years like clockwork.
16) $300-$500m is about the amount of money it takes develop a biotech drug. Tapeouts are much more predictable than they were (thank you EDA industry), but still a risk. The tapeout is like the phase 2 readout. If any of these co’s miss on their tapeout, it is over.
17) And if they have a great tapeout, they will likely only support 1-2 frameworks and a few algorithms. $NVDA supports all deep learning frameworks, all deep learning algorithms and can be used for traditional ML and HPC. Higher utilization b/c of this flexibility = lower cost
18) Supporting all of these frameworks is hard and expensive – witness how far behind $AMD still is from a GPU software perspective. $NVDA generally made better GPUs than $AMD, but their SW drivers were their real secret sauce in graphics – both more stable and more optimized.
19) i.e. $NVDA beat $AMD in graphics partially due to better software engineering. Software is a core competence. Mental model here is AI frameworks = game engines and AI algorithms = games. They can all be accelerated by recompiling code more efficiently.
20) This treadmill is very hard to get on and even harder to stay on. “Tick, tock” sounds easy but it is really difficult. And spending $500m plus every other year – just to develop the chip – let alone support it, market it, etc. is quite different than other tech industries.
21) And once successfully on this treadmill, the startups also have to compete with $GOOGL’s TPU and other cloud co ASICs. Given $NVDA’s gross margins and the gross margins the startups have to target, these inhouse ASICs can be much worse tech wise but still make economic sense
22) i.e. $NVDA really outperforms TPUs from a performance per watt perspective but it is still economic for $GOOGL to use TPUs in some cases given performance per watt per dollar (until the increased opex overwhelms the capex savings).
23) Lots of smart AI people say encouraging things about the AI chip startups. I would do the same! $NVDA having a semi-monopoly isn’t good for them. And some of the startups will succeed in specific use cases and take some share, but I don’t see major shifts in next 5 yrs.
24) And of the startups, my money would be on Cerebras to succeed in training and a top secret co. in inference (I’m under NDA). For the ones that fail, I think will be tougher to sell to $INTC given recent great engineering hires and more disciplined CEO (former CFO)
25) And all this is before $NVDA and $MLNX co-develop the next generation of interconnect and GPUs together. $MLNX was both accretive and strategic. Hearing more about the combined product roadmap is what I am most looking forward to at the analyst day.
