Nvidia’s Drive PX Pegasus Is A Royal Flush For Self-Driving Cars – NVIDIA Corporation (NASDAQ:NVDA)

The Auto segment of Nvidia (NVDA) is the smallest contributor in terms of quarterly revenue. It only contributed $142 million in Q2 FY18. The Auto segment can probably see a big spurt in growth when Nvidia starts shipping its Drive PX Pegasus next year. Drive PX Pegasus is Nvidia’s ambitious license-plate-sized in-vehicle datacenter-level processor intended for level 5 (or fully autonomous) robotaxis. There are now 25++ companies developing Nvidia Drive PX-based robotaxis.(Source: NVIDIA)

Anything that can help increase the quarterly revenue of Nvidia’s Auto business unit is an important matter that should be discussed here. To gauge just how important self-driving car technology-related products like the Drive PX Pegasus, read on Google’s (GOOG) (NASDAQ:GOOGL) billion-dollar legal suit against Uber allegedly receiving Waymo trade secrets.

The Drive PX Pegasus is 10x more powerful than the Drive PX 2 hardware. It can reportedly match the compute power of a 100-server datacenter rack. I consider this new hardware from Nvidia as the super-sized version of Intel’s (INTC) neuromorphic processor.

A little tweaking from Nvidia and Drive PX Pegasus can also self-learn like a human brain. Simultaneously, it can still deliver over 320 trillion operations per second compute performance without being tethered to the cloud.(Source: NVIDIA)

The Drive PX Pegasus could catapult Nvidia as the go-to processor supplier of the $38 billion/year ride-hailing industry’s adoption of robotaxis. Goldman Sachs also expects the ride-hailing industry to grow to $285 billion by 2030.

The Drive PX 2 Pegasus level 5 autonomous car processor can make Nvidia the enabler of some unicorn companies. The massive valuation of ride-sharing firms/taxi-hailing like Uber ($68 billion) and Didi Chuxing ($50 billion) can probably come true when they start augmenting their human-driven cars with Nvidia-powered robotaxis.

Nvidia can usher in the datacenter-in-a-car concept. Now that AMD is encroaching the high-end gaming/workstation GPU market with its Vega GPUs, Nvidia needs to work harder growing its Auto and datacenter GPU businesses.

Why I Am Optimistic About Robotaxis

Since Intel and Advanced Micro Devices (AMD) won’t have an equalizer to the Drive PX Pegasus, Nvidia can sell it at a high markup. After government regulators approve level 5 autonomous cars on the road, taxi fleet operators and ride-sharing companies will start using them. It is probably going to take 3-5 years before government regulators approve level 5 robotaxis. However, Nvidia’s surging valuation is already boosted by its potential role as the top supplier of semiconductor products for self-driving cars.

The three-year return of NVDA is more than 1,000%. Nvidia launched its Drive PX hardware/platform in March 2015. Nvidia priced its first Drive PX car processor at $10,000 and it still found acceptance. There are now more than 25 car manufacturers who work with Drive PX hardware. We have to conclude that the Auto segment is a contributor to the massive 3-year return of NVDA. No other company has matched Nvidia Drive PX’s level of acceptance among car manufacturers.

(Source: Morningstar)

Going forward, self-driving taxis can be better taxi-hailing service providers than human driven cars. More often than not, human drivers tend to burden me with stories about politics, their health, their family, their favorite TV shows and sports teams. I am ride-sharing to get from Point A to Point B. I do not need the additional aggravation of listening to another person’s ills/problems/thoughts.

If the travel time will take more than 30 minutes, I would prefer to nap rather than petty talk with a human driver. Robotaxis will assure me I won’t be driven around by chance by a sickly/DUI driver. Unlike a human driver, the Nvidia Drive PX Pegasus-based robotaxi won’t consider kidnapping or robbing its passenger.

Final Thoughts

The Gaming segment is currently the biggest revenue/profit generator of Nvidia. However, AMD will eventually take more market share in discrete gaming/mining GPUs. Nvidia developing more GPU products for Auto and datacenter is commendable. As far as I know, AMD still has no car-centric GPU in the pipeline.

As per released specs, the 500-Watt Drive PX Pegasus will come with 2 integrated Volta GPUs and 2 discrete post-Volta GPUs. The first Volta GPU, the Tesla V100 comes with 21 billion transistors, 5,120 CUDA cores, and 640 Tensor cores.

(Source: Wikipedia/Nvidia)

The Drive PX Pegasus is a stellar example of how far Nvidia has achieved in making its formerly gaming-only GPUs become in-demand for other enterprise-level, scalable applications. The rapid appreciation of Nvidia’s market cap over the last twelve months is because many investors believe in the future roles of GPUs in datacenter/deep learning and self-driving cars.

NVDA is a buy. According to the Artificial Intelligence-powered Relative Valuation Model of FundamentalSpeculation, NVDA is fairly valued.

(Source: FundamentalSpeculation)

FundamentalSpeculation derived a fair value of $182.74 for NVDA. It came about after FundamentalSpeculation’s AI computed the average valuation ratios of other companies with similar business fundamentals of Nvidia to get the Cohort Fair Value. After which, the Cohort Fair Value was modified with values derived from calculating the average valuation ratios of Nvidia’s peers in the Technology sector and Computer Hardware Industry.

Disclosure: I am/we are long NVDA, AMD, INTC, GOOG.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

First ‘Level 3’ SD Car May Not Include Nvidia’s Drive PX – NVIDIA Corporation (NASDAQ:NVDA)

Two days ago, Audi (OTCPK:AUDVF) revealed the first ‘Level 3’ autonomous vehicle, the A8. The driver of the vehicle can legally take his eyes off the road while the vehicle is self-driving at a maximum speed of 37 miles/hour. The A8 would be the first vehicle to legally contain this feature.

Audi and Nvidia (NVDA) announced a partnership a few months ago that includes AI and support for Audi’s infotainment systems. One would expect that the released A8 would include Nvidia’s Drive PX2 platform. However, that won’t be the case. Nvidia’s blog, a one which acts as a marketing tool for the company, oddly, hasn’t mentioned any direct involvement for Nvidia in the vehicle’s AD system. Instead, the blog mentioned that Nvidia’s chips are powering the vehicle’s infotainment systems, virtual cockpit, and rear-seat entertainment options, all using the Tegra K1 processors.

Instead of using the Drive PX2, the A8 would be using Audi’s zFAS system for autonomous driving, which is powered by Mobileye’s (NYSE:MBLY) EyeQ and Nvidia’s Tegra SoCs for image recognition. It’s worth to note that Audi and Mobileye are also having a partnership for self-driving technology.

Now, the question is, why the A8 didn’t include the Drive PX2 instead of the zFAS?

We believe that the answer takes us back to the Drive PX2 inference capabilities. While the platform is having the highest processing speed, it’s still just limited to ‘Level 2’ driving as speed is not the only relevant option for any platform. For instance, ‘Level 3’ driving should be highly safe. The Drive PX2 doesn’t have safety characteristics, thus, it won’t be used in the mass-production phase of autonomous driving, yet.

For being legally able to include ‘Level 3’ features in its vehicles, Audi used Infineon’s (OTCQX:IFNNF) Aurix architecture to handle the safety part of its platform.

Also, Audi used TTTech’s Ethernet capabilities to connect the two parts of the platform in such a way that a part can’t take a decision without the other’s consent, which would increase safety levels.

It’s worth to note that the platform is relatively slow when compared to Nvidia’s Drive PX2. For instance, the zFAS platform can just take 2.5 billion inputs per second while the Drive PX 2 can take more than 1 trillion inputs/sec.

However, as we said, it’s not just about speed. The platform needs many characteristics in order to be implemented on vehicles in the mass-production phase (a phase where we stated last time that competition would be intense).

How can that affect Nvidia’s revenues?

If other OEMs followed Audi’s path and just used Nvidia’s Xavier and Mobileye’s EyeQ4 while not using the coming Drive PX3, Nvidia would take a significant hit as it would be an SoC supplier and not a platform supplier. The huge difference in price between the two can affect Nvidia’s future revenues significantly. For instance, in its Drive PX2 platform, Nvidia would sell two GPUs and two SoCs in addition to the platform’s board and software. However, by just being an SoC supplier, it would sell a limited number of SoCs (GPUs are much more expensive than SoCs), which would translate to a lower revenue per vehicle.
Also, Nvidia doesn’t have an edge in the SoC market as its Tegra processors haven’t recorded any significant success. While Tegra’s clock speed is double that of the EyeQ3, the former lagged significantly in performance per watt (whether Xavier is better than EyeQ4 on this front is still to be known). Also, there are many other competitors with deep expertise in the SoC design segment such as ARM Holdings, Qualcomm (QCOM), and Samsung (OTCPK:SSNLF), which would depress Nvidia’s margins going forward and divide more the TAM.

What’s the effect on the stock price?

Nvidia’s stock is currently trading at 12.3x LTM sales. This is well above the average of 3 it had prior to Q1 2016, the period which first witnessed high data center and automotive growth. Nvidia generated $7.5 billion in revenues over the LTMs. This means, by giving Nvidia a market value of $96 billion, the market is expecting Nvidia to generate $32 billion in annual revenues at some point. The additional $24.5 billion that Nvidia needs to generate to justify its valuation can only be generated from the automotive and data center segments as the TAM for gaming is relatively limited.

At its 2017 Investor Day, the company stated that the TAM for the accelerators for data centers would be worth $30 billion by 2020. Well, that is highly unlikely, ARK Investment Management explained why in details.

Let’s assume that Nvidia is right and the company dominated half the market (an optimistic assumption as ASICs and FPGAs are being heavily used). The company would need another $9.5 billion to just justify its current valuation, which can only be derived from the automotive segment.

Now, the question is if Nvidia’s Drive PX won’t be used as expected, can Nvidia reach the $9.5 billion target? We don’t think so.

Over the LTMs, Nvidia’s automotive segment generated $0.5 billion, recording a 44% growth Y/Y. For it to reach the $9.5 billion needed, the segment needs to increase its revenues by a 1,800% (!), which definitely can’t be achieved by selling SoCs, a chip that is cheap at wholesale level.

For Nvidia to justify its valuation, it needs to increase its data center revenues by 1,400% (LTMs data center revenue was $1 billion) and increase its automotive revenues by 1,800%. We believe that Nvidia can reach the target for the data center segment as revenues increased nearly 200% in the last year. However, we are skeptical towards Nvidia’s ability to reach the automotive segment’s needed target due to the reasons above.

If other OEMs followed Audi’s path, we believe that the upside for Nvidia is very limited from current levels. However, if that turned out to be a one-time event, we believe that Nvidia stock can reach $200/share.


We are long Nvidia because we believe that its edge in the data center segment is unreachable by any competitor. However, we are skeptical about its edge in the automotive segment, a segment that must be a huge success for the company to push the stock upwards. The AD’s future is still unclear with so many innovations happening on a daily basis. We sold two-thirds of our Nvidia position last week, and we won’t add to our position until we see a detailed roadmap for Nvidia’s part in the mass-production phase of AVs.

For the short term, we rate Nvidia as a “hold”.

Important note to readers: This article is for informative purposes only as we are not recommending to buy or sell the stock.

Disclosure: I am/we are long NVDA.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Rádio comunicador tático MJ270MR Motorola – Para Emergências, sobrevivência e Bushcraft.

Ótimo para garantir a comunicação em acampamentos, jogos de airsoft, paintball o Rádio comunicador tático MJ270MR Motorola também funciona para Emergências, sobrevivência e Bushcraft, opera muito bem em regiões ermas e de baixo sinal de celular . Inscreva-se!
Siga nossos links:
Bunker Food:

Curtiu, então apoie nossos projetos:


Linha de facas Survivalist

O Guia do sobrevivente é um canal focado em sobrevivencialismo e um dos pioneiros no Brasil neste segmento. Nossa meta é mostrar que tecnicas de sobrevivência são úteis no dia a dia e que pessoas comuns tem total condição que garantir o básico para viver com dignidade mesmo no pior cenário possível.