A Retirement Journal: Parallels between Cisco 1999 and Nvidia today

Semiconductor company NVIDIA’s stock price has been on a tear the last 12 months (picture above on right) rising over 5X and propelling its market cap above $1 trillion. Another tech company on a similar type of run over 20 years ago was networking company Cisco. ?Between 1998/1999 and 2000 Cisco’s stock price rose by a similar multiple (picture above left). ?I spent over 10 years at Cisco. Unfortunately for me financially, I joined at the peak not during the rise. ?My sign on shares were priced at $64, a level the company 23 years later is only now starting to even approach with its share price in the $50s.* What propelled Cisco’s meteoric rise in the 90s was the emergence of the data communications protocol TCP/IP and Cisco going all in on it. For NVIDIA, their core product at founding (and still) was the Graphics Processing Unit or GPU. Today, the sweet spot of NVIDIA’s A100 GPUs is in the intensely hot AI field. The company can’t meet demand and can price at almost any level driving the stock price sky high. ?I posit here there are parallels between Cisco in 1999 and NVIDIA today.

Cisco was founded out of Stanford when the husband & wife team of Len Bosack and Sandy Lerner realized they were on to something when they created a work around for a domestic situation. The couple who worked in different departments of the university wanted to send messages to each other during the day as to who had the responsibility to feed their cat. Finding the phone system inconvenient, they preferred the emerging medium of e-mail. However, the respective departments where the couple worked used different computer systems with different protocols for communicating or messaging. The couple jury-rigged a crude multi-protocol router that enabled the incompatible computers to communicate with each other. As data traffic grew with the explosion of e-mail and other types of data communications, the increased traffic hastened the standardization of the TCP/IP protocol eventually adopted by all computer manufacturers. Cisco rode that standardization wave which turned out to be the start of the public Internet. The company was the figurative supplier of picks and shovels for the Internet Gold Rush. ?TCP/IP eventually became the standard for communications all other forms of communications besides data. The TCP/IP migration started with voice on the traditional Public Switched Telephone Network (PSTN), then to broadcast and eventually streaming video, then to the machines on the factory floor, and finally to sensors everywhere including our bodies. There were hiccups along the way. The initial explosive growth of the Internet spawned hundreds of the early dot com e-commerce companies. These new companies in turn spawned new telecom companies and spurred existing ones to build out capacity. ?Most of the early dot com business models were not well thought through and the bust of 2000 caused hundreds of these (e.g., Pets.com) to disappear. Telecom carriers who had overbuilt capacity suffered as well. ?Cisco’s wings were clipped. The picks and shovels company that supplied this gold rush came down to earth. Cisco had its first layoffs and a massive one at that in early 2001. However, the fundamental build out of the Internet was going to continue, dotcom hiccup notwithstanding. A chastened Cisco continued with much more realistic expectations. ?Cisco transformed from a growth stock to a value stock paying dividends. As for me, though I did not have the wild returns from the gold rush years, the company has still been a significant part of my family’s financial security. I do muse on the very rare occasion on the would have, could have, should have!

I see strong parallels with NVIDIA and its GPU roots. ?Started in 1993 on a napkin drawing in a Silicon Valley Denny’s restaurant by Jensen Huang and his co-founders, NVIDIA believed the small but distinct market for gaming would benefit from the acceleration of the graphics and images a Central Processing Unit (CPU), the engine of PCs and laptops, just could not provide. We witnessed the emergence of the first NVIDIA GPU. While the company’s early years were good, NVIDIA’s TCP/IP moment came last year with the explosion of Open AI’s ChatGPT and other AI Large Language Models (LLM). ?NVIDIA’s GPU architecture is far better suited for processing needs in this hot new field compared to CPUs. Before AI burst on the scene, NVIDIA ?GPUs had also caught a moderate size wave in the processing ?needed for data mining in crypto currencies. That said, the AI wave has been orders of magnitude bigger. Looking forward, NVIDIA engineers are already finding the next set of waves or use cases. One interesting one is the processing and simulation of Quantum Algorithms, an alternative before the emergence of full-blown Quantum Computers. Jensen Huang and his company will continue to grow and profit for many years.

Nonetheless, I find NVIDIA’s today situation and hype comparable to Cisco’s in the dotcom era. Many of the initial LLM AI use cases will fail with unproven business models. Today’s unlimited demand for GPUs will I predict fall, just as Cisco’s demand curve did in 2000. ?NVIDIA stock will come back to earth. But, like Cisco, NVIDIA’s GPU, the core building block will continue to play a critical role in building out the next and more sustainable wave of AI. ??Even then, NVIDIA’s trajectory won’t exactly parallel Cisco’s. The semiconductor industry has cycles different than networking. The ecosystems are different. Cisco during the 90s sold directly to the enterprise dot com companies as well as the telecom companies building out the internet. In the case of NVIDIA with the AI application, established cloud players like AWS and Microsoft Azure are NVIDIA’s customers providing the brawn of AI. Software companies like Open AI are the brains of AI. Both these categories are more secure anchors than what Cisco had in 2000. ?NVIDIA will, like all companies reaping monopoly profits, attract competitors. ?Other forms of computer architectures like Google’s Tensor Processing Unit (TPU) might turn out to be better suited for neural network processing required in AI than GPUs. Even so, I am sure NVIDIA will gravitate toward the technology architecture that makes the most sense. ?It will be a fascinating next era for the company.

What goes up must come down. However, it does not have to come all way down. ?What do YOU think about NVIDIA’s future trajectory?

Jake

*Cisco’s just announced acquisition of Splunk for $28 B has driven the share price down slightly but I expect it will not drop too much.

?