AI versus dot-coms from another angle: shovels, expiry and circular money

Everyone compares AI to dot-coms as the same story: a bubble, a cap on a box, an 80% fall. That is boring and off target: there they compared trucks, here they compare factories. Here are four angles rarely put together.

Peak multiples Cisco and Nvidia

Fig. 1. Peak multiples. Data: Creative Planning.

The first angle is the price of the dream. Cisco at its March 2000 peak traded at more than 150 times forward earnings and 29 times sales at a 579 billion dollar market cap - the most valuable company in the world. Nvidia at about 5.2 trillion dollars trades at roughly 50 times earnings and 20 times sales on about 253 billion in revenue. So the bubble today is cheaper per dollar of earnings.

Factory versus highway​

The second angle is what is being built. In dot-com days they built fiber optics that live for decades, with 98% of capacity unused. In AI they build data centers and GPUs that age out in three years. Fiber overbuild hurts but lasts; GPU overbuild hurts and quickly deprecates: amortization eats everything without utilization.

Peak revenue Cisco and Nvidia

Fig. 2. Peak revenue. Data: Harding Loevner.

Who pays​

The third angle is the customer. Cisco sold routers to excited startups on vendor financing. Nvidia sells shovels to four giants: 39% of revenue from just two customers and 46% from four, concentration almost double Lucent's back then. Plus circular financing: Nvidia invested about 100 billion in OpenAI, which then buys Nvidia chips. That is like Cisco paying for its own buyer.

Revenue growth Cisco and Nvidia

Fig. 3. Revenue growth speed. Data: Trefis.

Cash flow Cisco and Nvidia at the peak

Fig. 4. Quarterly cash flow at the peak. Data: Trefis.

The fourth angle is cash inside. Cisco made 1.3 billion in quarterly operating cash flow. Nvidia makes about 25 billion per quarter, with 55% CAGR at Cisco versus 70% at Nvidia on a 13 times larger base. According to Harding Loevner, Cisco's peak annual revenue was about 19 billion, and 22 years later 52 billion, so the business survived but the stock fell 88% and took a decade to come back.

⚠ IMPORTANT The takeaway: AI is dot-coms in reverse. There the bubble was the promise of traffic, here the bubble may be the promise of utilization. Cisco fell not because the internet did not win, but because everyone bought shovels at the same time.
 
Back
Top