From Fossil Fuels to Neural Networks: Trillion-Dollar Shifts Shaping Our World.
The AI Craze is crazy!
Over the last 5 years, the Capital Expenditure by top tech firms, especially the BIG 5 (Apple, Microsoft, Meta, Alphabet and Amazon) has reached ~$1 trillion.
Over the same period, OIL&GAS majors have done a CAPEX of about $800 Billion.
Result: Tech now dwarfs Oil & Gas in CAPEX.
If utilization and cost curves keep improving, AI spend becomes the digital economy’s infrastructure.
But at what cost?
Are the economic gains from ~US$1Trn in AI build-out greater than the tangible, visible gains from Oil & Gas? Oil & Gas still accounts for over 50% of world energy consumption by source.
What does AI drive at a commensurate scale?
LLMs correcting grammar, gamers beating people in chess, prompt engineers writing new way to ‘search’? May be these observations are a bit too harsh. There are some serious big-ticket improvements that is doing in the field of medicine, finance and education.
History tells us that there are very few technologies which change the world. May be AI is one of them, or it isn’t. I have no clue. But history also tells us that the very few firms survive the first onslaught of new tech development and the mortality rates of firms in innovative sectors is very high, leaving a handful of survivors.
The question to ponder. How much are we paying for this innovation. By one measure, $36 trillion globally (as per GICS classification). That’s one third of Global GDP.
On S&P 500 definitions, Oil & Gas, the so, called “sunset” sector carries ~US$1.6T in market cap, while Technology, the anointed “future”, stands near ~US$28trillion (just in US). Has the market prepaid too much for that future? Perhaps many times over. The answer hinges on testable unit economics: rising utilization of deployed compute, falling $/inference, reliable power, and proof that AI productivity lifts diffuse beyond tech vendors into the broader economy. Until those show through the P&L, the valuation spread is a bold assumption, priced as inevitability.
Another extremity.
The technology sector market capitalization is now approaching $30 trillion in MSCI ACWI Index. This is equal to the combined valuations of the following ‘Sleeping 7’ sectors combined:
- Consumer Staples
- Energy
- Materials
- Utilities
- Real Estate
- Consumer Discretionary
- Industrials
The tech sector’s implied earnings are about $600 Bn while the 7 sleeping sectors make about $1.5 trillion in net income. The tech sector trades at about 38x trailing earnings while the sleeping 7 trade at about 22x.
The reason is of course the huge differential in ROEs. Tech sector operates at a whopping 25% ROE while the sleeping 7 are at 12%.
The problem is, at extremes, the extremities look very likely to continue. Extrapolation of extremes cause mistakes. Beware.
This is turning out to be the biggest bet on the ‘unknowbale’ future. One of the most outrageous extrapolations we have ever seen in a long time.
Maybe it will go right. But at what cost?
What’s the ‘Margin of Safety’?
As sceptics, we are watching.