2 days ago
1
If you’re reading this, you probably know about the artificial intelligence (AI) boom currently underway, and you’ve likely heard some crazy numbers being thrown around casually. Whether this boom and its maths tethers to reality or is just another bubble, I will leave that conclusion to you (comment with your thoughts below this article )…
Instead – and following Nvidia’s much-hyped results released last week – I ran some numbers to provide context on exactly how large the current and expected capital expenditure (capex) cycle for AI actually is.
And it is quite mind-blowing!
Nvidia’s well-spoken founder and CEO, Jensen Huang, dropped some amazing numbers in his results conference call (though it’s worth taking them with a pinch of salt):
- Hyperscalers (i.e. Big Tech) are currently spending c.$600 billion on AI and datacentres, and
- Jensen expects AI spend to reach $3 trillion to $4 trillion (yes, trillion) by 2030.
Listen: Nvidia results shoot lights out … again
Read: Nvidia forecasts decelerating growth after two-year AI boom
For a moment, let’s assume Jensen is not biased towards hyping the very cycle that his business sells their products into, and let us assume that these figures are accurate.
How big is this spend? And how does it compare globally? And versus wee little South Africa?
Well, the US economy’s latest quarterly figure implies a nominal GDP figure of approximately $30.34 trillion, so current AI spending of $0.6 trillion represents around 2% of the American economy now. And, if Jensen’s prediction is correct (and the US economy remains flat outside of AI spending), the AI bill will grow to about 10% to 13% of the economy by 2030 (though it’s more likely to be closer to 5% to 7%, as the non-AI economy will also grow).
This equates to roughly 0.5% of the current global economy (circa $114 trillion per annum) being spent on AI today, increasing to around 2.6% to 3.5% of the global economy by 2030.
In fact, if we compare this with the number of humans on planet earth (approximately 8 billion), the current AI spend is about $75 per person, expected to grow to $375 to $500 per person by 2030. This is equivalent to roughly R1 324 per person right now, or R6 622 to R8 829 per person by 2030.
Read:
Jack Ma-backed Ant’s profit dives 60% after AI, global expansion
Apple faces calls to reboot AI strategy with shares slumping
(Are you currently spending more or less than R1 324 per year on AI? Do you think you will be spending R7 000 to R10 000 per year on AI by 2030? Just a simplistic thought, as revenues are not margin, and neither is capex…)
Spending in SA context
Which brings us to comparing this AI spend with South Africa: the current AI spending bill by the hyperscalers is one-and-a-half times the entire size of the South African economy (which is only around $400 billion per year)! And – assuming the government continues to stifle our GDP growth and keeps it flat – by 2030 the AI spend could be eight-to-ten times the size of our local economy!
Read: How South Africans are actually using artificial intelligence
Listen/read: AI offers hope in South Africa’s growing water crisis
In fact, looking at the current $600bn AI spend and pretending that it was aggregated into a company and listed on our JSE as a single entity that traded at a 1.0x book value (with no costs), its market cap would be around R10.5 trillion.
This would make it the same size as Prosus (market cap of R2.6 trillion), BHP Group (R2.5 trillion), British American Tobacco (R2.3 trillion), Anheuser-Busch InBev (R2 trillion) and Richemont (R1.6 trillion) combined!
We are living through the largest, loudest capex cycle the world has ever seen – if not in real terms, then certainly in nominal ones. How does this end? Where does this go? How is this monetised?
I have no idea, but tell me again how Big Tech are “capital-light” business models?
* Keith McLachlan is CEO of Element Investment Managers.
English (US) ·