You're Long AI - You Just Don't Know It Yet
I got worries on my mind
A brain dump of what's on my mind:
- Oil, I think we're all kidding ourselves if we think this is resolved. Prices are pinned a bit too low.
- Trump himself said "Bedlam in 4 weeks" 4 weeks ago when the MoU was executed. I'm worried for what energy looks like in a month's time.
- Have you heard about the one where the Iran war was put on pause just so SpaceX could get the IPO done?
- Rates pricing in maybe 1 more rate rise by the end of the year. Oil goes higher, inflation persists, energy cost keeps rising, RBA forced to act.
- A few more months go by, food prices creep higher as no fertiliser and we have a brutal summer due to el nino, RBA can't cut again and maybe rates go higher and higher.
- El nino persists and we have another period of high oil prices and high food costs all while AI datacentre buildout in Aus causes crowding out of the construction sector again. Can't cut again, maybe rates go higher.
Truth is probably somewhere in the middle, the datacentre buildout will be linked to higher gdp growth too.
Datacentre buildout
Potential massive boon for Australia, if we see significant buildout it sounds exciting for GDP Growth. But the frontier of GDP growth is going to be unevenly distributed and the government hasn't quite worked out how to tax hyperscalers yet.
The government really needs to get its act together as they're seemingly concerned about IP that was likely all lost 3 years ago. If Australia can be a "safe" jurisdiction for US based "labs" and hyperscalers, there is so much upside for Aussies. There is convexity in the government getting a deal done, I just hope one can get done in time.
Imagine if we can't get enough trades to build homes because everyone is too busy building datacentres, ha!
Zigging while zagging
This happens every once in a while where you see US tech holdings across institutions fall substantially. It's happening now, on the back of the memory degrossing event.
At some point the money will follow the growth.
Have a look at Amazon for example, they own 20% of Anthropic and get to collect a margin straight off the topline revenue from any instance of Claude run off their servers. You have leverage to the fastest growing company in the world via public markets.
Be aware though that every hyperscaler is going to generate negative free cashflow as they're all investing in the future. There is a good reason why public companies don't usually go negative FCF to invest in the future and this is better kept to private equity, j-curve etc.
Investing in infrastructure for the future is not that bad! But share prices might not necessarily reflect that in the near future.
Do your job
I've just outlined an area that potentially delivers convexity and inoculates against missing out on returns. Timing matters however!
Map it out, AI is here to help, you can get a shocking amount of work done to understand what your portfolio might look like in certain scenarios.
You don't even need to think through stuff, just prompt away to build a framework. Even a partly built thing has more value than doing nothing.
My only suggestion is look at your best performing exposures over the last year and work out what drove performance. You're not avoiding the AI trade by going underweight Mag-7 because some "quality" global equities manager told you so, you've probably picked it up elsewhere. Either via "value" global equities or via infrastructure or via commodities or via credits.
You're long AI - you just don't know it yet.