I split my time between making investments and working to improve quality of life for people by reducing costs in key industries
Joined August 2018
- Tweets3.7K
- Following111
- Followers2.2K
- Likes55.2K
The AI industrial complex is pointing to charts showing growth in number of tokens being used, and in doing so is knowingly omitting an inconvenient truth. One need simply Overlay flattening token dollar spend, $300B+ chips on the sidelines, apparent rapid model compression, and new capacity in order to see the writing on the wall.
Everyone wants to know what will happen next with AI in markets and industry, and the near term is unknowable, but there are some signals.
What we have right now is increasingly extreme efficiency gains in models as evidenced by the new Deepseek release and so many others. The momentum in improvement of performance per dollar appears to be a combination of harness improvements, compression, and model efficiency gains. Combined, these three are improving inference/dollar dramatically. This naturally puts into question the amount of compute needed to support dramatically more efficient models which are being increasingly operated on edge, or not through data centers.
On top of this we have hardware improvements which are separate and agnostic of the individual models, but rather act on the market for compute as a whole. The effects of new hardware are delayed and distorted, because the process of getting the new hardware plugged in and operational is slowing down dramatically and there is a huge amount of construction in progress, whereas with the model improvements they flow through right away. So even though the new chips are an order of magnitude more inference/dollar, the supply does not flow through in the near term because they aren’t being plugged in for 1-3 years.
overall, supply demand inflection is unknowable in the near term, but it does seem to be the case that the direct and indirect supply of inference is now reaching escape velocity relative to demand, which is also growing but running into the law of large numbers and transmission limitations of digital intelligence vis a vis actual useful work in society…
Live by your inner scorecard. Comparison is the thief of joy
More important then ever in these days of fast food media abound
The issue with Elon companies is not the product. It is the distribution and customer service. A product like cell phone service will amplify this issue given the critical customer service and brick and mortar touch points around phones.
Tesla built the best cars and still plateaued. Deliveries peaked near 1.8 million in 2023 and have not grown since, because service stayed spotty and sales never matched the engineering.
Phones are harder. A dead handset today means a T-Mobile store down the street. If the carrier is SpaceX, do you wait three days for a shipped part? That does not work for something people treat as urgent.
No physical footprint, same ceiling. Best product is not enough if support and distribution lag.
The idea that chip and memory companies should be worth trillions of dollars because AI is important is akin to the idea that lumber and cement companies should be worth trillions of dollars because housing is important. It misses the size and transmission of the value chain relative to the commodity components within it.
There continues to be a whole lot of talk about the AI models getting smarter and smarter (which they are), but not as much about how much money is actually being saved using AI or examples of AI actually doing useful work all on its own at meaningful scale. It does seem that AI will ultimately get there, but the timeframes on which investment is being based are increasingly unrealistic.
So we have risk free real yields on 30 Year TIPS at 3%+ inflation hedge. These are selling off.
At the same time, we have SP500 at a CAPE north of 42, +~ 0% real earnings yield inclusive of all kinds of late cyclical margin tailwinds and circular hardware revs built in. Pushing all time highs with room to run.
So markets are preferring 0% real yield on late cycle equity risk over 3% risk free real yield in treasuries.
What could go wrong here?
In twenty years we will look back on the current AI infrastructure explosion and tokenmaxxing boom much like every other previous major innovation-Ultimately transformational, with much over excitement, capital destruction, and adjustment of expectations in the process.
Nobody can truly afford the closed AI models at the current economics, which are net negative for all but the hardware manufacturers. Further, compression is happening so rapidly across ALL of the models that the frontier model of 18 months ago now runs on the edge of an old iphone…Ultimately the costs are simply too high for usership not to converge on the cheapest high quality models and for margins to commodify.
AI retweeted
Proud to unveil the Luce Migliore
A Ferrari electric vehicle designed by me and ChatGPT in 8 seconds at a cost of $0.10
🤖 Made with AI
Hot (but not so hot) take on GME EBay proposed deal
Classic case of an operator with great vision (cohen) not having good skill in capital allocation. From a synergies standpoint it’s obvious that eBay/Gamstop is a formidable if not dominant player in collectibles. But that is then end of the good part of the proposed deal. The bad part;
80%+ dilution to GME shareholders in order to issue the shares and raise the cash required for the base offer
75%+ loss in book value per share to get it done
6X+ resulting debt/ebitda creates a leverage overhang with no maneuverability and certain death in any recession or collectibles market slowdown.
It’s a no for us
AI retweeted
In my teens and 20's I would spend way too much time playing Starcraft and Civilization. Harvesting resources, building things, and expanding was super addictive to my brain - to an almost unhealthy degree.
Later I realized that entrepreneurship and business is the ultimate game. It scratches the same itch for me (resources, building, expanding), but you're actually contributing to humanity at the end of the day, which can be much more fulfilling.
Business is also much more positive sum than video games. In Starcraft, the other player has to lose for you to win. In business, there is competition, but in a growing market there can be multiple winners. And gains compound long term (it's a infinite game) instead of starting over each time.
Now days I prefer to watch pros play video games to unwind, instead of playing video games myself. But a quick game can still be fun here and there to unwind. By contrast, the game of business is played over many decades.
AI retweeted
Crazy idea: Let's split a country in socialist and capitalist halves and check in on them in 75 years.
AI retweeted
People underestimate how long it takes to win big.
You struggle for 10 years. Eventually, in one day, you achieve more than you did your entire life.
Be patiently aggressive.
AI retweeted
“I lived so carefully, thinking someone was watching. But the stage was empty, the audience never came.”
— Osamu Dazai
AI retweeted
When you’re 5 years old, a year is 20% of your life. And when you’re 50 years old, a year is 2% of your life. This is an explanation given why time speeds up as you age. It's called Janet's law. It states you’ve experienced roughly half of your perceived by life by 20 years old. Or to put it another way: A summer holiday for a 5 year old feels as long as the 10 years from 40 to 50 years old.
But Janet's law can be broken with high agency.
You have agency over the speed time. You're not a passive victim. A better explanation of why time speeds up as you age is because you have fewer new experiences as an adult, so your brain deletes the memories. If you take agency over your life, do new things and create memory dividends, time slows down.
If you live your life on autopilot, you may die at 80, but feel like you died at 20 years old.
If you take agency over your life, you may diet at 80, but feel like you died at 200 years old.
A lot of AI tools look impressive in isolation. Very few of them actually do useful work inside real organizations.
Every enterprise has its own rules, manuals, technical expertise gaps, edge cases, and regulatory environment. AI only becomes valuable when it is deeply connected to those realities.
White glove development, implementation, and maintenance will remain a critical service and an opportunity for great teams of engineers
Force AI is doing this for insurance, however use cases will emerge and persist across all enterprise
Second order consequences
California's Central Valley produces 80% of the world's almonds. Each almond requires 3.2 gallons of actual irrigation water to grow. Not rainfall. Actual tap water pumped from aquifers.
One gallon of almond milk requires 162 gallons of irrigation water. Compare that to dairy milk at 8 gallons of tap water per gallon, with the rest being rainfall that falls on pasture anyway.
But here's where it gets properly grim. Almonds bloom for exactly three weeks in February. During those three weeks, California needs every pollinating bee in North America transported to the Central Valley or the crop fails entirely.
Commercial beekeepers truck in 31 billion honeybees. That's two-thirds of America's entire managed bee population, all concentrated in one valley for three weeks. The bees are packed into trucks, driven across the country, dumped into almond groves drenched in pesticides, worked to exhaustion, then packed up and shipped to the next crop.
The mortality rate is catastrophic. Beekeepers report losing 30 to 50% of their hives annually. That's billions of bees dead. Not from natural causes. From being used as disposable pollination machines for your almond milk.
The pesticides don't help. Almond groves are sprayed with neonicotinoids which scramble bee navigation systems, fungicides which weaken their immune systems, and herbicides which eliminate the wildflowers they'd normally forage on between almond blooms.
Meanwhile the aquifer depletion is permanent. The Central Valley has sunk 28 feet in some areas from groundwater extraction. That water took 10,000 years to accumulate. It's being drained in decades for almond milk.
Your vegan latte killed more bees and used more water than a year's worth of dairy milk. But it's got "plant-based" on the label so you're definitely saving the planet.
