I'm a CPA with a focus on equities, derivatives and currencies that benefit from chaos and shortages. Conversely, I short anything that's oversupplied.
Toronto
Joined February 2014
- Tweets3.3K
- Following278
- Followers843
- Likes6
There are several types of traders - fundamental, technical, dividend and momentum.
If or when the $mu chart looks better, it will attract the momentum trader. For the moment, only the technical and fundamental are in $mu. It will take a bit of patience for the chart to look better for new money to come into $mu.
Long time follower and MU bull but $1800 in less than two quarters?? I would love to see that but without the type of volume we had in the summer and the cautious market sentiment around semi, partically memory, what’s the catalyst? Buybacks?
1
1
432
The fact that China targeted Jensen underscores the geopolitical significance of AI leadership.
That means both Democrats and Republicans will need to build lots of data centers. $NVDA
1
180
The government will backstop AI. No chance the West loses the AI race to China. The government has already invested in $intc and thrown some money at Quantum computing.
Read up on the history of credit tightening. Government also steps up to keep the economy going.
1
2
290
Laughable to believe that $crwv and #memory is a correlated trade. If $crvw goes out of business another customer will take its allocation of chips. Moreover, $crvw is tiny in context of the entire memory chip market, which is dominated by the hyperscalers.
@elonmusk would like to see $crwv and other neoclouds go out of business so that $SPCX gets a larger allocation of #memory.
Claiming AI compute demand peaked after 3 years is like saying that in 2003 $amzn is nothing more than online bookseller.
20 years from now, when wars are fought with AI and robots, you'll wish you'd invested a lot more in AI in 2026
Technological revolutions don't end after 3 years. AI will be the most transformative tech in our lifetime.
Agree but in the short term market is feeling the AI fatigue. Capex growth is no longer accelerating, capital markets is now a constraint, token maxxing has ended. The winners will be the ones who can convert AI into cash flow ($MSFT etc). Play the winners and leaders in each space is the move imo
What % of AI infrastructure spending do the neoclouds have and could the 4 hyperscalers take their place if the neoclouds go bankrupt?
You're right—WorldCom and Global Crossing were the classic infrastructure overbuilders that went under after the fiber glut. Closest AI parallel today is the debt-heavy neoclouds financing GPU clusters and data centers without hyperscaler balance sheets. CoreWeave (and peers) already show stress with high yields and leverage. Shakeouts likely concentrate there; the big spenders have profits and scale that the pure telco plays of 2000 lacked.
108
Apart from the smaller neoclouds, I really can't think of any AI related companies that will go bankrupt. The companies that are spending the vast majority of the money $amzn, $msft, $googl $meta are financially solid. These four companies hold up the entire AI trade.
I think it would be quite healthy to get a real multi-month basing here and take out the trash in some of the bottleneck sectors, see laggards forced to sell/consolidate and have common holders wiped out in the worst names.
127
Every weekend, when people have too much time on their hands, they argue about some idiotic topic in which they have no knowledge. This weekend it's about LTA.
LTA are irrelevant. If #memory demand exceeds supply $mu will be fine. If not, #memory is toast.
I can't wait for next weekend when #twitter / X moves onto the next stupid topic.
Ok, I have said a lot about the LTAs myself too and I still think Jukan makes some good point worth noting. Of course, right now, none of this matters because the industry remains in an upcycle. The conditions made by SCAs/LTAs on pricing, volume, and duration look good on paper because the industry is growing, so the better question to ask is:
How reasonable does it sound to keep pushing your customer to buy your inventory in a downcyle? Would it make sense to ruin your relationship with your customer?
Yes, HBM is not something you can buy and store and inventory standalone but even if the LTAs remain honored, I have a hard time not seeing the market continue repricing memory stocks to price in for a future after the LTAs expire.
147
Read @Twaharashid's tweet. Making good investment decisions requires you to read and understand.
I have no idea if his tweet is accurate. If it is, tankers will get hit. $eco $dht $fro
You can add 50 ships of supply you still need 50 ships worth of crew & diesel for the additional supply
258
Cashflow is really the only metric that matters. The winners in AI will earn trillions of $.
Stop looking at AI like it’s just a chatbot that summarizes emails. We are barely in the first inning. The real revolution is happening behind the scenes in medical research, weapons and robotics. The hype might peak, but the actual utility is just getting started. And that's where the money is made.
Comes down to whether HBM can sustain the ROE at current levels. Question is this really the cycle peak? If not then indeed cheap but if not and ROE trends lower, P/E can ramp in a few quarters and it is a value trap
1
1
190
You don't use charts to tell you what happened. You use them to anticipate when technical traders will buy or sell. Knowing what other traders are likely to do, and doing it before them, gives you an edge.
If you don't use a chart, you're buying or selling at random prices.
Once you’ve sold a stock, it’s hard to decide when to buy back!
You need to make two right decisions to profit: when to sell and when to buy back.
For those relying on charts instead of fundamentals, you often end up buying and selling too late because charts are like rearview mirrors!
102
If the below happens, it will be very bad for tanker bulls. It immediately adds 50 ships to tanker supply. $eco $fro $dht
Trafigura, Vitol, Mercuria and ADNOC are seeking US approval to acquire or lease around 50 sanctioned tankers as freight costs explode. Chartering a supertanker from the US Gulf Coast to China reached nearly $80m in October, roughly 10x pre-war levels.
3
2
521
$mu bounced at the 50% Fibonacci retracement of $659. It could retest it, but if $659 holds, the #memory carnage is over.
This is why charts are important, it lets fundamental investors know where technical traders are buying, like me, who bought at $660. If this doesn't hold, $540 is the next stop. Yikes !!
Misleading narratives is how extraordinary profits are made. Eventually the market figures it out and will correct.
In the meantime you can buy. Imagine a second opportunity to buy $mu below $750 or $sndk below $1,000. Wouldn't happen without FUD.
Nope. Healthy markets call for skepticism. If you want an efficient market, you need to embrace moments like these.
Won't be the first ever, won't be the last ever.
1
333
In the short term, $MU’s chart looks bad, and technical traders are selling. This has nothing to do with fundamentals
Not all investors trade the same way. Different investors have different strategies, time horizons, and reasons for buying or selling.
153
The worst type of analysis is any that only considers one variable. The goofy chart below does not factor the importance of first mover advantage or revenues.
DO NOT BUY THIS AI DIP
Most of the companies at the top, the hyperscalers, are investing way too much way too quick
The laws of capital dictate current capex/depreciation ratios will eventually lead to declining and even negative ROIC rates
All present for $GOOGL $META $AMZN $MSFT
But even more so for chipmakers like $MU $SNDK $SKHY $DRAM
Don't be surprised to see this fall 20%+ from highs, and most already did
82
AI infrastructure delays mean little when chip supply is already the bottleneck.
If demand is for 10 factories but chips can only equip 7, delaying 3 factories simply gives the supply chain time to catch up, especially power.
That's deferred demand, not lost demand. $mu $dram $SKHY
@heyfinlo
The AI trade is apparently dead again after last night’s blue tsunami but I think this is another example of the narrative following recent weakness in stocks rather than anything actually changing about the demand for compute.
Yes a democrat winning Texas governorship given that about a quarter of planned U.S. data center capacity sits there but Hinojosa isn't saying AI infrastructure can no longer be built in Texas.
All she wants is to pause new development until communities have more control and developers cover more of their infrastructure costs which is very different from permanently banning new projects.
What actually changes is how quickly and cheaply new capacity can come online (not demand for compute) which could actually make already-powered sites more valuable while increasing importance of on-site power generation, electrical distribution and efficiency as developers look for ways to keep building.
2
331