top of page

Monthly rollup | August 2026

Sep 7
7 min read

Stocks Mentioned Atlassian (TEAM), Alphabet (GOOG), Nvidia (NVDA), Apple (AAPL), Super Micro Computer (SMCI), Tsakos Energy (TEN), Standard Chartered (STAN.L), Methanex (MEOH), Financials sector ETF (XLF), CoreWeave (CRWV), Nebius (NBIS), IREN (IREN), Hut8 (HUT), Cerebaras (CBRS)


Highlights


August 03, 2026


Anyone remember the era when software was considered un-investable because cash flows could dive to zero when a better product came along?


That's going to happen to the AI labs at some point.



August 04, 2026


Buying panic in AI-heavy US equities after a prominent AI bull blew up his fund.


This comes on the heels of South Korean investors losing their shirt chasing AI names.


Zero lessons learnt by the bubble chasers.



August 07, 2026


Celebrating a big win in Atlassian


Atlassian (TEAM) is up 31% pre-market and >100% from my entry earlier this year.


Software is dead, they said. Saas-apocalypse, the AI bulls called it.


While commodity guys were getting frustrated buying gold and oil stocks, I bought a GARP stock in this beaten down sector.


Terrible macro, terrible sentiment, poor looking chart, and worse - the company has a history of losing money.


But I did the work, and subscribers who paid for my work profited from this trade.


As they did with Tsakos Energy (TEN), Standard Chartered (STAN.L), Methanex (MEOH) and numerous other picks across sectors and countries.



Alphabet betting it all on AI


Alphabet (GOOG) is raising another $25B in debt to blow on AI capex.


The stock does not deserve to trade at a $4.3T valuation given that AI has cannibalized search and its primary business is past its peak.



August 10, 2026


Canva calls frontier AI models uneconomic


$42B Canva says serving AI requests at scale using frontier models is a money loser.


Saying so is currently as unfashionable as shale drillers refusing to drill a negative ROIC well in 2014.


AI usage is already heavily subsidized and it is still uneconomic. What happens when the labs are forced to pass on the real cost of training + inference workloads?



"Texans must come first."


Following on the heels of New York banning data centers, Texas is auditing data center projects which require access to the power grid.


The audit also scrutinizes water use and local community impact.


ERCOT has received requests for 474 GW of power, which is 5x the state's peak consumption. No way that's happening.


When the AI bubble pops, chips idling in warehouses will get scrapped without ever being plugged in.


Nvidia (NVDA) and private credit funds are going to be the new subprime.



Bitcoin technicals


TA-wise, bitcoin is in no man's land. If I had to trade it, I would bet on the direction of the trend and go short with a stop at $68,000.


The momentum is going to be epic when this breaks out/down.



August 14, 2026


Can AI eliminate white collar jobs?


Current AI ARR = $80 billion

Global software industry = $1 trillion

Knowledge workers income = $13.5 trillion


The AI bulls predicted a SaaS-apocalyse because Codex and Claude could write code. They said AI would grow to $1T by taking market share from application software.


Last month, software made a comeback, while AI hedge fund(s) blew up.


So the bulls moved the goalpost. AI is now supposedly going after a much bigger market - white collar jobs.


Claude will replace Dave in accounting and maybe even the CFO. The $13.5T in wages that knowledge workers earn will go to agentic AI actors. The TAM is yuuge!


Here's my counter:


Below is a chart showing the branch growth of India's largest private sector bank.


In 2016, the country demonetized 85% of cash in circulation, forcing its citizens to go digital.


UPI grew from near zero to 83% of digital payment volume. Yet, HDFC keeps adding branches, as do other banks.


Digital payments didn't replace bricks-and-mortar branches.


AI didn't replace Microsoft Office. And it's certainly not going to replace knowledge workers.


The AI bulls successfully sold the $1T dream while falling 90% short. Now they need to replace it with a bigger dream to keep the circular financing going.


The numbers get bigger when the lies get bigger.



"Build it and they will come."


$80B in AI lab revenue, $2.2T in AI infrastructure spending. The dreamers need demand to quadruple just to meet the depreciation on this spend.


The bust is going to be epic.



August 17, 2026


Buying gold for 21 consecutive months


In H1 2026, China's gold imports were up 138%. The PBoC has been buying for 21 straight months, with purchases amping up on declining prices.



AI fakery


Sure, a single OpenAI data center can generate 15x the current ARR of the entire company and more than Apple (AAPL) makes in a year, or 2x the sales of Costco.


Why is Barrons reporting such obvious nonsense? This NVDA pump is getting silly.



August 19, 2026


Gold miner short interest at decade highs


I'm not a fan of these ETFs but good golly, that's a sentiment tell!



Is AI becoming TBTF?


ChatGPT was launched in November 2022.


The hyperscalers spent trillions on AI infrastructure, invested downstream on AI labs, all to capture ~$100 billion in yearly inference spend.


And this $100 billion in ARR is set to shrink as open weight models and edge inference eat into the pie.


The goal was never revenue - it was to build until the sector became Too Big to Fail.


During the sub-prime bubble, the famous Ninja/ liar loans needed a guy with a pulse to sign the paperwork. Similarly, the AI labs need customers just for the optics.


The goal is a data center in your backyard, financed by private credit that's now TBTF.



Bitcoin lift-off


As expected, we got epic momentum out of the channel. Ethereum and Solana have confirmed the move.


There is ZERO fundamental reasons for bitcoin to go higher, so if it continues higher, something is up. Likely a sign of massive liquidity to come.


Bessent wants to lower bond yields and save the Japanese Yen at the same time. Fed money printing is a quick fix for both.


Bitcoin isn’t good for much, but it is a great liquidity barometer.




August 20, 2026


At the height of the AI bubble, ETF investors plowed money into financials sector ETF (XLF)

The passive bid is tired of propping up memory stocks.



Bitcoin


Bitcoin: Never argue with a channel breakout that is also a trend reversal. There is literally no resistance until $74.7k.


How much higher can it run? Who knows.


Can it give it all back tomorrow or next week? Who knows.


No position.



The dragon rises


While everyone's distracted by the Trump tweet of the day, offshore Chinese Yuan has been steadily climbing vs the USD.


The CNH (offshore) is consistently stronger than the PBoC controlled CNY (mainland).


Little known fact: China has ruled that offshore trusts are subject to Chinese taxes.


There is quiet pressure to bring capital home. Is that driving yuan strength, or something else?



August 23, 2026




This will reverse hard. And soon. Both markets are overvalued but India is less overvalued now.


August 24, 2026



Color me unsurprised.


Super Micro Computer (SMCI) was a great short (+70%) for me in 2024. Everyone chased it because it had a historic RSI of 100. I spotted it for the fraud that it is.



August 25, 2026


When US equities tank, as is inevitable with the bursting of the AI bubble, the world will feel the pain.


US equities are the most over-owned assets on the planet.


Long US equities is the most crowded trade on the planet.


No amount of Fed money printing can overcome The Great Unwind when foreigners dump US dollars, bonds and stocks.


India went through something similar in 2022-26.


Initially, Indian mutual funds bought whatever foreign investors dumped. Then the rupee started to sink faster, foreigners accelerated their selling, and the doom loop kicked in.


You can either look stupid before the crash or after. There is no in-between.



Apple gaining share in edge inference


Apple (AAPL) is making fast inroads in edge inference, the Achilles heel of the neoclouds and NVDA.


"Mac Studio provides a complete, end-to-end platform for AI development — from experimentation and training to deployment."


The AI economy needs $1T in annual consumer spending, flowing wholly to the frontier models and then upward to data centers, Nvidia and the memory layer.


What does this actually mean?


Imagine buying a bank of fax machines on credit, leasing prime downtown office space, then waiting for customers to show up to send/ receive fax messages.


Then a company decides to install a fax machine in its office.


Sure, the utilization isn't 100%, but the machine is reliable, doesn't deliver moral lectures or ask you to go to sleep, and keeps your data siloed within your own environment.


It is good enough for most fax machine users, which is all you need to gain market share.


If you think the fax machine analogy sounds ridiculous, look up Zapmail, FedEx's $350M failure in the mid-'80s.


Apple has always played the Fast Second game well.


The neo-clouds (CoreWeave (CRWV), Nebius (NBIS), IREN (IREN), Hut8 (HUT), Cerebaras (CBRS)) are doomed.



August 26, 2026


"I mean, if you look at their [Anthropic] revenue growth, $65 billion is their run rate. They were $9 billion last year. That's a sevenfold increase. I do think that market opportunity is out there." - Connor Group to Yahoo Finance


The logic: $30T TAM doesn't sound crazy because ARR grew 7x to $65B.


The math: $30 trillion is 461.5 times $65 billion. It is 30x the revenue of the global software industry.


I always knew AI was capable of hallucinations, but the switchover to infecting humans is new.



August 27, 2026


Nvidia, like the US government, keeps making promises it cannot keep while going deeper into debt.


Absolutely insane that people are focused on revenues instead of looking a few paras below the headline.



The company has added an additional section to its off-balance sheet obligations - $108.5 billion in guarantees.


Like parents co-signing a lease, NVDA has to guarantee its customers obligations in order to book sales.


Total obligations now stand at a record $530.5 BILLION.

Total cash on the balance sheet: $22.4 billion.


It is simply impossible for Nvidia to pay its bills when they come due.


And no one is talking about this.


That's the opportunity.



Like what you read? Get more exclusive content by subscribing to my premium newsletter!


Good Trading!

Kashyap




Comments


bottom of page