1. Nvidia PE falls to 14-year low.

"This is a key-person event, not a demand event. Nobody is canceling GPU orders because the CEO resigned. Historically, stocks take time to find a bottom after an abrupt CEO exit while the market works through the investigations and the search”.

Nvidia reports quarterly earnings on Tuesday, with investors watching whether its sharp growth can survive without Huang.

2. Wall Street withstands the AI blues.

In the past month, waves of selling have engulfed the artificial intelligence trade that amounts to more than $2 trillion in erased market value.
Yet, despite all this, the S&P 500 is only around 3% off its all-time high.
The generally accepted view has long been that if tech crumbles, Wall Street will implode as well. But this has not happened.
The “average stock” is actually holding up quite well. The equal-weighted S&P 500 hit a record high last week and is up nearly 5% since October 6. It is outperforming the more widely followed benchmark index so far this year too, up 16% compared with 11% for the S&P 500.
This resilience is due largely to rotation, as investors have switched into more defensive, value-oriented sectors.

3. “Buy the drop!”, says Morgan Stanley.

“Given the recent market drop, we see buying opportunities among AI enablers, infrastructure, and adopters”.
“In our view, Texas election results will not impact '27 data center construction, but the new DC review process will shift ~20% of US projects from '28 to '29.
Our Powering AI model shows >30% powered DC shortfall through '28, while MS US data center builds remain ~flat from '27 to '28, with ~40% growth in '29.
Expect an even higher importance on exponential compute efficiencies from the labs to enable tech diffusion across the economy.
We see the market moves as an overreaction and think the ultimate impact will be a push-out of some projects from 2028 to 2029, rather than a derailing of the capex story. This has the potential to elongate the AI capex cycle, which has positives including: a smoother supply-demand rebalancing across the data center supply chain, improved credit outlook, and higher lease rates for compute.”

Below: Total token volume growth continues to be exponential

4. Foreigners continued to be net buyers of the KOSPI for a fourth consecutive day.

The steady upward trajectory of tech heavyweights suggests investors remain focused on underlying global demand for high-bandwidth memory and next-generation chips, viewing the domestic giants as insulated from Nvidia's internal executive shakeups.

Below: The definition of cheap… Korean chips at 2.8x forward earnings!!!

5. US powered shell providers at attractive risk/reward level.

Powered shell providers (PSPs)—often written as "power shell" providers—are data center companies that build and lease the physical building, land, and high-capacity electrical infrastructure to artificial intelligence (AI) and cloud tenants without buying or operating the computer servers themselves.
Many prominent PSPs, such as Hut 8 and TeraWulf, are former Bitcoin mining companies. Because cryptocurrency mining already required massive electrical grid connections and industrial cooling, these firms repurposed their power-advantaged sites to meet the severe electricity shortages facing the AI boom.
Instead of acting as traditional cloud operators, PSPs function essentially as specialized industrial landlords for high-performance computing (HPC).
• What the Provider Supplies. The PSP secures the real estate, grid interconnects, substations, cooling systems, and the exterior data center building (the "shell").
• What the Tenant Supplies. The tenant—typically a major tech hyperscaler or an AI lab—brings in, owns, and operates its own expensive graphics processing units (GPUs) and server racks.
• Contract Structure. PSPs sign long-term triple-net or colocation leases that typically span 10 to 20 years, locking in predictable, recurring cash flows.

There is a big difference between neoclouds (GPU rental firms) and PSP’s:

6. Terawulf is Barron’s favourite PSP.

“TeraWulf, the former Bitcoin miner turned data-center operator, offered weary investors some proof that the artificial-intelligence infrastructure payoff wasn’t just a distant promise. And for the company, that momentum is just getting started.

TeraWulf, which has shifted its operations to developing AI data centers from Bitcoin mining, signed a 20-year lease with Anthropic in July valued at about $19 billion in contracted revenue.

TeraWulf needs significant financing to build out its data centers, sparking investor concerns that the company may need to issue additional shares to cover its construction costs, diluting existing shareholders.

Analysts have an average Buy rating and an average price target of $30.72 for TeraWulf, suggesting about a 205% upside from Thursday’s close, according to FactSet data.”

Morgan Stanley is a buyer of Terawulf, Hut8, Riot and Cypher.

Not a subscriber yet?