Contrarian & Deep-Cut Hunt

What happened: A Reuters review found more than 700 GW of very-large-load requests across parts of the Midwest, Mid-Atlantic and South, but highlighted duplicate and underfinanced requests. ERCOT separately reported more than 438 GW of requests, 89% from data centers. Exelon said stricter collateral reduced its high-probability data-center demand tally by about 40% to 11 GW.

Why it earned coverage: The post-cutoff investigation supplied a cross-region comparison and an observable queue-shrinkage receipt after stricter collateral.

Transmission: Deposits, site control, ownership disclosure and study fees force developers to demonstrate capital and project maturity. Queue attrition changes probability-weighted utility capex, generation need and supplier backlogs even if headline requested load remains enormous.

What happened: Bitdeer paid approximately $100 million cash for about 200 acres near its Rockdale, Texas facility. The company says the Milam County sites have 563 MW currently interconnected with plans to scale to 742 MW; the acquisition removes lease-renewal risk and gives outright site control for financing and developing AI/HPC infrastructure. The additional 179 MW is planned, and no AI tenant was announced.

Why it earned coverage: Independent monitoring surfaced a primary-filed transaction that the prior briefing omitted. It is retained as catch-up because the cash outlay changes site control and financeability, not because the 3.0 GW headline is new demand.

Transmission: Fee-simple ownership eliminates ground-lease renewal risk and can improve collateral and long-duration financing, but value is created only if planned power becomes firm and a creditworthy AI tenant funds or signs capacity.

What happened: A Bank of Canada staff article found no broad change in the labor market's overall structure but said job seekers may find it harder than in 2019 to secure work in the most AI-exposed occupations. It identifies lower job-finding rates rather than higher separations as the early channel. Statistics Canada separately reported 19.2% of businesses used AI in the prior 12 months in 2Q26, triple the 2024 share.

Why it earned coverage: A central-bank labor-flow analysis provides a falsifiable middle path between 'no labor effect' and mass-layoff claims, with implications for entry-level software, clerical and customer-service work.

Transmission: Firms can reduce vacancies or raise output thresholds before firing incumbent workers. That depresses job-finding rates and career entry while leaving aggregate separations initially unchanged, delaying the signal in headline unemployment and earnings.

What happened: Independent analysts argued that AI financing increasingly resembles vendor-supported reflexivity; Nvidia's official third-party platform supplied a primary receipt for the mechanism, though not proof of a bubble.

Why it earned coverage: The strongest contrarian case is no longer that demand is fake. It is that commitments, financing capacity and recognized profitable demand are being conflated.

Transmission: If capital formation runs ahead of cash-producing workloads, refinancing and residual values become the clearing mechanism.

What happened: Moonshot released full Kimi K3 weights for a 2.8T-parameter model. Independent analysis placed it near the proprietary frontier, while public terms showed a roughly 1.56 TB download, $3/M input and $15/M output API pricing, and separate-agreement requirements for large model-as-a-service businesses.

Why it earned coverage: A Chinese lab combined near-frontier performance with actual weight release, then exposed the serving and license frictions behind the word 'open.'

Transmission: Near-frontier weights shorten the capability lead and broaden diffusion, but model size, reasoning-token intensity and commercial terms preserve hosting economics and deployment barriers.

What happened: OpenAI said Astra meets its Critical cybersecurity threshold, reported that privileged internal access found two zero-days in a recent-vulnerability set and produced working exploit chains against a hardened browser/OS environment, and said advanced cyber capability will initially be limited to trusted users with additional monitoring and safeguards.

Why it earned coverage: OpenAI replaced a possible Critical assessment with a formal designation and disclosed concrete zero-day/exploit evidence plus an access-control plan.

Transmission: If models can discover and operationalize unknown vulnerabilities, labs must restrict tools, users and execution environments. That slows or segments deployment while increasing demand for monitoring and defensive automation.

What happened: Crusoe was no longer the developer for a proposed 1.8 GW Cheyenne project, but Black Hills said the project was not paused, remained targeted for early-2028 service and had received $201M of refundable customer contributions to reserve generation equipment.

Why it earned coverage: Contradictory headlines about a Crusoe pause were resolved by the utility's project-level receipt.

Transmission: Utility tariff, direct customer relationship and funded long-lead equipment can preserve project continuity even when an intermediary developer exits.