Sept 24: Barclays "[The Oracle Force Majeure] is neutral from a credit perspective, in our view, given the possible capex offsets, and we do not believe ORCL spreads should move in any meaningful way on the headline today." Sept 25: ORCL CDS blows out +10 to a record 237bps

AI compute demand is going parabolic and we need more compute than ever before (Save this). OpenRouter’s chart shows weekly token usage growing 25 times in one year and doubling in the last month. That means more demand for GPUs, memory, networking equipment, electricity, cooling systems and data center capacity. The companies building this infrastructure could become some of the biggest beneficiaries of the AI boom. CoreWeave and Nebius are two of the clearest examples because they provide rented GPU capacity to AI labs, developers and enterprises that cannot build enough infrastructure themselves. CoreWeave is benefiting from strong demand from hyperscalers and AI companies that need access to GPUs quickly. Nebius is also expanding rapidly as customers compete to secure GPU capacity, power, and data-center space. NVIDIA benefits by selling the GPUs, while Micron and SK Hynix supply the HBM and DRAM that help those systems operate efficiently. Arista and Broadcom benefit from the networking equipment required to connect thousands of GPUs, while Vertiv and Eaton provide power-management and cooling infrastructure. The opportunity also extends to data center developers, utilities, construction companies, and equipment manufacturers. The AI usage will continue to grow as models become cheaper, more capable and increasingly integrated into coding, research, customer service, and business automation. When AI agents begin running continuously in the background of millions of users, inference demand will grow far beyond today’s chatbot usage. That would turn compute into a recurring infrastructure expense and make access to GPUs and power even more valuable. The companies with available GPUs, electricity, and data center capacity could maintain strong pricing power while the rest of the industry races to keep up. If you enjoyed reading this, make sure to follow @MelvinInvests for more AI infrastructure insights and if you want to see exactly what I'm buying as an analyst at Milk Road Pro, check out the link below.

Probably the strongest pro-data center case in the US is Loudoun County, VA. The fiscal numbers are real but it is also an outlier case. Loudoun says data centers sit on ~4% of commercial parcels and generate 38% of General Fund revenue. FY27 budget has the industry closer to ~$1.3B, or ~45% of local taxes. Loudoun taxes the equipment inside the buildings at $4.15 per $100 of assessed value, with a 60/45/30/15/10/5% depreciation schedule. JLARC, using older FY23 data, found mature Virginia markets ranged from <1% to 31% of local revenue from data centers - Loudoun was the 31%. There are limits to all the positive externalities (which doesn't mean they are negatives too btw). JLARC estimates a typical 250k sq ft facility supports ~50 full-time workers vs ~1,500 at peak construction, and a typical Dominion-system residential customer could see +$14-37/month of generation and transmission costs by 2040. Loudoun also ended by-right approvals in March 2025 over siting, noise and infrastructure, which I find interesting given it's the pinnacle example. To be transparent, the clip itself is advocacy. Innovation Council Action is a pro-Trump, pro-AI 501(c)(4); Axios reported plans for $100M+ of political spending in 2026. The speaker is a Loudon resident, though. We need voices on the ground speaking their truth and we should listen to them. I’d use Loudoun as evidence of how valuable data centers can be when a locality actually captures the tax base. I would not use it as evidence that the next 300 MW project in a different tax code, power market, and greenfield site will look always like Loudoun. There’s enough real economic upside here that I don’t think the pro-data center case needs exaggeration. The better argument is the factual one: these projects can bring enormous investment and tax revenue, but the benefits depend heavily on tax structure, power costs and where/how they’re built. Here for communities and partners (labs, hypers, sponsors) that do this the right way and quickly - we need more of them!

Every college syllabus should include these graphs. Use AI for homework, you will get it done faster and get a higher grade, and then get crushed on the exam.

Americans would rather have a nuclear power plant built in their community than a data center, this may be the worst corporate public-relations fumble of all time

In Loudoun Virginia home to more than 200 data centers, tax revenue from data centers has exploded. As a result, over the past 10years, they lowered the real property tax rate every year from $1.145 in tax year 2016 to $0.805 cents per $100 in assessed value for 2026.

Absolute state of this. Young lad purchases a baseball bat in "Modern England". As soon as he leaves the store, police grab him for "carrying a weapon". The bat is literally still in it's plastic covering.

When I first proposed a national moratorium on AI data centers it was called a radical idea. Not anymore. Let me thank the people of Prince William County, VA for standing up to the greed of Blackstone and stopping the construction of the largest data center campus in the world.

Ages of Founding Fathers in 1776: James Monroe, 18 Aaron Burr, 20 John Marshall, 20 Alexander Hamilton, 21 James Madison, 25 John Jay, 30 Thomas Jefferson, 33 Thomas Paine, 39 John Adams, 40 George Washington, 44 This nation was built by brilliant young men.

Data center forecasts need better evidence. SynMax tested Vulcan’s forecasts against 168 projects representing 11 GW of capacity. The result: when a reported online date came before Vulcan’s earliest satellite-based completion estimate, 99% of the GW did not come online on schedule. Read the full Vulcan blog to see why observed construction milestones matter for power demand and infrastructure forecasting: 🔗blog.synmax.com/vulcan/how-a… #DataCenters #PowerMarkets #EnergyInfrastructure #SatelliteIntelligence #InfrastructureForecasting #VulcanResearch

'Why is a flat-screen TV affordable and a college education not?...Because Congress has spent 60 years trying to make college affordable and has spent zero years trying to make TVs affordable.' By @reason's Aaron Brown, Michael Mendelson, @CliffordAsness thedispatch.com/article/affo…

If only ratepayers paid wholesale prices this would mean something to the people of California. Here’s the truth…

“California almonds use 4x-7x more water than all data centers in North America used combined in 2025.” Golf course owners:

🚨 Exposing California's corrupt "Stop Nick Shirley Act", instead of going after the fraudsters California is now going after the people exposing the fraud. This bill AB 2624 will: - Criminalize journalists with misdemeanors, $10,000 fines, imprisonment, and content takedown - Let immigrant based NGOs' funding be confidential - Take away freedom of the press from journalists - Protect any "immigration support services" information from being public (healthcare, legal services, etc) This bill was created by the Attorney General's WIFE Mia Bonta to stop fraud from being exposed. Please like and share this video everywhere! By trying to silence and intimidate journalists, they are trying to hide the truth from you. EXPOSE ALL THE FRAUD.

~40% of US data center projects may be delayed due to labor, permitting, and power constraints. Leveraging data from SynMax, the @FT recently released an article highlighting the growing gap between AI investment and infrastructure delivery. Read the full article: hubs.li/Q04cBXF30