📢BREAKING : Some contractors involved in the Obama Presidential Center project have reportedly raised concerns about delayed payments ahead of the facility's opening in Chicago. Construction costs have reportedly climbed close to $1 billion, adding to scrutiny surrounding the project.

💻The AI Buildout- I broke the trade into 10 categories — MY top 3 picks in each with their stock prices as of last market close (~6/11/26): Bookmark this & also tell me what I missed/what are your favorite names? 1. GPU/Compute Silicon $NVDA $205 | $AMD $178 | $AVGO $436 2. HBM/Memory $MU $1,010 | $SNDK $720 | $WDC $314 3. Optical/Photonics $COHR ~$350 | $LITE ~$430 | $AAOI ~$150 4. Networking/Interconnect $ANET ~$155 | $ALAB ~$210 | $CRDO ~$165 5. Foundry/Packaging $TSM ~$300 | $ASML ~$1,050 | $AMKR ~$38 6. Servers/Hardware $DELL ~$170 | $SMCI $31 | $HPE ~$28 7. Power/Electrical $VRT $289 | $ETN ~$390 | $GEV ~$940 8. Power Generation/Grid $CEG ~$370 | $VST ~$215 | $TLN ~$420 9. GPU Clouds/Neoclouds $CRWV $102 | $NBIS ~$140 | $KEEL ~$5.50 10. Data Center REITs/Connectivity $EQIX ~$780 | $DLR ~$185 | $GLW ~$95

AI Data centers in North America consumed nearly 264 billion gallons (1 trillion liters) of water in 2025. This massive consumption figure, highlighted by market research firm Mordor Intelligence, is equivalent to the entire annual water demand of New York City. The issue has become a flashpoint as roughly 63% of the United States experienced severe drought conditions during the same period.

Can we utilise the heat released by data centers in generating electricity?

🚨 BREAKING! SHOCKING RUMORS EXPLODING RIGHT NOW—WE'RE GONNA RUN OUT OF DATA CENTERS BY 2028 AND THE INTERNET AS WE KNOW IT IS DOOMED! 😱

☀️ Solar Power Surpasses Coal in the U.S. However, reports on this milestone are accompanied by a 404 error. The original source is unavailable to verify figures, dates, or context. This underscores the importance of rigorously verifying information. Despite the lack of access, discussions about renewable energy continue to grow daily. The transformation of the power system is ongoing, but clear data is essential. Stay informed about the future of energy.

For this round of AI data centers, I’m focusing on the bill of materials: copper is being consumed across everything from rack power distribution, busways, transformers, and UPS systems to liquid cooling pumps and valves. A 1 GW AI campus can consume up to approximately 500,000 tons of copper. By 2026, demand for data centers alone is projected to reach 475,000 tons. As computing power expands, the ultimate impact lands on the BOM—copper, an old material, has suddenly become a shadow curve of AI capital expenditures.

Many companies have already outpaced their free cash flow with massive capital expenditures on data centers and infrastructure. While these companies maintain strong balance sheets and high credit ratings, a key risk lies in whether AI revenue can keep pace with spending growth. Accelerated earnings will support growth; delays could put pressure on credit metrics and refinancing. And if current capital allocation trends continue, the eventual AI bubble could be significantly larger than the dot-com bubble of the late 1990s. History doesn't repeat, but it often rhymes.

Due to issues at Tesla’s Supercomputer data center, resources have been leased to Anthropic. Instead of using Colossus 1—the massive AI supercomputer Musk boasted about building quickly—Tesla has leased its computing resources to Anthropic. The facts are that SpaceX encountered latency and hardware compatibility issues in its large-scale AI training network, prompting it to lease out computing capacity to external parties like Anthropic to generate revenue. This is less a case of a completely failed data center and more a situation where the infrastructure proved bottlenecked for training cutting-edge AI models, leading to its repurposing as an external leasing and revenue stream. Simply having a large number of GPUs does not guarantee effective AI training. When generations such as H100, H200, and GB200 are mixed and network synchronization becomes unstable, even the latest accelerators fail to deliver their full performance. Ultimately, the core of AI infrastructure lies not in the number of chips, but in design, networking, and homogeneity. In other words, a loosely assembled cluster won’t suffice—what’s needed is a homogeneous cluster of the latest GPUs paired with a high-performance network. Moving forward, the importance of high-performance networking capable of rapidly and reliably connecting GPUs will grow even beyond the significance of the GPUs themselves.