The Biggest Bottleneck in AI

Dear Readers,

AI is racing ahead – but the pace is no longer set by the next model leap, but by the infrastructure: electricity, networks, water. As parameter figures explode, transformer stations, power lines, and cooling circuits are becoming the real battleground. The exciting question today is: Who is scaling smarter – not just with GPUs, but with reliable energy and short connection times?

In this issue, you'll take a look at the biggest bottleneck in the AI economy and its consequences: why PPAs, location selection, and water concepts are suddenly product features; how new models are shaking up European software values; why hedge funds are pushing back into big tech; and how emerging markets could benefit from AI value creation. Plus, there's a future benchmark to try out, fresh OpenAI signals, the chart of the day, and a quick poll on energy dominance. Ready for clear findings and tangible takeaways in just a few minutes of reading time? Then read on.

In Today’s Issue:

All the best,

The AI revolution has an inconvenient bottleneck: energy, grid connections, and water. It's not GPUs that are slowing things down first—it's transformer stations, power lines, and cooling towers. In short, AI can only scale as fast as the infrastructure can grow.

What's the issue? Data centers for training and inference are energy-intensive and require reliable cooling. The IEA expects the electricity demand of data centers to roughly double by 2030 – a pace that puts pressure on grid expansion and approvals. At the same time, water is becoming a location factor; providers are countering with liquid and direct/hot water cooling, and even “zero-water cooling.”

For the AI community, the race is shifting: models, data, and computing power count – but whoever secures electricity and water contracts first will win time to market. This affects startups (location, PPAs), hyperscalers (nuclear and renewable deals), and politics (grid planning, approval times).

The good news is that the solutions are on the table: load flexibility, waste heat utilization, on-site generation, water reuse. The question is less “if” than “how fast” we build – and who will coordinate it.

Why it matters: Without predictable electricity and water use, AI growth stalls, costs rise, and innovation slows down. Those who think of infrastructure as a product feature scale more reliably – and more sustainably.

Sources:

🔗 https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centres-while-offering-the-potential-to-transform-how-the-energy-sector-works?utm_source=chatgpt.com

🔗 https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centres-while-offering-the-potential-to-transform-how-the-energy-sector-works?utm_source=chatgpt.com

🔗 https://www.deloitte.com/us/en/insights/industry/power-and-utilities/data-center-infrastructure-artificial-intelligence.html?utm_source=chatgpt.com

🔗 https://emp.lbl.gov/queues?utm_source=chatgpt.com

When the founder who sold his last company to Zillow for $120M starts a new venture, people notice. That’s why the same VCs who backed Uber, Venmo, and eBay also invested in Pacaso.

Disrupting the real estate industry once again, Pacaso’s streamlined platform offers co-ownership of premier properties, revamping the $1.3T vacation home market.

And it works. By handing keys to 2,000+ happy homeowners, Pacaso has already made $110M+ in gross profits in their operating history.

Now, after 41% YoY gross profit growth last year alone, they recently reserved the Nasdaq ticker PCSO.

Invest for $2.90/Share

_{Paid advertisement for Pacaso’s Regulation A offering. Read the offering circular at invest.pacaso.com. Reserving a ticker symbol is not a guarantee that the company will go public. Listing on the NASDAQ is subject to approvals.}

A new benchmark has been introduced that has not existed in this form before. The benchmark tests predictions about the future—and therefore cannot be trained using data. Try it out for yourself!

New, significantly more powerful models (including GPT-5 and Claude for financial services) have triggered a wave of sales at European software and data companies (SAP, Dassault, LSEG, Sage).

The market is pricing in the risk that GenAI will attack established software margins and data moats; highly valued stocks are becoming vulnerable to downgrades. For stock pickers, deep workflow integration is now more important than mere data sovereignty.

13F data shows that Bridgewater, Tiger Global & Co. have massively increased their holdings in Nvidia, Microsoft and Alphabet. This increases concentration and crowding risks, but could support multiples in the short term – especially if the Fed eases policy soon. For portfolios, this means consciously managing factor exposures (momentum/quality/growth) and hedging cluster risks.

Asset managers are rotating into emerging market stocks with direct AI exposure (e.g., semiconductors, model providers). AI-driven companies are the biggest contributors to the EM index in 2025; some firms are explicitly shifting their EM strategies toward the AI value chain. Signal: Structural capital flows into EM tech, higher factor concentration, and new alpha beyond commodities/China cycles—but with crowding and geopolitical risks.

Working on the future of finance through AI? From trading algorithms to credit modeling and fintech infrastructure, we’re looking for work that explores the intersection of AI and financial systems.

Submit your paper or project to Superintelligence, the top AI newsletter with 200k+ readers, by emailing [email protected] with the subject line “Finance Submission”. We’ll contact you if we’d like to feature it.

Looks like OpenAI has a much better model internally

Jimmy Apples argues that GPT-5 is actually greatly underrated and that opinions on this will change in the very near future.