8/17: Nvidia + OpenAI + SB Energy = PORTS-Pike

It’s Monday. Over the weekend we saw some amazing discourse between Gavin Baker of Atreides Management, Sholto Douglas and Dario Amodei of Anthropic, and David Sacks of the All-In Podcast on AI and concentration of power. All parties agreed that there is much more work to be done in ensuring that AI does not overly concentrate power in the hands of just a few people. We covered it this morning on MTS. Be sure to monitor with us live on X and YouTube, and follow us on Instagram.

In October 2025, Roon posted “ill believe we’re over investing in computational substrate once i see real interest rates above even 3%”. Today, the real market yield on 30-year treasuries stands at 3.00%, the highest level since the Great Recession. There are many reasons for this: a large federal deficit, inflation remaining above target, the Fed not cutting rates, and a new one: AI hyperscalers borrowing so much money to fund compute buildouts that they are raising the cost of capital of the US government.

In economics, “crowding out” is a phenomenon where large-scale government borrowing in order to fund deficit spending soaks up available savings and drives up interest rates for private companies. Right now, we’re seeing the reverse. Hyperscalers, data center providers, and AI infrastructure projects have issued $308 billion through July to fund compute buildouts, nearly 13x the same period last year. Because data centers are long-lived assets, much of this debt is long-dated, directly competing with 10- and 30-year Treasury bonds.

There’s another reason to expect AI to raise real interest rates, explained by economists Trevor Chow, Basil Halperin, and J. Zachary Mazlish in their paper “Transformative AI, existential risk, and real interest rates”. If people expect to be much richer in the future, they will borrow against the future and save less today. If they expect AI doom, they will front-load consumption. Either way, lower available savings today forces the interest rate upward. When the first version of the paper was released in January 2023, real interest rates stood at 1.47%, substantially lower than today. So either markets are efficient but believe transformative AI is 30-50+ years away, or markets are incredibly inefficient, and we should expect real rates to rise. Right now, we may be seeing the first glimmers of this hypothesis being borne out.

• Nvidia, OpenAI, and SB Energy are partnering on a massive Ohio datacenter. SB Energy will build the facility, known as PORTS-Pike, on the site of a decommissioned uranium enrichment plant on federal land in rural southern Ohio. OpenAI will lease the facility, paying rent and energy costs to SB Energy and buying chips from Nvidia. Nvidia is the exclusive chip provider of the facility in exchange for guaranteeing up to $105B in lease and power payments to SB Energy if OpenAI can’t pay. Nvidia will also invest $1.5B in SB Energy1. The datacenter is expected to open in phases beginning in 2028 and scale to 8 GW of computing capacity, powered by 10 GW of new power generation.

• Anthropic’s revenue run-rate passes $65 billion, meaning the company is expected to generate annualized revenue of $65 billion based on its current performance. In July 2025, Anthropic had $4 billion of run-rate revenue. In Q2, Anthropic reported over $11.5B in revenue, compared to $787M in Q2 2025, and more than the entirety of Anthropic’s 2025 revenue. Anthropic also reported positive adjusted operating income.

• Cursor releases Origin, a GitHub competitor for code hosting. The early beta, available for paid users, includes Git features like repos, pull requests, and code browsing, and allows you to sync with your GitHub repos. More agent-native features will be released over time. Some may remember that GitHub released the first mainstream AI programming tool (GitHub Copilot) in October 2021, a full year and a half before Cursor. Now, Cursor is competing directly with GitHub’s core business model2.

• Meta will stand trial over accusations of addicting minors to their platforms. 29 state attorneys general are seeking penalties of up to $20,000 per violation, which could add up to hundreds of billions of dollars or more3, as well as other penalties like data and algorithm deletions and mandatory design changes. Google, Snap, TikTok, and other tech giants are facing similar lawsuits. The trial will be presided over by US District Judge Yvonne Gonzalez Rogers, who also presided over Musk v. Altman in May.

• Google buys Spirit Airlines’ data for $10 million. Spirit started operations in 1983 as Charter One Airlines and ceased operations on May 2, 2026. The data includes emails, Teams messages, spreadsheets, calendars, and more, and will be used for AI training. Google outbid a $7.5 million offer from Mercor.

• Alibaba launches music generation model HappyShrimp 1.0, competing with other music gen models like Suno and Udio.

• Ron Alfa (Co-founder and CEO of Noetik)

• Tim Hwang (Founder, Institute for a Christian Machine Intelligence)

• Campbell Hutcheson (Senior Researcher, Epoch AI)

• Martin Borch Jensen (CSO, Gordian Biotechnology)

• Mitchell Troyanovsky (Co-founder, Basis)

• Keller Cliffton (Co-founder and CEO, Zipline)

• Nalin Semwal (Co-founder, Capy) and Haris Mehrzad (Head of Growth, Capy)