Why Ohio could host a $500B AI campus

The Frontier

Science & tech breakthroughs in AI models, hardware, and computing that define what's possible

Nvidia may back an Ohio data center with up to $105 billion. Nvidia plans to back an Ohio data-center project with up to $105 billion in funding, according to the New York Times. The facility, which would be leased by OpenAI, could cost as much as $500 billion and rank among the world’s largest data centers.

Nvidia’s up-to-$105 billion backing makes financing—not chips—the immediate constraint on an Ohio project that could cost $500 billion. The test is whether OpenAI discloses funding commitments tied to its Ohio lease next year.

New York Times

OpenAI joins the PORTS-Pike project. OpenAI said it has agreed to secure approximately 8 gigawatts of IT capacity at the PORTS-Pike Technology Campus in Pike County, Ohio, alongside SB Energy, Nvidia and the U.S. Department of Energy. The six-year buildout through 2032 is expected to create 35,000 construction jobs and 2,500 operating jobs.

OpenAI’s 8 GW agreement makes delivered IT capacity—not job projections—the key benchmark for PORTS-Pike. The announced buildout runs through 2032; watch for capacity delivery by then.

Openai

China wants its data to power global AI. China is seeking to expand its influence over global AI systems by exporting data alongside AI models, according to the New York Times. The effort raises concerns that Chinese state narratives could become embedded in chatbots built or trained using Chinese data sources.

Huawei’s hardware-and-standards push was narrower: China’s bid to pair NLP models with censored data could shape overseas AI training inputs. Watch for ASEAN chatbot pilots to disclose use of Chinese training data by 2027.

New York Times

Model Developments

Latest AI model releases, capabilities, and technical breakthroughs shaping the competitive landscape

Claude will begin watermarking AI-generated text. Anthropic said it will modify Claude’s text-generation process to embed a detectable watermark using small randomized choices in output generation. The change is intended to comply with European Union regulation, though it raises questions about whether the altered process could affect text quality.

Anthropic is making Claude watermarking an EU compliance differentiator ahead of the December mandate. Private-key verification creates a concrete output-checking mechanism; the test is whether rivals offer comparable detection by December.

The Guardian

Capital & Control

Tracking the flow of capital and influence that shapes the tech landscape

Nvidia discloses a $21 billion SpaceX stake. Nvidia disclosed ownership of nearly 123 million SpaceX shares valued at almost $21 billion at the end of June, according to an SEC filing. The holding originated from Nvidia’s January investment in xAI before Elon Musk combined the AI company with SpaceX; SpaceX has said it will use Nvidia equipment exclusively for its data centers.

Nvidia’s $21B SpaceX stake turns a supplier relationship into direct balance-sheet exposure: it benefits from SpaceX’s value while SpaceX uses Nvidia equipment exclusively for data centers. The test is whether that exclusivity translates into sustained equipment demand.

Ars Technica

Shein targets a valuation of up to $27 billion. Shein is targeting a valuation of as much as $27 billion in a Hong Kong initial public offering, according to the Financial Post. The company built its business through low-priced, trend-driven clothing sold globally and shipped directly from suppliers.

Shein’s $27 billion IPO target turns on whether Hong Kong investors will pay up for its direct-from-supplier clothing model.

Financial Post

Infrastructure & Power

The physical backbone of AI: compute, data centers, supply chains, and the geopolitics of energy

Microsoft’s AI ambitions may be constrained by chip deployment. A Guardian investigation identified an apparent gap between Microsoft’s stated AI-capacity ambitions and the number of advanced AI chips reported to be operating in its data centers. Microsoft reportedly targeted 1.8 million AI chips installed across its data-center network, raising questions about the pace at which planned capacity has been deployed.

Microsoft’s 1.8 million installed-chip target is not usable AI capacity: the Guardian-identified gap with reported operating chips makes deployment—not procurement—the constraint. The test is whether the next disclosure shows operating chips closing that gap.

The Guardian

Digital Defense

Threats and defenses shaping the global security landscape

The defender’s window is narrowing. OpenAI said the OpenAI-Hugging Face incident exposed how AI capabilities could rapidly expand the capacity of typical threat actors. The company said models are increasingly able to automate elements of cyberattacks, including finding vulnerabilities in human-written software and exploiting overlooked permissions, while calling for accelerated defensive modernization.

OpenAI’s incident makes model-registry attestation more urgent than source-code secrecy: a tainted checkpoint can reach thousands of downstream projects without code exposure. Watch for attestation to solidify by 2027; the test is whether breach costs stay above $4.5M.

Openai

The Performance Edge

Peak performance science for sustained execution in high-stakes environments

New guidance urges fewer unnecessary shoulder scans. A Monash University evidence review published in JAMA Internal Medicine concludes that most shoulder problems improve through self-management, education and time. The guidance challenges routine early imaging and surgical intervention, recommending a more conservative approach for general practitioners and physiotherapists.

Monash University’s JAMA Internal Medicine review puts GPs and physiotherapists on notice: for most shoulder problems, self-management, education and time—not early scans or surgery—should be the default. The test is whether early-imaging referrals fall within 12 months.

Medicalxpress

Market Pulse

Daily market data: indices, sectors, top movers, and volatility

Markets eased as energy outperformed. The S&P 500 fell 0.4%, the Nasdaq 100 slipped 0.1%, the Dow Jones declined 0.5% and the Russell 2000 dropped 0.5%. Energy led sectors with a 1.2% gain, while Consumer Staples lagged at -1.6%; SMCI fell 4.7%, META declined 3.8% and MSFT dropped 3.2%.

1.2% energy gain looks defensive, not a broad-market turn: all four major indexes fell as Staples lost 1.6% and SMCI, META and MSFT declined. Watch whether energy still leads by the next market close if large-cap tech remains under pressure.

Yahoo Finance