Money and trust ran through this week. Nvidia brought in Wall Street to pay for AI data centers, and Anthropic signed a long lease with a former bitcoin miner. Anthropic also started marking what Claude writes, and courts in England and Wales decided smart glasses stay at the door.
Nvidia lines up $500 billion in financing with six investment firms
Nvidia announced on August 10 that six firms, Apollo, Blackstone, BlackRock, Brookfield, Goldman Sachs and KKR, will help raise $500 billion to fund AI infrastructure, The National reported, citing Bloomberg. The plan uses computing power as collateral for loans and bonds. Jensen Huang called Nvidia's compute "an investable infrastructure asset." The report notes that some investors worry Nvidia's deals with its own customers inflate demand.
Our take: AI computing is being financed like real estate, so data center capacity will keep growing. That is good news for prices and availability of AI services, as long as the demand holds up.
Anthropic signs a $9.1 billion data center deal with a bitcoin miner
Riot Platforms signed a 20-year, $9.1 billion deal to supply 191 megawatts of computing capacity from its Texas campus, and Bloomberg reported the customer is Anthropic, according to Channel Insider. The first capacity is expected in late 2027. It is Anthropic's second large deal with a bitcoin mining company, which already has the power and land that AI data centers need.
Our take: The limit on AI right now is electricity, not ideas. Expect the AI tools you rely on to have busy periods and occasional outages for a while yet, and keep a backup tool for critical tasks.
Claude starts watermarking text and images
Anthropic plans to put invisible watermarks in text and images made by Claude models released after August 2, SiliconANGLE reported on August 11. The goal is to meet the EU AI Act's rule that AI content be marked. Images get C2PA content credentials, an open standard that records how a file was made. Anthropic says detection tools are coming but won't be perfect, especially on short or heavily edited text.
Our take: Assume AI-written text may be detectable, now or later. If you publish AI-drafted marketing or client work, edit it properly and be comfortable saying AI helped.
Courts in England and Wales ban Meta smart glasses
Court buildings in England and Wales will confiscate Meta smart glasses at the entrance and return them when the visitor leaves, Legal Cheek reported. Phones are still allowed as long as they are not used to record. The move followed a High Court case in which a litigant was found to have been "assisted or coached" through smart glasses during cross-examination, which he denied. Restaurants, pubs and New York's court system have made similar moves.
Our take: Wearable cameras are reaching your office, store and meeting rooms. Decide now whether recording glasses are allowed in client meetings or staff areas, and post the rule.
xAI releases Grok 4.6
xAI released Grok 4.6 on August 12, available through its API, the Cursor code editor and its own Grok Build tool, iPhone in Canada reported. The model handles up to 500,000 tokens of context. On the Artificial Analysis Intelligence Index it scores 61, the same as OpenAI's GPT-5.6 Sol, according to the report.
Our take: The top models are now roughly tied, so the choice comes down to price, privacy terms and how well a model fits your own tasks. Test two or three on your real work before you commit to one.
One thing to do this week
Take one task your team does with AI every week, like drafting proposals or summarizing calls. Run the same input through two different AI tools and compare the results side by side. You will learn more from that hour than from any benchmark chart.
Want a system like this in your business?
A discovery conversation is where we map it: what to automate first, where the human checkpoints go, and what it would take.
Book a discovery conversation →