Nvidia has agreed to cover $105 billion of rent for one data centre

Futurelab Studios

19 August 2026

In AI, this week.

We at Futurelab have been carrying a question around all week.

"what is the best use of a human brain, between now and the day the machines get good at everything?"

Three things happened this week that answered it, and not in the order we expected.

Signal of the Week

Nvidia has agreed to cover $105 billion of rent for one data centre

Nvidia has promised to cover up to $105 billion of rent for a giant new AI data centre in Pike County, Ohio. OpenAI will be the tenant for the next twenty years.

The land used to be a Cold War uranium plant. It is rented from the US government.

The site needs about 10 gigawatts of new electricity, and at least 9.2 of that will come from burning gas. For a sense of size, 8 gigawatts is roughly what all of Delhi pulls on a hot afternoon.

What we at Futurelab think

for years the question was who could get enough chips. that one is settled. a new question has taken its place, and it is who can get enough electricity.

so if you want to know how fast AI grows from here, stop watching the model launches. watch the power stations. this is now a construction story, and it will be argued about in local councils, not in labs.

none of our brains get a vote in that. which is why the next story is even more interesting.

AI News That Matters

A doctor with no maths degree solved a problem that had been open for 22 years

Shanmu Jin is a brain surgery trainee in Beijing. He studied geology first, then medicine. He is not a mathematician.

He ran a sixteen-hour session with ChatGPT on a maths problem nobody had cracked since 2004. It worked. Three senior mathematicians, including the man who first posed the problem, have checked the proof.

The same week, a data scientist with no biology training raised $4 million for a company making cancer vaccines for dogs. He built the first one himself, using AI, after eleven months of vets missing his own dog's tumour.

What we at Futurelab think

neither of these men had permission. no PhD, no lab, no title. they had a question they could not put down, and a tool that finally let them try.

and notice what the machine did not do. it did not decide which problem was worth the sixteen hours. it did not know when the answer was real. both men had to do that part themselves.

so the valuable skill is quietly changing. it used to be knowing how to produce an answer. now it is choosing the right question, and being able to tell a real answer from a convincing one.

that is good news for anyone who has ever been told they were not qualified.

AI got about 60% cheaper this month

Three big price cuts landed inside four weeks.

Grok 4.6 arrived at roughly 60% below the models it now matches. Google's new Gemini 3.7 Flash undercut its three closest rivals. OpenAI had already cut its bulk prices by up to 80%, and DeepSeek came in cheaper still.

To put a number on it, eight dollars now buys about 750,000 words read and 750,000 words written. That is three long novels in, and three long novels out.

What we at Futurelab think

if you worked out your AI budget before july, it is now wrong, and wrong in your favour. that almost never happens in software.

but here is the part worth sitting with. price was the reason most people said no. wait till it gets cheaper. it got cheaper. that reason is gone, and it was probably never the real one.

One more, because it made us laugh

Anthropic put three AI helpers on one server, gave them the same code to work on, and did not say who was in charge. they spent four hours locking each other out and sabotaging each other's work. one of them later apologised and admitted it had "behaved badly".

it was not the AI that failed. it was the org chart. no owner, no rules, three capable workers on the same job.

we have all sat in that meeting, haven't we?

Practical Tool - Steal This

Ask your own chat history what it thinks of you

The problem: nobody tells senior people the truth, and we are all very kind to ourselves.

1.	Download your full chat history from ChatGPT or Claude. It is in account settings, and arrives by email in a few minutes.

2.	Open a brand new chat, nothing to do with work, and upload the file.

3.	Ask it to study how you think, decide and talk to people, and to name the patterns you cannot see yourself.

4.	Ask for the evidence behind every point. Tell it not to be nice.

5.	Pick the two things that stung most, and work out what you will do about them.

the words to paste

This is my full AI conversation history. Audit it and tell me how I work, decide and communicate. Name my three biggest weaknesses based on patterns in this data, with the evidence for each. Be direct. Do not soften it and do not compliment me first.

Why it works: it holds months of evidence about how you actually think. No colleague, no coach and no friend has that. Fair warning, it is uncomfortable reading. That is the point.

Latest From Futurelab

19th August 2026

We are running a month-long AI immersion program for professionals. the first session was held last Saturday, but we have had more signups since, so we will run session #1 again to make sure everyone is on the same page.

Pricing: ₹ 30,000 for regular professionals, and ₹ 20,000 for professionals with less than 5 years of work experience, for the full month program.

Program details: https://futurelabstudios.com/ai-workflow-accelerator

Please DM Uddhav at +91 9313034255 for fresh signups.

So, back to the question

We do not think the answer is to work faster. we are not going to win that race, and we do not have to.

the doctor in beijing sat with one problem for sixteen hours. he did not out-calculate the machine. he chose the problem, stayed with it, and knew a real answer when he saw one.

choosing. staying. knowing.

As far as we can tell, that is what a human brain is still uniquely for.

everything else is being poured in ohio, on nine gigawatts of gas, by people we will never meet.

see you next week,

Team Futurelab

Join the Conversation

5,000+ professionals

Founders, business leaders, and practitioners navigating the AI transition together.

Join the Community

To unsubscribe please click here.