π Whatβs up, fellow Capitalists β
Welcome to the very first issue of Prompt Punk β the AI business newsletter for business people.
Hereβs what I promise:
β
Signal, not noise β the most important AI business moves, minus the jargon
β
Opportunities β tools, tactics, and trends to help you close deals, or outsmart slower people
β
Irreverence and humour β because in the long run, weβre all dead.
Letβs dive inβ¦
π₯ Biggest Story of the Week
π OpenAI drops GPT-5 β and itβs gunning for Google Search and Microsoft Office
OpenAI just launched GPT-5, its most advanced model yet. Itβs faster. It codes better. It writes better. It gives smarter health advice. And it hallucinates less β so it lies to you more politely.
But hereβs the real story: Chat is replacing Search. And Office.
Need research? Chat.
Need to write? Chat.
Need to think? Honestly⦠Chat.
Just writing this newsletter, I used AI hundreds of times β to pull quotes, build charts, do calculations, rewrite headlines, the whole shebang. And Iβm not alone. Any company doing knowledge work will end up buying something like ChatGPT for every employee.
Why? Because using an LLM is faster, easier, and (soon) cheaper than the Frankenstein combo of Google + Docs + Excel.
GPT-5 isnβt a product update. Itβs a platform shift.
π° Whereβs the Money Going?
π§ From $29B to $500B: OpenAI Hits Escape Velocity
Itβs OpenAI week in AI. Thereβs a planned employee stock sale that values OpenAI at half a trillion dollars. Thatβs up from $29B just two years ago. Incredible. But donβt get too jealous, OpenAIβs unique non-profit compensation model caps employee stock growth at 10-20x. Rich, yes, Filthy rich, probably not.
Why it matters:
AI companies of all stages will continue to be a magnet for talent when compensation can hit these levels. Legacy software companies will have a hard time hiring. Who wants stability when thereβs a seat on a rocket ship.
πΈ Anthropic wants $5B. Mistral wants $1B. The model wars are devouring capital.
Everyoneβs raising. Everyone thinks their model will win. How is this possible?
Anthropic is reportedly raising $3β5B to scale Claude
Mistral is seeking $1B to stay in the game.
Why it matters:
In the last 6 months the market has shown there can be multiple winners for different markets. OpenAI leads in chat, Anthropic in code, Google in video, Midjourney/BackForest in images, Grok in science, Mistral in on-premise? The markets are bigger than we dreamed 24 months ago.
OpenAI is on track for 13.6B in annualized revenue by the end of 2025, surpassing AirBnB and Spotify. So...if these markets are so huge, why not plough in some more cash. We can always rely on our friends from McKinsey to confirm our biases with an outrageous statistic: AI could be worth at least $13.6 trillion to the global economy. If thatβs the case, I might as well sell my children and buy NVIDIA stock (not investment advice π )
ποΈ Hyperscaler CapEx hits $350B. Thatβs not a typo.
In 2023 Amazon, Microsoft, Google, and Meta dropped $153B on AI infrastructure. In 2025? Projected to hit $350B.

Hereβs why the hyperscalers canβt spend fast enoughβ¦.
βAs fast as we actually put the capacity in, itβs being consumed β¦ We could be helping more customers and driving more revenue β¦ if we had more capacity.β
βWhile we brought additional data center capacity online this quarter, demand remains higher than supply. Even as we continue bringing more data center capacity online, we currently expect to remain capacity constrained through the first half of the fiscal year.β
βWith this strong and growing demand for our Cloud products and services, we are increasing our investment in capital expenditures in 2025 to approximately $85β―billionβ¦β
Why it matters:
This isnβt hype. Itβs real demand β and itβs straining supply chains. Whoever controls compute, wins. Why else would OpenAI unleash the ginormous Stargate datacenter project.
π₯ Clay raises $100M to kill boring sales work
Never heard of Clay? Thatβs because your company blocks tools made after 2015. Donβt you just hate those information security policies. π
Clay helps teams find leads, enrich them, and automate outreach. They just raised at a $3.1B valuation β and theyβre hiring in Europe.
Why it matters:
AI in sales is no longer a RevOps fantasy. Itβs happening. And if youβre not using tools like Clay, your competitors are.
π Whoβs Winning with AI?
βοΈ Lawyers are loving AI. You read that right.
A personal story: Outside an event at my daughterβs school I spoke with a Partner at a law firm about how they were using AI. Of course, they were already using Legora and loving it. It sped up research, review, and prep. I am not surprised people love the product, I once met the CTO of Legora at a dinner, the guy was totally cracked.
βIt hasnβt changed everything. But weβre 20% faster. And less burnt out.β
Why it matters:
AI isnβt replacing full roles. Itβs replacing what people hate doing. And itβs happening bottom-up β one task at a time.
π§ Jason Lemkin cloned himself. You can now talk to him.
SaaStrβs Jason Lemkin, our personal tech sales guru turned over decade of blogs, talks, and decks into a chatbot. Itβll answer anything from comp plans to when to hire a CRO.
Why it matters:
This is what productizing content looks like. Every company with a dusty Google Drive full of training material should be paying attention.
πΈ $854B invested in AI. $120B earned. When will it flip?
At Prompt Punk we like to keep on ROI on all this AI craziness. Letβs see where we stand. Since 2023:
$854B invested into AI (infra, models, chips, startups)
$120B in revenue (mostly to NVIDIA, OpenAI, AWS, Accenture)

Why it matters:
It looks bad. But this is how infra gets built. The smart moneyβs now in:
Apps built on top of that infrastructure
Firms who help enterprises deploy AI
Anyone close to chips, power, or latency
π Accenture clocks $1.8B in GenAI revenue, but How?
Thatβs not a typo, a projection or promise. Thatβs money in the bank. Projects range from Telecoms to Food Processing to Oil and Gas. Use cases include customer service, business intelligence, supply chain and more. Accenture CEO explains why it is workingβ¦
Our clients continue to be focused on reinvention and Gen AI is a catalyst for reinvention..For our clients, the twin themes of achieving both cost efficiencies and growth continueβ¦The number of clients embracing Gen AI is increasing significantly and we are starting to see some tangible examples of scale in data and AI.
Why it matters:
Generative AI is an excuse to get big things done in slow moving organizations.
π§ Whoβs Hiring and Moving?
π Mistral expands to Singapore
The open-weight French model shop is now hiring in Singapore, off the back of a partnership with the countryβs Ministry of Defence.
Why it matters:
Smart geopolitical move. Itβs about more than talent β itβs about building outside U.S. and EU.
𧨠Job of the Week: Granola, Founding AE β London
Granola is a fast-scaling AI notetaker. Reportedly at $10M+ ARR, now hiring a founding sales rep in the UK.
Why it matters:
Real product, real traction. If you want to sell into a hot category, this is worth a look.
π οΈ Whatβs New in Tech?
π§© OpenAI drops GPT-OSS β and goes open(ish)
OpenAI released an open-weight model: GPT-OSS, running on AWS and other clouds.
Why it matters:
This is a chess move. It helps OpenAI:
Compete with Meta + Chinese open models
Unshackle from Microsoft
Win back developer trust
Donβt be fooled β itβs strategic, not spiritual. Sam Altman wants to be on the right side of history.
π» King Claude fights to retain the AI coding crown
Anthropic has released Claude Opus 4.1. Better coding. Better docs. Better reasoning.
Why it matters:
30M+ devs and builders are going to use AI coding assistants. Claude has become the one you actually want in a workflow. Now letβs calculate 30m multiplied by $2,400 per year = $7.2E10
π³π΄ OpenAI to launch European data center in Norway
OpenAI will bring 100,000 GPUs online by 2026 in Norway β their first in Europe.
Why it matters:
Solves EU data sovereignty
Reduces political risk
Positions OpenAI as a full-stack platform, not just an API layer
This is OpenAI preparing for world domination β one GPU cluster at a time.
π Coming Next Week
Can AI replace an inbound sales rep?
Iβve been testing ElevenLabs-powered voice agents. Next issue:
What works
What breaks
The use cases that are closer than you think
Iβll share it all!
β If you liked this:
Forward it to someone who likes the technology business
Hit reply and give me feedback or complete the poll below
What do you think of today's newsletter?
See you next week.
β John