Inside the math: How OpenAI’s $700B makes sense

👋 What’s up compadres!

This is Prompt Punk — the best AI sales newsletter! 😉

TL;DR: inside today’s newsletter:

Let’s rip…

OpenAI has everyone’s head spinning with unprecedented data center deals. 5GW (gigawatts) with Microsoft, 4.5GW with Oracle, 10GW with NVIDIA, 10GW with Broadcom, 6GW with AMD, 0.5GW with CoreWeave.

That’s enough electricity to power the United Kingdom. The entire country.

For the uninitiated, data centers are compared in their power usage, since square meters and number of installed chips don’t allow for apples-to-apples comparisons.

The cost of all this build out? $700B+

Does this make sense? Yes, if you believe the growth curve. And you should.

Is there demand? Hell, yes. Google CEO Sundar Pichai talked about AI usage growing 50x YoY on Gemini. OpenAI’s revenue is going parabolic, adding more than $2B in annual recurring revenue every month. Now reportedly at 800M active weekly users (about to surpass TikTok).

And we’re not just talking chatbots. We’re talking images, video, drug discovery, drones, robotics, military and more. Just see the compute usage required for each AI application below.

Ok, so how does 700B make sense?

First, this isn’t $700B on OpenAI’s balance sheet. Most of the concrete, copper, and GPUs sit with partners (Microsoft, Oracle, CoreWeave). OpenAI commits capacity, not pure capex.

Second, the revenue mix is getting richer. More enterprise seats, more applications built on their API, an advertising network, robotics and military applications can quickly add up to $100B+ in revenue.

At scale isn’t this tech worth more than search advertising? ($175B)

Let’s look at the math….

So….spending $700B over 5 years to earn $30B+ in profits per year is a 23x multiple. Roughly the same valuation multiple as Google. Not crazy. The risk is in the execution. Can OpenAI pull it off and get to escape velocity?

Bottom line. If AI just stayed as a “chat” interface, these amounts would be reckless. It’s not. In my opinion, this CAPEX spend is justified.

But OpenAI better keep adding billions in ARR per month. More consumers, more enterprise, more $ per user.

AI generated launch videos for any product, service or business in 72 hours.

What they did
Absurd pairs creative directors with autonomous AI agents to plan, script, generate, and edit cinematic launch videos. They pick a mix of models for motion, faces, and lighting, then stitch scenes into a story you can ship in ~72 hours.

See their own launch video below…

Result
Their first client launch video pulled 450K+ views on X, giving the team a high-velocity “hero asset” they could clip into ads, posts, and prospecting messages. (Speed + story = oxygen for a launch week.)

Why it matters (for sales teams)
A punchy launch video isn’t just brand it's a reply generator. Video can help build a narrative around your product. Now imagine custom videos for your top customers. Dynamite!

I wish I had this tool when I was running my SaaS startup. Service companies, SaaS builders and advertisers are wiring YourAtlas into their lead forms so every new lead gets an AI phone/SMS call in <60 seconds.

It’s a pretty basic Startup idea but it works.

The AI qualifies on the spot, and books a meeting or warm transfer, 24/7.

Why isn’t every company using this?

Results
Customers report 3× more conversions, 10× Return on Ad Spend and +30% website conversion when speed-to-lead is automated.

Testimonial data shows 75–85% connection rates.

How are end customers winning?
Business moves faster, no longer waiting days to schedule a call with a rep.

Customers of Your Atlas didn’t just “add a chatbot”, they operationalized response time. Form submit → Atlas dials/texts instantly → qualifies → books to calendar or transfers live, with prebuilt integrations and a speed-to-lead dashboard to tune the funnel.

Why it matters (for sales teams)
Speed is the conversion moat: contacting within 5 minutes can make teams 100× more likely to connect, within 1 hour is ~7× likelier to qualify than later. Most orgs still take hours or days! It’s 2025 not 2005.

What they did
Rolled out Recruit 360, an AI/ML prospecting program that sifts 30M+ applicant files across ~1,700 variables to hand recruiters refined prospect lists.

Result

How they won
Treated recruiters like a revenue team: contact-within-48-hours SLA, prioritized queues, and new pipeline visibility.

Why it matters
Recruiting is sales with a different contract. AI-ranked leads + disciplined follow-up = more qualified conversations per recruiter and fewer misses.

Commercial teams running giant inbound/outbound funnels can steal this playbook tomorrow. Score leads on intent + eligibility, then enforce SLAs and iterate.

What it does: AI meeting agent that joins Zoom/Meet/Teams, captures real-time notes, action items, and summaries, then syncs them where work happens (Calendar, Docs, Salesforce/HubSpot). Live guidance on calls included.

Example: Discovery call → highlights + next steps auto-captured → pushed to CRM → follow-up draft ready before you hang up.

Why it’s valuable: Never lose signal from conversations. Faster follow-ups and cleaner CRM with coaching baked in.

Website: otter.ai

What it does: A self-driving browser with AI companions that click, scroll, and type across tabs to research, source data, and automate repetitive web tasks.

Example: Hand it a target list → it visits sites, pulls the useful bits, compiles a sheet, and drafts follow-ups. No copy-paste marathons.

Why it’s valuable: Turn tedious web work into repeatable automations so you can spend time where it moves the number.

Website: strawberrybrowser.com

Amazon Quick Suite is a new AI workspace from AWS. You type a question it searches your company’s data, explains the answer, builds a chart if needed, and can take actions in your apps (create a ticket, update a record, send a summary).

This puts Quick Suite in the same space as Glean, Microsoft 365 Copilot, Google Gemini Workspace.

It bundles chat + research + business intelligence and automation in a single screen. It plugs into AWS data (S3, Redshift) and popular SaaS (Salesforce, Google Workspace, Microsoft 365, Snowflake, Atlassian, etc.)

Why it matters:
Enterprise AI internal knowledge bases are an AI killer app, just like consumer chat, image generation, research, coding agents and workflow agents. Every hyperscaler needs one. AWS has theirs.

Microsoft’s internal forecasts show data-center shortages dragging into 2026.

To ease the crunch, Microsoft aligned with a $40B consortium buying Aligned Data Centers (~5GW portfolio) and struck an expanded deal with Nscale to deliver ~200k Nvidia GPUs across Texas, Portugal, and Norway—helpful, but not instant relief. Microsoft is also adding new regions in Asia, yet power/land/builder bottlenecks keep timelines stubborn.

Why it matters:
If your 2026 plan assumes “capacity loosens,” rethink it. Demand continues to outstrip supply. The AI boom continues.

Europe raised $13.1B in Q3 (flat QoQ, +22% YoY). The headline: 40% ($5.2B) went to AI, up from $2B a year ago driven by big rounds like Mistral ($2B) and Nscale ($1.1B). Early-stage is the engine (~60% of all €) while late-stage is thin (~9% of global late-stage). No OpenAIs or xAIs in Europe. Still…Europe isn’t just talking AI, money is flowing especially in frontier, infra, and applied enterprise AI.

Why it matters:
Budgets and hiring are real across the EU/UK, but skew earlier-stage. Still missing BIG late stage dollars.

Which one are you?

Thanks for reading! Respect! 👊

If you liked this:

📨 Forward to a friend
💬 Hit reply and tell me what you want more (or less) of
📊Vote in the poll below 👇

— John
Prompt Punk