Latest / Ctrl AI Profit / Ep. 140 | Chevron Just Bet 20 Years on AI Power
Transcript
- 0:20SPEAKER_00: So Chevron just signed a 20-year deal with Microsoft to power an AI data center.
- 0:26SPEAKER_00: 20 years.
- 0:27SPEAKER_01: That's not a contract.
- 0:28SPEAKER_01: That's a marriage.
- 0:29SPEAKER_00: A marriage where the dowry is natural gas and the offspring are your future AI bills.
- 0:35SPEAKER_01: Let me set the scene.
- 0:37SPEAKER_01: Chevron is going to supply natural gas to a Microsoft data center in West Texas for two decades.
- 0:44SPEAKER_01: This isn't a pilot program.
- 0:45SPEAKER_01: This isn't let's try this for a year and see.
- 0:48SPEAKER_01: This is we are betting the next 20 years of AI compute on fossil fuels.
- 0:53SPEAKER_00: And that should make every small business owner's ears perk up because when Microsoft is locking in energy for 20 years, they're telling you something.
- 1:02SPEAKER_01: They're telling you they believe AI demand is permanent, not a trend, not a bubble, permanent enough to justify a utility-scale energy commitment that extends well beyond any current technology cycle.
- 1:14SPEAKER_00: Think about what a 20-year contract means in tech years.
- 1:18SPEAKER_00: Twenty years ago, the iPhone didn't exist.
- 1:21SPEAKER_00: Cloud computing barely existed.
- 1:24SPEAKER_00: And Microsoft is saying whatever AI looks like in 20 years, we'll still need this much power for it.
- 1:30SPEAKER_01: That's the infrastructure story that nobody's really digesting.
- 1:33SPEAKER_01: AI isn't just software anymore.
- 1:35SPEAKER_01: It's a decades-long energy commitment.
- 1:38SPEAKER_01: It's pipelines and power plants and land deals.
- 1:40SPEAKER_00: It's the kind of commitment you make when you're building a railroad, not an app.
- 1:45SPEAKER_01: And railroads have tolls.
- 1:48SPEAKER_01: Everyone who ships goods on the railroad pays.
- 1:51SPEAKER_01: Everyone who uses the data center, which means everyone who subscribes to an AI service, pays.
- 1:57SPEAKER_00: So let me connect the dots.
- 1:59SPEAKER_00: Microsoft signs a 20-year energy deal.
- 2:02SPEAKER_00: That energy powers AI data centers.
- 2:05SPEAKER_00: Those data centers run the AI tools you subscribe to, your co-pilot, your Chat GPT, your automation platforms.
- 2:13SPEAKER_00: The cost of that energy flows downstream to your monthly bill.
- 2:17SPEAKER_01: Compute requires power.
- 2:19SPEAKER_01: Power requires fuel.
- 2:21SPEAKER_01: Fuel requires long-term contracts.
- 2:24SPEAKER_01: Long-term contracts lock in rising prices.
- 2:27SPEAKER_01: And those prices get baked into every API call, every subscription tier, every usage-based billing model.
- 2:35SPEAKER_00: Right now, AI companies are mostly eating those costs.
- 2:38SPEAKER_00: They're subsidizing usage to drive adoption.
- 2:42SPEAKER_00: That's not sustainable.
- 2:44SPEAKER_01: Not even close.
- 2:51SPEAKER_01: But a 20-year energy contract says the training wheels are coming off.
- 2:56SPEAKER_01: They're planning for the era where AI has to pay for itself, including the power bill.
- 3:01SPEAKER_00: So what does that mean for the business owner paying 40 bucks a month for ChatGPT Plus or 20 bucks for copilot?
- 3:09SPEAKER_01: It means those prices are going up, not might.
- 3:12SPEAKER_01: Will.
- 3:13SPEAKER_01: The question isn't if your AI costs increase, it's how fast and how much.
- 3:19SPEAKER_00: And there's another layer.
- 3:20SPEAKER_00: Chevron is supplying natural gas, not solar, not wind, natural gas.
- 3:26SPEAKER_01: These data centers need power that's reliable and always on.
- 3:30SPEAKER_01: Solar and wind are great, but the sun sets and the wind dies.
- 3:34SPEAKER_01: Natural gas plants run 24-7.
- 3:37SPEAKER_01: When you're committing to 20 years of AI compute, you need 24-7 power.
- 3:42SPEAKER_00: So the biggest tech company in the world is choosing fossil fuels over renewables to power the future of AI.
- 3:48SPEAKER_01: It's a pragmatic choice.
- 3:50SPEAKER_01: Microsoft has made huge renewable energy commitments, but when it came to guaranteeing uninterrupted AI compute for two decades, natural gas won.
- 3:59SPEAKER_01: Reliability beats sustainability.
- 4:01SPEAKER_00: And that tension isn't going away.
- 4:03SPEAKER_00: Every AI company is going to face this same choice.
- 4:07SPEAKER_00: Do I go green and risk downtime or go with what works and take the PR hit?
- 4:14SPEAKER_01: Most of them will choose what works.
- 4:15SPEAKER_01: An AI service that goes down loses customers instantly.
- 4:19SPEAKER_01: An AI service that uses fossil fuels, most customers never know.
- 4:23SPEAKER_00: I want to dig into the location.
- 4:25SPEAKER_00: West Texas.
- 4:27SPEAKER_00: Why there?
- 4:28SPEAKER_01: Cheap land, existing natural gas infrastructure, and proximity to the Permian Basin, one of the most productive oil and gas fields in the world.
- 4:36SPEAKER_01: Chevron can pipe gas directly.
- 4:38SPEAKER_01: The fuel is literally in the ground next door.
- 4:41SPEAKER_00: So this is a data center built on top of the fuel source.
- 4:44SPEAKER_01: It's the digital equivalent of building a factory next to the mine.
- 4:47SPEAKER_01: You cut out transportation costs and supply chain risk.
- 4:50SPEAKER_01: Chevron and Microsoft are vertically integrating energy and compute, which tells you how serious this energy problem is.
- 4:57SPEAKER_00: They wouldn't go to these lengths if the existing power grid could handle what they need.
- 5:02SPEAKER_01: It can't.
- 5:03SPEAKER_01: Data center power demand is growing faster than grid capacity in most regions.
- 5:08SPEAKER_01: That's why these companies are doing deals directly with energy producers instead of just buying from the local utility.
- 5:13SPEAKER_00: Let's talk about what the small business owner should actually do.
- 5:16SPEAKER_00: This isn't just a headline, it's a financial planning issue.
- 5:20SPEAKER_01: Three things.
- 5:21SPEAKER_01: First, lock in your current AI subscriptions at annual rates right now.
- 5:26SPEAKER_01: Month-to-month pricing is going to get more expensive faster.
- 5:30SPEAKER_01: Second, audit your AI usage.
- 5:33SPEAKER_01: Which tools are actually driving revenue?
- 5:35SPEAKER_01: Which ones are nice to have?
- 5:37SPEAKER_01: When prices go up, the nice to have become the first thing you cut.
- 5:41SPEAKER_01: Know which ones those are before the price increase forces the decision.
- 5:45SPEAKER_01: And third, start exploring local and open source AI options for tasks that don't need cloud level compute.
- 5:53SPEAKER_01: If you're using GPT, four for something a local model can handle, that's a cost you can insulate from the energy price spiral.
- 6:01SPEAKER_00: That third point is huge.
- 6:03SPEAKER_00: Document summarization, email drafting, basic data analysis, a lot of that can run locally.
- 6:09SPEAKER_01: And local AI is getting better fast.
- 6:12SPEAKER_01: Open source models that fit on a laptop can handle most routine business tasks.
- 6:16SPEAKER_01: The gap between cloud and local AI is narrowing.
- 6:19SPEAKER_00: Plus, local means no data center, no energy contract, no usage-based pricing.
- 6:25SPEAKER_00: You're not paying Microsoft's natural gas markup.
- 6:29SPEAKER_01: A single large AI data center can draw as much power as a small city.
- 6:34SPEAKER_01: Multiply that by the dozens being built right now, and you see why Microsoft needs a 20-year gas contract.
- 6:39SPEAKER_01: The grid wasn't built for this.
- 6:41SPEAKER_00: Small city power consumption for what amounts to a single building.
- 6:46SPEAKER_00: That's staggering.
- 6:47SPEAKER_01: The math is straightforward.
- 6:49SPEAKER_01: More compute means more energy.
- 6:52SPEAKER_01: More energy means higher costs.
- 6:55SPEAKER_01: Higher costs mean higher prices.
- 6:58SPEAKER_01: It's not fear-mongering, it's supply chain economics.
- 7:02SPEAKER_00: I've been through enough business cycles to know what a cost wave looks like.
- 7:06SPEAKER_00: This is a cost wave.
- 7:07SPEAKER_00: It's building right now and it's going to hit in the next 12 to 24 months.
- 7:12SPEAKER_01: Your AI subscriptions go from$40 to$60.
- 7:15SPEAKER_01: Your usage-based API bills double.
- 7:17SPEAKER_01: Your unlimited plans start having caps.
- 7:20SPEAKER_00: The free tiers disappear.
- 7:22SPEAKER_00: The businesses that plan for it will absorb those costs and keep competing.
- 7:26SPEAKER_00: The ones that didn't, playing catch-up while their margins get squeezed.
- 7:30SPEAKER_01: This deal also signals that the biggest players are locking in their energy costs now.
- 7:35SPEAKER_01: Microsoft is essentially prepaying for power at today's rates for 20 years.
- 7:40SPEAKER_01: Small businesses can't do that.
- 7:42SPEAKER_00: No kidding, we can't sign a 20-year energy contract.
- 7:46SPEAKER_01: But you can lock in annual AI subscription rates, buy compute credits in bulk, reduce your dependence on the most compute-intensive services.
- 7:54SPEAKER_01: It's the same strategy, just scaled down.
- 7:57SPEAKER_00: You can't negotiate a natural gas deal, but you can negotiate your software contracts.
- 8:03SPEAKER_00: And the time to do it is before everyone else figures out what the Chevron Microsoft deal really means for pricing.
- 8:11SPEAKER_01: Because once the market catches on, the deals get worse.
- 8:14SPEAKER_00: The AI boom just became an energy story.
- 8:19SPEAKER_00: And energy stories always end up on your bill.
- 8:23SPEAKER_00: Every single time.
- 8:24SPEAKER_00: That's the episode.
- 8:26SPEAKER_00: Lock in your AI subscriptions, audit your stack, and start planning for higher costs.
- 8:31SPEAKER_00: This is Control AI Profit, and we'll talk to you next time.