Latest / The Cloud Business Podcast with Fexingo: AWS, Azure, GCP, and Enterprise Infrastructure / How Cloud Contracts Now Mandate Carbon-Aware Data Placement
Transcript
- Lucas: If you signed a new cloud contract this year, there's a good chance it includes a clause that would have sounded like science fiction five years ago: the provider now commits to placing your data in a specific region based on the real-time carbon intensity of the local power grid. Luna: So the provider is essentially saying, 'We promise to store your data where the electrons are cleanest right now.' That's a pretty big shift from the old model where the customer chose the region and the provider just promised uptime. Lucas: Exactly. The old approach was carbon offsetting — buy enough renewable energy credits to match your usage, and call it green. But that didn't change where the compute actually happened. Carbon-aware data placement is different. It ties the physical location of your workload to the carbon intensity of the local grid at that moment. Luna: And that's a contractual guarantee, not just a feature you can opt into. Which providers are actually writing this into enterprise agreements? Lucas: All three hyperscalers have pilot programs, but Microsoft Azure is furthest along. They've been offering carbon-aware scheduling for batch jobs since 2024. What's new in 2026 is that several large enterprise contracts now include a clause that data placement itself — not just compute scheduling — will be carbon-aware, with a specific carbon-intensity threshold baked into the SLA. Luna: Is the threshold uniform across providers, or is each one choosing its own number? Lucas: It varies. The most common figure I've seen in contracts this year is 250 grams of CO2 equivalent per kilowatt-hour — that's roughly the average carbon intensity of the US grid. Below that threshold, the provider considers the region green enough. Above it, they commit to moving the workload to a cleaner region if possible. But here's the catch: the clause applies only to workloads that the contract designates as "carbon-placement eligible." Luna: Which workloads qualify? I'm guessing not your latency-sensitive real-time systems. Lucas: Right. Typically it's batch processing, data analytics, machine learning training jobs — anything where a few milliseconds of added latency doesn't break the application. A good example is a multinational retailer I spoke with last month. Their Azure contract covers all their batch analytics workloads with a carbon-intensity cap of 250 grams. So when their European data centers spike because of a coal-heavy hour, those analytics jobs get routed to a region in, say, Scandinavia or Quebec where hydropower is plentiful. Luna: But that retailer also has data residency requirements in Europe. If the analytics contain personal data, can they legally move it to Canada? Lucas: That's the tension. The carbon-aware placement clause typically includes a carve-out: the provider can only move data within the same geographic data-sovereignty zone. So the retailer's European data stays within the EU plus maybe the European Economic Area. Azure has a specific list of regions that satisfy the GDPR adequacy standard. So the carbon-aware routing is restricted to that subset. Luna: Which means the provider needs a pretty granular carbon signal for each region within that zone, and they need to update it in near real-time. Lucas: Exactly. AWS and Google Cloud both use the real-time carbon intensity data from Electricity Maps, which aggregates grid data globally. Microsoft uses its own Emissions Impact Dashboard plus third-party data. The contract typically says the provider will use "commercially reasonable efforts" to place data in the lowest-carbon available region within the approved geo-boundary. It's not a hard guarantee every second — it's a best-effort commitment, but with a carbon-budget cap that the provider tracks quarterly. Luna: And if they exceed that cap? Is there a penalty? Lucas: Yes, and this is where it gets interesting. Some contracts now include a carbon service credit mechanism, similar to uptime credits. If the provider exceeds the agreed carbon-intensity threshold for more than a certain percentage of compute hours, the customer gets a credit on their bill. I've seen thresholds like: if more than 10 percent of eligible workload hours run in regions above 250 grams, the customer gets a 5 percent credit on that line item. Luna: So the provider has a financial incentive to actually route workloads cleanly, not just claim they're doing it. Lucas: That's the theory. And the early data suggests it's working. A study from Uptime Institute earlier this year looked at five large enterprises with carbon-aware placement clauses and found that, on average, their eligible workloads saw a 30 percent reduction in carbon intensity compared to a static region assignment. But the study also noted that the carbon savings depend heavily on the diversity of regions available within the sovereignty zone. Luna: So if you're a company with data that can only stay in the US, and your contract limits you to US regions, you're at the mercy of a grid that varies wildly from Iowa's wind-heavy mix to the coal-heavy PJM region. Lucas: Right. The US is tougher for carbon-aware placement than Europe or Canada because the grid is more fragmented and dirtier on average. Some CIOs I've talked to are pushing for a lower threshold — 200 grams — for their us only workloads, because they know there are regions like the Pacific Northwest that can hit that during certain hours. But the providers push back, arguing that 200 grams would trigger too many data movements and potentially degrade performance. Luna: So the negotiation is really about balancing the carbon ambition with operational stability. What about Google Cloud? They've been the most aggressive on carbon-intelligent compute, but are they writing placement into contracts? Lucas: Google does have a carbon-aware load-shifting feature for their data analytics and AI workloads, and they've been including it in some large enterprise contracts since late 2025. But their approach is slightly different: rather than a fixed threshold, they commit to a quarterly carbon-budget target for the customer's eligible workloads, and they use their internal model to optimize placement. The customer gets a quarterly report showing actual grams of CO2 per kilowatt-hour per workload. If they exceed the target, Google issues a carbon credit. Luna: That actually sounds more flexible. But it also puts more trust in the provider's model. Lucas: Exactly. And that's the debate right now in procurement teams. Do you want a hard threshold with a simple binary — green or not green — or do you want a more nuanced carbon-budget approach that the provider optimizes on your behalf? I think we'll see both models coexist for a while. Luna: Let me ask about the edge case: what happens if the only eligible regions for a workload all exceed the threshold at the same time? Does the workload just not run? Lucas: The contracts I've seen include a fallback: if no region meets the carbon-intensity cap within the sovereignty zone, the workload runs in the lowest-carbon region available, and the provider does not count those hours toward the penalty threshold. So there's a kind of force majeure for the grid. But the customer can also set a maximum latency tolerance — if moving to a cleaner region would cause unacceptable latency, the workload can stay put and the provider logs the exception. Luna: So it's not a perfect system, but it's a huge step forward from where we were even two years ago. I'm curious what the adoption looks like. How many enterprises actually have these clauses? Lucas: Still a minority. A survey by CloudZero in Q1 2026 found that about 12 percent of enterprises with annual cloud spend over $10 million had a carbon-aware data placement clause in at least one of their contracts. But that's up from essentially zero in 2024. And among the largest spenders — over $100 million annually — it's more like 35 percent. So it's scaling fast, especially in Europe and among companies with public net-zero commitments. Luna: And I imagine this is also a way for the providers to differentiate in a market where compute is increasingly commoditized. Lucas: Absolutely. When pricing on raw compute is basically a race to the bottom, adding a carbon-aware placement guarantee is a way to justify a premium. One AWS enterprise sales rep I spoke with off the record said that these clauses are becoming a "check-box item" in RFPs for sustainability-minded clients. If you don't offer it, you're not even considered. Luna: That's a big shift. And it ties into the broader conversation we keep having on this show about how cloud contracts are becoming more like partnership agreements than simple utility contracts. Lucas: You know, it's funny you say that. Because we put a lot of work into researching these contract trends, and we deliberately don't run ads on these episodes. If you want to support that choice, the link is buy me a coffee dot com slash fexingo. That's all — no perks, just knowing the content stays independent. Luna: Yeah, it matters to us that these conversations are useful for what you're building or running, without a sponsor shaping the narrative. So thanks for listening. Lucas: Alright, back to the contracts. There's one more nuance I want to flag: the carbon-intensity data itself. The accuracy of these clauses depends entirely on the quality of the grid data. If the provider is using hourly averages but the grid actually shifts every 15 minutes, the placement decision might be based on stale information. Luna: So the temporal granularity matters. Are providers moving toward sub-hourly data? Lucas: Some are. Google has been using 5-minute granularity for their carbon-intelligent compute since 2023. AWS and Azure are at 15-minute intervals for their placement models. But the contract language is usually written in terms of "average carbon intensity over the billing hour" — which is less precise. So sophisticated procurement teams are starting to ask for the granularity to be included in the SLA. Luna: That's the kind of detail that separates a well-negotiated contract from a greenwashing one. Any final advice for listeners who might be renegotiating a cloud contract soon? Lucas: Yes. Three things. One: ask for a carbon-intensity threshold that matches your company's own sustainability targets, not just the provider's default. Two: make sure the sovereignty carve-out is clearly documented, so you know exactly which regions are in play. Three: ask for a quarterly carbon report broken down by workload, so you can verify the placement is actually happening. Don't just trust the billing line. Luna: Good advice. I'd add: start with a pilot workload that's not mission-critical, test the carbon savings and the operational impact for a quarter, then expand. Lucas: That's smart. And on that note, I think we've covered the essential ground. Next episode, we'll look at how these carbon-aware clauses interact with the new SEC climate disclosure rules that take effect next year.