Latest / The Cloud Business Podcast with Fexingo: AWS, Azure, GCP, and Enterprise Infrastructure / Cloud Contracts Now Penalize Data Gravity Waste
Transcript
- Lucas: If your cloud contract doesn't have a data gravity waste penalty clause yet — give it six months. It's coming. Luna: Data gravity waste — that's a new one. I've heard of egress fees, storage tier penalties, but waste? Lucas: Yeah, it's a newer breed of fine print. The idea is simple: hyperscalers have noticed that a huge portion of stored data in their clouds is never touched by compute. It just sits — consuming disk, replication, networking overhead, without generating any utilization revenue. Luna: So they want to charge you for data that's idle. Not just storing it, but for the gravity of having it there — the drag on their infrastructure? Lucas: Exactly. And the numbers are not trivial. I was looking at a recent negotiation between a big Fortune 500 retailer and Google Cloud. The retailer had about 8 petabytes of historical transaction data sitting in us-central1. None of it was being read by any active workload — it was kept for compliance and the occasional audit query. Lucas: Google proposed a new line item in the renewal — call it a 'data gravity maintenance fee.' It came out to roughly $2.3 million extra per year. Based on the ratio of stored data to compute hours in that region. Luna: Wait — $2.3 million for just leaving data there? Was it negotiable? Lucas: It was, but only if they agreed to either move the data to a colder region — which would increase latency if they ever needed it — or set up a scheduled compute job that touches each data block at least once a month. Basically, prove the data isn't completely dead. Luna: Huh. So the hyperscaler is basically saying, 'If your data is a corpse in our data center, we're charging you a funeral fee.' Lucas: That's one way to put it. And look — these clauses are showing up across AWS and Azure too. AWS has something similar in their Enterprise Discount Program renewals, though they call it a 'storage utilization adjustment.' It's less aggressive than Google's, but the direction is the same. Luna: Right — and if these shows are useful for what you're building or running, a couple of dollars a month is genuinely what keeps these going. Buy me a coffee dot com slash fexingo, if you've gotten something out of them. Lucas: Absolutely. Even a few bucks helps us keep digging into these contract details. So back to the data gravity angle — why now? Why are hyperscalers suddenly getting aggressive about this? Lucas: I think it's a combination of two things. First, storage density has outpaced compute density in their data centers. They've got racks and racks of NVMe and SSDs, but the CPU utilization in those same zones is flat or declining. So they need a lever to either get you to use compute on that data, or move it out. Luna: Second being: AI training workloads are compute-heavy but they don't always touch historical data. So if you're storing a decade of clickstream logs and never using them for models, that's dead weight. Lucas: Exactly. And the hyperscalers are under pressure from their own investors to show return on capital. Building a data center costs billions. They want every petabyte to earn its keep. So data gravity waste becomes a negotiable term in the contract — but only if you know to ask about it. Luna: So what should a cloud buyer — say a VP of Infrastructure or a CFO — look for in their next renewal? Lucas: First, look for any language that ties storage costs to compute activity. Phrases like 'active data ratio,' 'compute-touch frequency,' or 'data gravity factor.' Second, ask your cloud provider for a data gravity report — show me which regions have a low compute to storage ratio. They all have that data internally. Luna: And if they won't share it? Lucas: Then you run your own analysis. Use tools like AWS Storage Lens or Azure Storage Insights. They can show you which buckets haven't been read in the last 90 days. That's your vulnerability. If you've got 30 percent of your storage untouched, you're looking at a potential penalty. Luna: So the remedy is either to delete that data, archive it to a cold tier, or set up a dummy compute job that touches it. Which do you recommend? Lucas: It depends on the data. If it's genuinely needed for compliance, archive to something like Amazon S3 Glacier or Azure Archive. That costs pennies per gigabyte but has retrieval delays. But if the data has potential value for analytics or AI, then set up a scheduled job — a small serverless function that reads a few blocks per day. That can satisfy the 'compute-touch' requirement. Luna: I've also seen companies adopt a data-streaming architecture as a result. Instead of storing everything in one region, they shard data across regions based on where compute actually runs. So data gravity is reduced naturally. Lucas: That's smart. And it's a trend I'm seeing more of — region-aware data placement. It's not just about cost; it's also about resilience. If you spread data, you reduce the blast radius of a regional outage. But it does increase complexity. Luna: Right. So the trade-off is between simplicity and avoiding these new fees. For a smaller company, the penalty might not be worth the engineering effort to shard. But for enterprises with petabytes, $2 million is real money. Lucas: Exactly. And the hyperscalers know that. They're betting that the inertia of existing data will keep you locked in — unless you proactively restructure. These clauses are designed to capture value from that inertia. Luna: One thing I'm curious about: do these penalties apply to all customers, or just the big ones? Lucas: Right now they're appearing mostly in Enterprise Agreement renewals — so $1 million plus annual commitments. But I wouldn't be surprised if they trickle down to mid-market within the next year or two. AWS has a history of piloting terms with big customers and then rolling them into standard contracts. Luna: So small and mid-size companies should watch for this too, even if they don't see it yet. Lucas: Absolutely. And the best defense is to measure your own data gravity now. Know which regions have cold data. Start a conversation with your account rep about it — because once it's in the contract, it's much harder to negotiate out. Luna: I wonder if we'll see third-party tools that automatically detect data gravity waste and suggest remediation, kind of like how CloudHealth does for general cost optimization. Lucas: I think so. There's already a startup called Gory — g o r y — that does exactly that. They analyze your cloud storage patterns and flag high-waste regions. They also help you simulate what a data gravity penalty would cost under different providers' pricing models. Luna: Sounds like a useful tool for procurement teams. Lucas: Yeah. And it highlights a broader trend: cloud contracts are getting more granular and more punitive. The age of 'unlimited storage for a flat fee' is over. Every byte has to justify its existence. Luna: So let's say I'm a CFO listening to this. What's the one thing I should do this week? Lucas: Pull your top five storage accounts by cost. Look at the read to write ratio over the last six months. If any account has a read count near zero for more than 90 days, flag it. Then ask your cloud provider for a data gravity assessment. If they don't have one, ask your internal team to run one. That'll give you a baseline before your next renewal. Luna: And if you find a lot of cold data, consider a proactive archiving project. It might cost a bit upfront but could save millions in penalties later. Lucas: Exactly. And it's not just about money — it's also about efficiency. Data gravity waste means you're paying for storage, replication, and networking on data that provides zero value. Cleaning that up is good practice regardless of contract terms. Luna: True. So the new penalty clauses might actually force companies to do something they should have done years ago. Lucas: Right. And that's the upside. The hyperscalers are essentially using pricing to push better data hygiene. It's just that the push is coming with a price tag attached. Luna: So next time you hear 'data gravity' in a contract negotiation, you'll know exactly what it means and what to ask for. Lucas: Yeah. And hopefully you'll have a strategy to avoid the penalty — or at least negotiate it down. That's it for this episode of The Cloud Business Podcast. I'm Lucas. Luna: And I'm Luna. Talk to you next time.