Cloud Costs
Archera CEO Aran Khanna joined Scott Hanselman on Hanselminutes to talk cloud spend, insured commitments, and why on-demand pricing is basically the convenience store version of compute.
I kept thinking “we have heard this cost visibility, cloud tagging and attribution story one too many times.” For me, the game changing moment was when Aran began talking about reducing risk, proactive planning, and creating a secondary marketplace.
TLDR:
Scott Hanselman has been paying as-you-go in the cloud for 17 years. He's sharp, he's built things, and he's never felt comfortable committing to a year of compute he might not use. Which makes him the perfect person to explain why cloud commitments are broken, and why Archera exists.
The conversation landed on Hanselminutes last week. It's a good one. Here's the rundown.
When cloud computing took off, the pitch was simple: spin it up, spin it down, pay for what you use. No data center, no CapEx, no waste.
That promise still holds…kind of. But somewhere along the way, reserved instances, savings plans, private pricing agreements, and committed use discounts got piled on top of each other until nobody could make sense of it anymore.
Aran was on the AWS team that launched SageMaker back in 2017. He watched it happen in real time: every new capacity problem got "solved" by throwing a new commitment type at it. Eventually customers were staring at a spaghetti mess of overlapping instruments and still leaving money on the table.
The discounts are real, by the way. If you move from on-demand to a three-year commitment, you can save north of 80%. That's not rounding error, that's headcount. But the risk of committing to three years of infrastructure when your business might look completely different in 18 months? Also real.
Aran used an analogy in the episode that's worth repeating.
Building a data center is like building an apartment building. It costs a lot upfront, and you've probably got a loan against it. If you put all the units on Airbnb, you might get a good nightly rate, but you'll have vacancies, and your lender isn't going to love the unpredictability. What you really want is one-year and three-year leases. Guaranteed income, every month.
Cloud providers are the same. They need committed revenue to justify the capital they're putting into new data centers. That's why they offer deep discounts for commitments. They're not being generous, they're managing their own balance sheet.
Archera's product is basically the right to break that lease. You get a three-year rate, but if your circumstances change after year one, you're not stuck. We pay you back for the unused portion.
Scott's follow-up was great: "Do the cloud people know you exist? Because if I break my lease, the landlord would be upset."
The answer is yes, and they're actually fans. Most of Archera's revenue runs through AWS, Azure, and Google Cloud marketplaces. The hyperscalers like it because it helps their own sales teams close bigger deals. Customers who were afraid to commit suddenly aren't.
This was Scott's line, and it's the one that's going to stick.
He bought 900 Advil at Costco for $25. Then bought two Advil at a hotel for $5. Same product, wildly different price, because one requires showing up and one requires committing to bulk.
On-demand cloud is the hotel Advil. You're paying for the convenience of not committing. That's fine when you're prototyping. It stops being fine when you're running a real business at scale and the premium is compounding every month.
The math gets serious fast. At $100K–$200K a year in cloud spend, the savings from commitments can fund a new initiative. At $500K+, you're talking about actual headcount.
There's a newer wrinkle that came up in the conversation: in 2026, you often have to commit just to get a GPU.
The latest Nvidia hardware requires liquid cooling and high power density. Supply is tight. So cloud providers are using commitments as an allocation mechanism. If you want a Grace Blackwell, you might need to sign a commitment just to get in the queue.
For startups trying to plan their AI infrastructure, this creates a real dilemma: how much do you overcommit now to guarantee capacity later, knowing you might not use all of it? That's exactly the kind of risk Archera can take on with you.
Aran was clear that Archera works with everyone from two-person YC startups to some of the largest cloud customers in the world. The sweet spot where it starts getting meaningfully impactful is around $100K–$200K a year in cloud spend. But they've also helped individual developers running side projects save $75 a month on a $200/month bill; which, if it's coming out of your personal credit card, is real money.
There's also a free tier. Archera will automate your commitment lifecycle management (tracking, renewing, optimizing) at no cost. The insurance is an add-on you opt into when it makes sense.
The full episode is on Hanselminutes. Scott asks the questions a smart developer would actually ask, which means the analogies are good and nothing gets too deep into the weeds.
And if you want to see what Archera could do for your cloud bill, visit archera.ai
