Show HN: Usage 2.0 – Cut AWS Spend by 57% in 5 Minutes
Hey HN: Kaveh here, the founder of https://www.usage.ai/
We launched on Hacker News for the first time early last year, and we've made a lot of progress since then. We've saved tens of millions of dollars for companies, and we are even more excited to announce the launch of a new product: insured reservations for RDS! We worked closely with AWS on this feature and are excited to finally make it generally available.
We help companies drive down AWS EC2 & RDS spend. Why? Because the way it's done now is a pain. DevOps and Software Engineers end up spending time managing costs and reservations rather than focusing on business problems.
In the early days, we saw horror stories of customers with millions of dollars in monthly on-demand spend simply because their finance team didn't want them committing to AWS. Worst yet, we've seen AWS users who ended up overspending by hundreds of thousands of dollars a month because they overcommitted their Savings Plan commitment.
Here's how it works: We are typically brought in by a DevOps manager to cut AWS EC2 costs. The app is entirely self-service and the savings are generated automatically, typically we do this live on a call. On average, we reduce AWS EC2 spend by 50% for 5 minutes of work, and RDS spend by ~30%.
To reduce by 50%+, we don't touch the instances, require any code change, or change the performance of your instances. We buy Reserved Instances on your behalf (a billing layer change only) and bundle them with guaranteed buyback. So you get the steep 57% savings of 3-year no-upfront RIs with none of the commitment.
We make money off of a 20% Savings Fee. Happy to chat directly kaveh@usage.ai
Have you experienced any issues with managing your company or organization's AWS expenses? We'd love to hear your feedback and ideas! > We buy Reserved Instances on your behalf (a billing layer change only) and bundle them with guaranteed buyback This is by setting your AWS account as the payer account so you can aggregate? I’m curious how the finances will work out - at least one of them had challenges balancing their RIs with customer changes since there wasn’t a feedback cue for developers not to change instance types casually. On a marketing level, I have an instinctive negative reaction to the claims like the title here has because I know it’s not true for me (my accounts are dominated by storage and network egress so even if EC2 were free you couldn’t get 57%). I’m wondering how best to phrase it to help people understand they can save a lot but not make it seem like you’re misrepresenting the possibilities. It’d also be useful to compare with compute savings plans. This is mainly in reference to the fact that Reserved Instances don't have any bearing on the instances themselves (ie. no code change, performance chance, server downtime, etc.). The 57% savings is the difference between the 3-year, no-upfront, standard Reserved Instance rate and the On-demand rate (for RDS it is 30% vs on-demand) As far as compute savings plans: 1-yr SP is anywhere from 26-29% savings vs on-demand 3-year Sp is anywhere from 49-52% savings vs on-demand ... but note that these commitments are non-transferrable. Customers find our tailorable commitments to be a healthy blend of savings + safety against over-committing to volume they may not need If you take 20% of the savings, my max savings are ~45%. Even that assumes my AWS bill is entirely instances. I think you have a product that some companies will really want. It’s a good product; don’t let your marketing promises exceed the true savings by so much that it makes people leery of what else you might be hiding or stating in a less than straightforward manner. I'm a huge user of RIs, I like the idea of tools to balance them more easily, etc. Totally on board with that but the way the number is presented seems to me like you're setting yourself up to over-promise and under-deliver, not to mention turning off people who are familiar enough with this to already be using RIs / CSPs somewhat. Simply qualifying it to “Cut AWS EC2 spend by up to 57%” would avoid some of that, and you could probably address the latter users with some hard data about flexibility or total net savings by other customers so people could get an idea of what it looks like for normal users who aren't just buying acres of EC2 instances and nothing else. It just depends on what you have. I did get 20% right off the bat with RDS just buying no-upfront RIs. I could have probably gotten around 30% if I did upfront and longer commitment. Same here.
If you shaved of 100% of all the EC2 instance costs off of our 40K/month AWS bill we would save -> nothing. We are not running a single EC2 instance. This is like insurance in a couple of different ways, and makes me happy, but concerned. usage.ai is basically taking a bet that a large number of their customers wont cut back on RDS spend simultaneously. AWS massively made RDS instances cheaper when they released RDS on Graviton. I can only imagine that'd nearly put a company like usage.ai out of business. I'm curious what defense mechanism they have against this. > AWS massively made RDS instances cheaper when they released RDS on Graviton. I can only imagine that'd nearly put a company like usage.ai out of business. Are the on demand and reserved prices the same? If RI’s exist are still cheaper, it shouldn’t impact their business model. They made the new instance type much cheaper. Users already committed to RI cannot change instance types and remain bonded to the contract. Usage.ai and others give you an insurance that you can get out of the contract. If that existed before, users would return the old instances to usage.ai and get the cheaper ones. Usage.ai end up with a lot of reserved instances no one wants, with a large contract to pay. I guess it depends if they’re buying convertible or non-convertible RI’s behind the scenes. (Unless RDS doesn’t have the concept of convertibles) We've just taken over a platform for a client whose previous partner almost drove them bankrupt with insane AWS charges, support costs and so on. Most of their initial funding got burnt on charges that didn't relate to developing or marketing a product. AWS is just a horrible product for smaller companies. We've heard similar horror stories -- we're hoping to make the cloud a little more manageable! How does this differ with Vantage Autopilot https://www.vantage.sh/features/autopilot ? just from reading the site, looks like vantage takes a 5% fee vs Usage 20% - but that's a big difference, not sure if it's an apples-to-apples comparison. > Autopilot charges only 5% of the savings found to maximize customer savings: 75% less than other providers in market. Sounds accurate. CEO of Vantage here: You're correct - and we only charge for EC2 savings realized. Additionally we give away all other service recommendations for free: RDS, Elasticache, Redshift, etc. These are your savings and you should realize as much of them as possible. We also have a suite of visibility and reporting tools for AWS, Azure, Datadog, Snowflake, MongoDB, Databricks, etc. Most customers prefer using Vantage as a one-stop-shop FinOps vendor and in general we find significantly more in savings at a fraction of the cost of other vendors. Jeff bezos hates this one weird trick Jeff bezos gets to convert a set of small, flaky customers into guaranteed 3-year contracts while pushing all the risk to a third party. I'm gonna say he isn't going to be too mad about this. What is the minimum number of servers I need to buy from you? Can I buy just one 4GB server, for example? You can start with as little as one 4GB instance! Looks really good. I have been offered large discounts by an AWS reseller, and I have not understood exactly how or why they can do this. Are the resellers doing what you are doing but being less transparent about it? In a way. Resellers typically incorporate a company's AWS organization into theirs, and via resource sharing of savings instruments, are able to pass on savings to their customers within their portfolio. Since they have an AWS organization that is comprised of many companies and their AWS accounts, they are able to negotiate special pricing arrangements with AWS (typically in the form of an EDP) based on the total spend of that consolidated organization. We use similar instruments but allow you to maintain the independence of your AWS organization for what is usually higher savings. Why would they be in a good negotiating position with AWS, given that they have to be on AWS to even exist in the first place? I don't get why amazon would ever negotiate down with these guys, they're the ones who absolutely need the product. Like any reseller, they are aggregating demand, and then negotiating volume discount with the provider. Resellers will work with multiple cloud providers. In principle at least, customers can become loyal to the reseller and their value-added services, and the reseller can influence their clients to move to other providers. Obviously hosting is one of the stickiest SaaS services around since switching costs can be monumental, but the theory still applies. AWS SPP Resellers actually resell for better prices, see Partner Originated Discounts. https://aws.amazon.com/blogs/apn/introducing-new-aws-solutio... FYI, if you are struggling to find the link in the post like me:
https://usage.ai I'll reserve your instances for only a 10% savings fee. Or we can do a 20% fee and I'll kick you back 10% cash. Hit me up y'all. Interesting, the financialization of reserved instances. So, effectively, this ends up being similar to you buying reserved instances then renting them out to customers, taking advantage of the market inefficiency inherent in the pricing difference between reserved and non-reserved instances? I'm on a recommendation, it shows reservation 48.28% and $24.53 Lowest Risk then has a line of text that says "Select a recommendation to compare it to actual resource utilitzation" three things, what does select a recommendation to compare it to mean? I'm not sure what I should be selecting. And second, is it utilitzation or utilization? Or is that word like canceled and cancelled ? and the third thing, when approving, I see total savings of $24.53 and by selecting approve, you'll be charged with 20% of realized savings monthly. Should this give me some idea of how much this will cost ? I would think its $24.53/12 = $2.44 * .2 = $0.488 Is that correct? > We buy Reserved Instances on your behalf (a billing layer change only) > and bundle them with guaranteed buyback. > .. We make money off of a 20% Savings Fee. so this is similar business model, like https://archera.ai/guaranteed-reserved-instances/ "Find big-time savings with short-time commitments.
Unlike typical commitments, Guaranteed Reserved Instances can be sold back to Archera within a month of initial purchase and with higher savings than standard offerings." Archera appears to only charge for GRIs. They don't seem to tell you what they charge though. How does the buyback program work for your new releases (RDS, Elasticache, OpenSearch, DynamoDB)? I use the EC2 RI Marketplace but I can't find one for the other products mentioned. Also, I just reviewed your T&C and your "Program" states that you provide "Non-Usage Credits". What does this mean? I don't have any guarantee that in an off-scenario where there is an overload of RIs being sold I'll stop paying for them. I'd rather use Vantage and pay a 5% fee on my RIs. How does this compare to what's offered by doit? https://www.doit.com/flexsave/ From my conversations with doit sounds like they offer a similar service in that doit acts as an AWS reseller (~~some billing magic) and they apply commitment discounts etc. and pass along a portion of the savings. doit also claim to throw in free cloud architecture and support consulting services as a value add. > Using machine learning, Flexsave continuously monitors your cloud usage to identify compute instances that are not covered by existing commitments Huh. This sounds very cool! How does it work technically? I have checked your website but I am still not sure. How do I "buy" the instances using usage.ai instead of AWS? Do you need additional tooling or does it work with awscli/terraform? How do you integrate with my account and how do "your instances" appear in my account and how can I sell it back? As far as I understand, I can buy/sell reserved instances from you in a flexible way, so they can be used like on-demand instances. Anyone have insight into why AWS hasn’t offered 6 series instance types for Oracle RDS with license included? The conventional wisdom that a new instance type will improve dollar per unit if performance year over year is breaking down in this case. I could guess that Oracle licensing changes or negotiation has made it untenable for AWS. For my own curiosity, what's the benefit of a one-time engagement with Usage AI over a continuing engagement with a tool (plus proserv, if merited) like CloudHealth? Much of the trouble with AWS spend is due to deltas as the platform changes, rather than a single realization that AWS costs too much. (Disclaimer: I was an engineer at CloudHealth four years ago.) Former AWS employee here, and just generally been in the infra/cloud space for a long time. I've seen a few of these products (or this feature within existing products) over the years. The things that always gives me an extreme amount of pause, and why I'd never personally be willing to pursue this as a startup myself, are: * Amazon in general is famous for using scale to drive down costs and/or attack the margin of competitors. So much so that a Bezos quote of "your margin is my opportunity" regularly gets thrown around. This is a direct attack on what is essentially fat AWS margin. Which isn't itself bad. Maybe they've grown complacent and lazy? It does worry me though when the strategy is essentially taking on the reigning champion by using their own game against them, where they both have the home ground advantage and get to set the rules. You're one feature/packaging/pricing change away from becoming irrelevant. * The huge information asymmetry. AWS can see _all_ the data. They have entire teams dedicated to building models around it. They know who should be on RIs, for what term, the impact to both them and customers. For even the most moderate sized customers and above there are people who have a KPI on driving RI usage + cost reductions into those accounts. How much opportunity is there really? Prospects who have the most to gain from this should theoretically have SAs/TAMs/etc. telling them every month to take these actions. They've either done the low-hanging fruit already or aren't interested because of other competing priorities. I'm not sure another third-party tool changes those things. * The black swan risk of the insurance-like underwriting of carrying RIs. The models for all of these when I've looked into them meant there was tail risk around a large number of customer simultaneously deciding to hand back their RIs. And there's a lot of reasonable justification to believe that's unlikely. It also doesn't feel like a massive stretch of the imagination to think that a change in macro conditions + a shift to a far more aggressive red ocean strategy + a new instance family + AWS' history of discounting new instances on a $/performance basis to encourage adoption/migration = a huge price drop to drive both competitive and architecture migration en masse. The value of the RIs for what's now considered "legacy compute" would plummet overnight. All that skepticism aside, the things I continually find most interesting in this space are the general usability and UX improvements. AWS consistently has a pretty suboptimal user experience, and the billing aspects in-particular can be indecipherable at even a trivial level of scale. So maybe the opportunity here is less about attacking margin and more about the fact customers are leaving money on the table purely because AWS make it so dang hard understand and improve things? How would you describe the mechanics of this in regular finance terms... Almost like you are providing mortgages for reserved instances? I think I understand the model here, but I guess I'm curious how you model in risk of suddenly getting stuck with a bunch of 3 year RI commitments? I like to say we're a marketmaker for cloud contracts. Our recommendations take our current inventory into account and it's been very effective considering the amount of 'trades' we successfully make each day. That aside, We also have a considerable amount of cash set aside for an event like this. Is the inventory risk assumed by usage? Reading through your T&C it looks like you refund us with credits on your platform - sounds like monopoly money to me? I recently found out that I was overpaying by about 20% by just looking at "Recommendations" tab in AWS Console and buying RIs. It was mostly RDS. I would love a product that automatically manages the RI renewals (ex. every year) and instance type changes. Good to hear. :-) Curious what happens to resources and reservations if Usage goes out of business? Three years is actually a pretty long horizon to rely on a startup still being around. It's an asset on their books, but I doubt any of their investors/creditors can get any real use out of them. So my guess is AWS gets to keep that extra $$$. I think what they mean is, does the Usage.ai customer get stuck with long RI's that they now can't get rid of because usage.ai can't buy it back Yes I was thinking about this from a customer's perspective. In particular I can't tell if Usage is buying/selling/holding reservations in their account but on their customer's behalf, or if they are just automating the buying and selling of reservations but that's all done in the customer's account with the customer's money. If the latter then if Usage disappears one day I assume you might be stuck with reservations you don't want, or without the ones you do want, but in the former scenario the situation could be more chaotic if suddenly resources you depend on simply stopped being paid for. You can sell RIs on the RI marketplace. I guess you could still sell them on the RI marketplace. Any plans to launch this product on Google Cloud? For now we're all-in on AWS, but have plans on going to Google Cloud later this year! Is there a similar service out there for Azure? (Who just put their uk prices up by 9 percent) Why is AWS okay with this-- Isn't this profit that should be in their pocket? Yes it should be and will be in a due time. AMZN carefully watches the marketplace for "wonders" who are monetizing inefficiencies and if these inefficiencies worth big enough profit, AMZN will fill the gap themselves. Platform is never your friend or "partner" even if they say so. Why would AWS complain? A company coming along and converting some of their month-to-month instances into 3-year billing commitments while taking on all the risk is fantastic from their perspective. No wonder they are encouraging it. We actually have a strong positive relationship with AWS! They're helping us by introducing us to their customers, and we can also be purchased on the AWS marketplace: https://aws.amazon.com/marketplace/pp/prodview-3sq4hhmmwb5fg... I was wondering this .. could the fact they have a larger guaranteed long term spend, mean they can leverage that future income in some way? Can I do this twice to get a compound effect? Wow cool, hope you guys are successful. Appreciate your support! How is it different than nops.io? cut AWS stock prices by 57% in 5 minutes! Amazon to increase AWS prices by 57%