The cloud was sold as elastic. Zesty finally makes it act like it.
The company logo, photographed mid-pitch. The blobs are decorative; the savings are not.
Somewhere right now, a Kubernetes cluster is sitting two-thirds empty. The nodes are provisioned for a traffic spike that came last Tuesday and may never come again. The meter keeps running. Zesty's job is to notice - and to do something about it before the invoice does.
Zesty makes automated optimization software for teams that run on the cloud and have stopped pretending they can tune it by hand. Its flagship product, Kompass, plugs into a Kubernetes cluster through a single Helm chart, reads the workload, and then quietly resizes compute, storage, and cloud commitments to match what is really happening. No code changes. No begging engineers to fill out a capacity spreadsheet. The pitch is almost suspiciously simple: pay for what you use, not what you guessed you might need.
The cloud promised infinite, on-demand resources. What it delivered, for most companies, was a permanent surcharge for fear. Nobody wants to be the engineer whose under-provisioned cluster fell over during the Black Friday rush, so everyone over-provisions. The result: data centers full of servers running at a fraction of capacity, humming along, costing money, and - though nobody likes to mention it - burning power for nothing.
Kubernetes made this worse before it made it better. It gave teams extraordinary control over how workloads scale, and then handed them roughly forty knobs to turn, most of which interact in ways that only reveal themselves at 3 a.m. Horizontal scaling fights vertical scaling. Pods fragment across nodes. Persistent volumes fill up at the worst possible moment. The honest answer most teams reached was to throw hardware at the problem and look away.
Zesty's founders looked at all that idle capacity and saw two problems wearing the same coat: a financial one and an environmental one. Wasted compute is wasted money. It is also wasted electricity. Solve one and you solve the other - if, and only if, you can do it without making the application slower or breaking an SLA. That "if" is the entire ballgame.
Maxim Melamedov and Alexey Baikov started Zesty in 2019 in Tel Aviv with a contrarian read on the market. Everyone else was building dashboards - prettier ways to show you how much money you were wasting. Melamedov and Baikov, the CEO and CTO respectively, decided dashboards were the booby prize. Knowing you waste money is not the same as wasting less of it. They bet on automation that acts, not analytics that report.
The wager had a sharp edge to it. Automation that resizes live infrastructure is automation that can take down production if it guesses wrong. The skeptical engineer's first question - reasonably - is "why would I let software touch my cluster?" Zesty's answer was guardrails: you set the rules, the platform optimizes within them, and it earns trust one uneventful scaling event at a time. Investors liked the thesis. The company has raised roughly $129 million, including a $35M Series A in 2021 and a $75M Series B in 2022 led by B Capital and Sapphire Ventures.
In late 2024 Zesty pulled its work together into Kompass, an automated Kubernetes optimization platform. The headline claim is up to 70% off compute and storage with no SLA compromise. The interesting part is how it gets there - by attacking several dimensions of waste at once instead of picking a favorite.
Optimizes horizontal and vertical scaling at the same time, and plays nicely with HPA, KEDA, and any node autoscaler.
Cuts application startup time by up to 5x during scaling events, so right-sizing never costs you responsiveness.
Grows and shrinks persistent volumes based on real usage, ending the 3 a.m. "disk full" page.
Continuously aligns reserved and spot commitments with actual consumption across clouds.
That November also brought a tidy bit of corporate shopping. Zesty acquired Qubex, an Israeli startup that had built a way to scale Kubernetes roughly 10x faster and stay resilient through traffic spikes. The technology became HiberScale, which hibernates nodes at scale and wakes them in about 30 seconds - down from a typical five minutes. The pattern is consistent: buy back the time that overprovisioning used to insure against.
Maxim Melamedov and Alexey Baikov set out to automate cloud waste away, not just chart it.
Led by Next47, with Sapphire Ventures and Samsung Next, to scale AI-driven cloud automation.
B Capital and Sapphire Ventures back the move toward "dynamic cloud infrastructure."
Spike-resistant scaling folds into a single platform promising up to 70% savings.
Complete VPA and HPA optimization closes the loop on per-pod resource tuning.
Skepticism is the correct default for any vendor claiming 70% savings. So here are the customer-reported figures, which are less round and more believable than the marketing ceiling. Printify trimmed 40% off compute. Sennder shrank its cluster footprint by 43%. Wildflower Health went looking for a 30% reduction and landed closer to 65% - the kind of variance that makes a CFO nervous in the good direction.
"I was looking for a 30% reduction. It's going to be more like 65%."
Wildflower Health"Zesty keeps us flexible and we have saved 40% on compute costs."
Printify"Our cloud ROI has greatly improved since using Zesty."
Yonatan Deshel, CTO, WalkmeBehind the testimonials sits scale: roughly $5 billion in AWS spend under management across about 3,000 accounts, a 4.9/5 satisfaction rating, SOC 2 certification, and AWS Advanced Technology Partner status. Customers run the gamut - Zoopla, Armis, Alation, Priority - which is another way of saying overprovisioning does not discriminate by industry.
Zesty lists five tenets - Simplicity, Innovation, Impact, Sustainability, and, refreshingly, Fun. The one that does the most quiet work is Sustainability. Idle servers do not just waste money; they waste electricity, and electricity has a carbon price even when the accounting department forgets to mention it. Zesty's argument is that the financial case and the environmental case are the same case, viewed from two angles.
It is a convenient alignment, and the company knows it. Telling a buyer that cutting their bill also cuts their footprint is the rare pitch where the self-interested and the virtuous answer arrive at the same checkout. Whether you find that cynical or elegant probably says more about you than about Zesty.
Every company now wants to run more compute, not less - GPUs, inference, training, the whole expensive parade. Which means the gap between what is provisioned and what is actually used is about to get wider, not narrower. The work of closing that gap, automatically and without breaking anything, is not a finished problem. It is a growing one.
So picture that half-empty cluster from the start of this story. Before Zesty, it sat there, idling, charging by the hour for capacity nobody used, a small monument to fear. With Kompass watching, the same cluster shrinks when the traffic leaves and wakes a hibernated node in thirty seconds when the traffic comes back. The meter still runs. It just runs honest now.