Loading page…
A FinOps platform that cuts a company's AWS, GCP, Azure and Alibaba Cloud bills and charges 25% of whatever it saves, so the invoice can only exist after the customer is already ahead
priced at 25% of what the platform saves each customer · about 50 employees at the time of a friends-and-family SAFE of roughly $1M, the only outside money taken · roughly 70–80% of new product and feature bets are said to generate revenue
"We ended up in the model where we take 25% of what we save for the customer." His reason is not generosity: "if we stop giving value to our customers, they are stopping to pay us, and then we need to change something in the product, and we need to change it quickly."
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Andrew Alex grew up in a small city near Kyiv, dropped out of university because he found working at a marketing agency more interesting than the lectures, and built the first half of his career in marketing and product. He ran ads for large brands including MasterCard, moved to Conductor's enterprise SEO platform, then crossed to the consumer side running VPN apps with more than 80 million users. He describes that as getting both halves of the market: B2B deals of $100K and $500K that move slowly, and B2C where you run many tests, read the metrics and decide on the data. In 2020, in the middle of the zero-interest-rate hiring boom, he started a recruitment agency on the logic that when everybody is digging for gold you should sell shovels. It worked, and then it bored him. The first product came out of his own irritation inside that agency: people left the company and the company kept paying for their tools for years, not months. He built a subscription-management tool, tested it with a small group of customers, and deliberately did not raise for it — he wanted to prove to himself he could build something and be responsible for it — financing it from the recruitment business for roughly the first two years while gradually winding that business down. The customers are what produced the actual company. They told him they were overpaying not only for SaaS but for cloud, and Spendbase became a FinOps platform that reduces bills on the major hyperscalers — AWS, GCP, Azure — plus Alibaba Cloud, Huawei and other local providers. The pricing is the whole idea: 25% of what it saves. Four years in, the interviewer puts the company at more than $40 million in revenue, on about $1M of friends-and-family money and no institutional capital at all.
The first version was never sold to strangers. Andrew built the subscription-management tool against a problem he was living inside — an agency still paying for the seats of people who had left years earlier — and tested it with a small group of customers he already knew through the recruitment business.
The quick-reference card — acquisition channels, replication playbook, risks & traps, and our verdict — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself. The exact words are quoted below, with a timestamp that jumps to the second each figure is said.
“And now, you know, you guys have built this company, you know, that is 40 million plus in revenue uh with like, you know, I mean, when you did that safe, I mean, you had 50 employees.”
13:05 · hear it in the videoData credibility: Founder interview on a podcast; the numbers are self-reported or host-reported and unaudited. ⚠️ The $40 million-plus revenue figure is spoken by the interviewer, Alejandro Cremades, not by Andrew Alex — the founder neither states nor disputes it, and it also appears in the video's own title. The same applies to the roughly 50 employees and the friends-and-family SAFE of about a million dollars: both are the host's framing, accepted without correction. The founder's own numbers are the 25% savings share, the roughly 70–80% success rate on new product bets, the 80 million-plus users on the VPN apps he previously ran, and the $100K and $500K B2B check sizes from his earlier career. No dashboard, filing or third-party data is shown at any point. ⚠️ The auto-captions mangle the company name: the host says "Spendbase" in some places and "Spendesk" in others — Spendesk is a different, unrelated company, and the founder's own contact address in the interview is andrew@spendbase.co, so we read every mention as Spendbase. The figure he gives for the size of a single product bet is garbled in the captions as "like 100 to 100k" and we have not treated it as a precise number. Country comes from the founder's own account of growing up near Kyiv and moving to the capital; the company's current headquarters is never stated.
The first dollar did not arrive as a subscription, and that is deliberate. Andrew's stated objection to ordinary SaaS pricing is that a customer can pay for a year while receiving nothing, nobody from customer success ever calls, and the renewal goes through anyway.
Two loops compound here. The first is the one that created the company.
Inside the first four years Spendbase came close enough to the edge that payments to vendors ran late.
Spendbase is a FinOps platform: it connects to a company's cloud accounts and reduces what they spend on the major hyperscalers — GCP, AWS, Azure — as well as Alibaba Cloud, Huawei and other regional providers. The first product in the line was narrower: a subscription-management tool for company software, built because his own agency had spent years paying for tools belonging to employees who had already left. Both products solve the same shape of problem, which is money leaving a company on autopilot with nobody assigned to notice. The third product, described as what they are building now, is an internal-automation layer: they record the day-to-day flows employees actually perform, extract the workflows from those recordings, feed the transcriptions to a model, have planning agents produce automation plans from them and further agents implement the automations. The target he names is the copy-and-paste work — moving emails and passwords into spreadsheets, filling out vendor forms, drafting 150-page contracts — and his estimate is that about 80% of that work should be automated. The thread through all three is unglamorous and durable: find the recurring drain inside a company's operations, measure it, then remove it.
One number: 25% of what the platform saves the customer. Andrew frames it as the answer to a question he asks himself — name one thing everybody is wrong about — and his answer is standard SaaS pricing, where a customer pays for a year, receives no value, hears from nobody, and renews anyway. He puts the alternative in the same family as the AI tools now charging per resolved ticket rather than per API call: get paid for the result. The practical consequences are worth understanding before you copy it. Revenue is uncapped on the upside for a large, wasteful customer and near zero for a tidy one, which quietly selects your ideal customer profile for you. Renewal is not a decision anybody has to make, because the payment is a share of a benefit that already arrived. And the risk sits on the vendor's side of the table, which is why it works as a first sale into a sceptical finance team. Andrew's own summary is that it motivates the company to keep delivering, because the moment the value stops the payments stop with it. The company took one friends-and-family SAFE of roughly a million dollars and no institutional money, so the revenue has always had to fund the operation.
Our read: the interview contains no stack details, no infrastructure costs and no margin figures, so treat everything in this section as our analysis of what a savings-share model has to look like underneath, not as something the founder stated. The cost of goods in this business is not servers, it is the work of finding and realising savings in each account. Ingesting billing exports from AWS, GCP, Azure, Alibaba and Huawei is five different schemas, five different discount and commitment programmes, and five different ways of describing the same idle resource — that is where the engineering budget goes, and it is also the moat, since each additional provider is weeks of work a competitor has to repeat. The savings themselves fall into two buckets with very different economics: the automatable ones (rightsizing, orphaned volumes, storage tiering, commitment purchases) which the platform can find repeatedly at near-zero marginal cost, and the negotiated ones (contract terms, committed-use discounts, vendor renegotiation) which take a human and do not scale the same way. The number that decides whether the model works is how much of the saving is recurring. A one-off cleanup pays you once at 25%; a structural change to how a customer buys capacity pays every month it holds. Andrew's own comment that building is much easier now with cloud tooling and automation points the same direction — the platform's job is to push work out of the second bucket and into the first.
Copy the pricing. A share of a measured saving is the easiest first sale in software, because the buyer's downside is zero and the proof arrives on an invoice they receive from somebody else. Copy the sequencing too: a founder-led audit of a problem you personally had, tested on people who already know you, before any thought of raising. Change the scope. A multi-cloud FinOps platform covering five hyperscalers is a real engineering programme and the enterprise end of this market has well-funded incumbents. The version a small team can start this quarter is one provider and one waste pattern — AWS idle and oversized resources for companies spending $20K to $200K a month is enough surface for a solo operator, and you can deliver the first ten customers as a service with scripts before any of it becomes a product. Change the measurement discipline, because it is where this model fails. Define the baseline in writing before you touch anything, agree how long a saving counts for, and agree what happens when the customer's own engineers do the work. Our read: most of the disputes in savings-share pricing are about attribution, not about the percentage, and a two-page methodology attached to the contract is worth more than a sales deck. One more thing to copy: the 80% rule on new bets. He only launches products he is roughly 80% sure will earn, and reports 70 to 80% of them do. For a team with no second pot of money, a deliberately high bar on which ideas get built is a cash-management tool disguised as product strategy.
The revenue figure is the first risk for a reader, not for the company. The $40 million-plus comes from the interviewer, not from the founder's own mouth, and no dashboard, filing or third-party source is shown. Treat it as a number the founder did not dispute rather than a number he proved. The business risks are structural. A savings-share model is paid out of a budget line the customer is actively trying to shrink, so success eats the base: the better you are, the less there is left to save, and year-two revenue from a well-optimised account can be a fraction of year one. The cloud providers themselves ship free optimisation tooling and keep improving it, and every dollar of waste they help remove is a dollar you cannot charge against. Attribution disputes are the operational hazard — when a customer's own platform team rightsizes a cluster the same month you recommended it, somebody has to decide who earned the 25%. Then there is concentration. Andrew's account of the near-miss is that a couple of failed bets, with no outside capital, would have forced them to wind down parts of the operation. That risk does not disappear at scale; it changes shape. And the new direction — recording employee workflows to generate automations — puts the company inside customers' internal processes and their recordings, which is a privacy and security surface far larger than reading a billing export.
The four founder playbooks — first customers, first dollar, the flywheel, the silent stretch — are in the Unicorn Club, on this idea and on all 1,435 others. The story, the numbers, the sources and every later section stay open to everyone.
Join the Unicorn Club — $5/moCancel anytime, in one click