When Half Your Revenue Comes From Two or Three Clients
If half your revenue comes from two or three clients, then even while business looks healthy, your next big expense is already being decided by them. It's a familiar feeling for any freelancer or small supplier who takes on contract work or delivers to just a handful of accounts. You buy equipment and hire people to match the volume a major client wants, and when you get that money back depends on their next order.
A notable move at Amazon (AMZN) last week followed the same pattern. AWS (Amazon Web Services), the company's cloud division, has started taking on server manufacturing and GPU (graphics processing unit) procurement itself, just to keep pace with the AI (artificial intelligence) demand of a handful of large customers.
Amazon Took On Server Manufacturing to Serve Its Biggest Customers
AWS is spending $1.6 billion to expand server maker Wiwynn's manufacturing facility in Texas. It's a decision to shift the supply chain away from buying finished servers and toward building AI servers in-house. Over the same period, AWS agreed to buy 2 million more Nvidia GPUs. That's on top of the chips Amazon is developing itself, not instead of them.
Who this spending is really for shows up in AWS's customer mix. Forty-four percent of AWS revenue comes from its top four customers, and both the expanded manufacturing capacity and the added GPU purchases are aimed squarely at meeting those four customers' AI infrastructure needs.
The payback numbers came out alongside them. AWS operating income stands at $39.8 billion, or 58% of Amazon's total company profit. Even so, CEO Andy Jassy has disclosed 2028 as the breakeven point for these servers. The company itself has pushed the date when today's server purchases start turning a profit years into the future.
Big Customers Set Both the Spending and the Payback Clock
Read through the business logic and the sequence becomes clear. When its top four customers ask for more AI servers, AWS is in no position to say no — they account for 44% of revenue. Meeting that demand calls for 2 million more GPUs, and putting those GPUs into servers built in-house, rather than bought from a supplier, appears to be the better bet for keeping up with supply. That added spending doesn't turn into profit right away; the breakeven point is set at 2028. Strung together, the sequence looks like this.
Customer concentration is the path that determines both the spending line items and the breakeven timing.
This is a highly profitable business, and yet four customers are the ones deciding where its next dollar of spending goes. As long as those customers keep driving growth, this sequence doesn't look like a problem. It becomes one the moment any one of the four cuts back on orders. The manufacturing capacity expanded and the GPUs bought for that customer stay right where they are, and the breakeven point pegged to 2028 gets pushed further out.
It Starts With Writing Down Your Top-Customer Share as a Number
The scale is different, but the mechanics are the same. For a solo operator, a big client is a welcome source of growth and, at the same time, the party deciding where spending goes. The trouble is knowing that share only by gut feel. If you've never written down whether half your revenue comes from two or three accounts or more, it stays unclear who your next expense is actually for.
Reducing your dependence on a key customer starts with putting a number on your top-customer share. Just as Amazon's books show its top four customers at 44%, start by writing down what percentage of your own revenue your top three clients account for.
The next step is flagging, separately, the spending that's tailored to that one client: equipment you'd have no other use for, staff you added to handle their volume, inventory built to their specs. List each one on its own line, and next to it, write how many months of orders it takes to break even. Just as Amazon disclosed 2028 as its server breakeven date, pinning a date to your own payback lets you see in advance what turns into a burden first if that client's orders shrink.
Finally, for any expense whose payback date lands too far out, look for a way to share that cost with the client. Structures like upfront deposits, minimum order commitments, or equipment leasing shift part of the spending onto the customer's side of the ledger — and that's how you start cutting key-customer dependence without having to diversify your revenue right away.
One Thing to Try This Week
If you do just one thing this week, break down the last twelve months of revenue by client and work out what percentage your top three account for. Write that number next to the spending you added because of those three clients and how much time is left until it pays back, and it all fits on one page that shows exactly who is deciding your next expense. For Amazon, that same math showed up as 44% of revenue and a 2028 breakeven date. For a solo operator, reducing dependence on a key customer starts with building that one page.
