Why You Cave First in Price Negotiations
You've probably been there: one client accounts for half your revenue, so you're always the first to give ground in a price negotiation. Losing the contract would shake the whole business, so you lower your terms before the other side even has to ask. This week, Nvidia showed that even a wildly profitable company can find itself in the same spot — and revealed exactly where it spends its money first to climb out of it.
A Week When a Highly Profitable Company Sent Its Money Outward
Nvidia disclosed a $99 billion investment portfolio, including $30 billion in Intel and several billion dollars in CoreWeave. It's the company's third major capital move, following a 2,400% dividend increase in May and a $25 billion bond issuance. In the same stretch, it signed a $3.5 billion deal with robotics AI company Figure around its Vera Rubin platform, landed Anthropic as a new customer, and announced it was walking away from the Chinese market.
On the surface, the numbers look comfortable. Quarterly revenue of $96.2 billion beat expectations, and the CEO guided to 70% growth. But the CFO flagged that gross margin had slipped. The explanation: with OpenAI alone accounting for nearly half of revenue and the top four customers making up 44%, big customers now hold more negotiating leverage, and Nvidia can no longer pass costs on now that Meta and Google are building their own chips. Even that 70% growth guidance rests on three assumptions holding: OpenAI's concentration staying where it is, TSMC's supply chain remaining stable, and China export restrictions staying at current levels.
The Logic of Planting Outward When Margins Are Squeezed
A company that's still highly profitable but seeing its margins squeezed has two options: give more ground to hold on to the big customer, or build a game that works even without that customer. Nvidia chose the latter, and this week's moves trace a single chain.
As the reliance on a single customer began eating into margins, Nvidia put money into the next arena while it still had room to spare. The Intel and CoreWeave investments widen the base on which its infrastructure becomes the standard; the Figure deal lays down a pipeline for roughly 100,000 GPUs in the next market — robotics — before it's needed; and Anthropic is a second customer to absorb revenue currently concentrated in OpenAI alone. Walking away from China is a choice not to tie up capital in a market it can't defend. All of these moves point the same direction: instead of wrestling with today's big customer, Nvidia is using its own money to build an alternative that lets it sit at the negotiating table even without that customer.
Reducing Client Dependency Starts While Business Is Good
For a solo operator, the same principle applies — just at a different scale. Reducing your dependence on one client isn't something you scramble to do after revenue drops; it's something you set aside room for while business is good. When things get hard, you have neither the time nor the money to find a new client, and the other side knows it.
Start by writing down what share of this month's revenue comes from your biggest client. If it's over half, also write down exactly how far you'd be forced to bend if that client changed the terms. Just as Nvidia laid out the three assumptions behind its growth guidance, putting into words the assumptions your own revenue plan rests on lets you negotiate on numbers instead of nerves.
Once you can see that ratio, set aside a fixed share of your surplus, while things are going well, to spend on new clients and partners first. Just as Nvidia put money into Intel, CoreWeave, and Figure, that might mean doing a low-cost pilot project for a second customer who isn't generating revenue yet, or investing time up front in a partner who could expand your work. Compare the money you'd spend discounting to keep your current client against the money you'd spend planting the next one, and ask which leaves you more leverage a year from now.
You also need to watch for the chance that your biggest client might simply absorb your work itself. Just as Meta and Google building their own chips squeezed Nvidia's margins, the longer you've worked with a client, the more reason they have to bring your work in-house. Keeping a piece of the work that's hard for them to do themselves is another way to lower your dependency. And if a client relationship genuinely can't be saved, it's better to cut your losses — the way Nvidia walked away from China — than to keep spending to hang on.
One Thing to Try This Week
Try just one thing this week. Calculate what share of your revenue your top client represents, and if it's over half, set aside a fixed amount from this month's profit and reach out first to one prospect you don't yet work with. Just as Nvidia planted $99 billion outward in a quarter when its margins were under pressure, the room you have while business is good preserves more leverage when you spend it planting the next client — not defending the one you already have.
