Nvidia Just Spent $279 Billion to Make Sure Nobody Buys Without It

The score: Nvidia delivered a $96.2 billion quarter, up 106% from a year ago. Data center revenue alone hit $89 billion, up 117% year over year. The company guided the next quarter to $108 billion. The stock fell 1.3% today.

That divergence between the numbers and the price is not a mistake. It is the market telling you what it is afraid of.

The fear is not that demand is slowing. The fear is that the bottleneck inside Nvidia's supply chain has shifted from things it controls — its own GPUs — to things it does not: memory. And memory has a way of eating margins.

Here is what changed in the latest quarter.

Nvidia CFO Colette Kress revealed supply commitments of $279 billion over the next few years. That number more than doubled from $119 billion just three months earlier. The breakdown: $92 billion due in the rest of fiscal 2027, $87 billion in fiscal 2028, and $88 billion in fiscal 2029.

This is not a budget. It is a reservation. Nvidia is pre-paying suppliers — primarily SK hynix and Micron — to lock down high-bandwidth memory (HBM) for its next-generation systems. Without that memory, the Rubin and Rubin Ultra GPU platforms cannot ship.

The commitment is staggering, but it is not optional. Memory is now the constraint on how much AI infrastructure Nvidia can sell. The company has moved from being the seller everyone needs to being the buyer who needs everyone else.

The deeper question is whether this $279 billion bet strengthens or weakens Nvidia's economics.

It strengthens demand visibility. Nearly three years of supply commitments at this scale means Nvidia has order certainty that would make most companies weep. When you pre-pay $279 billion for the tightest component in the stack, you are telling the market that you expect at least that much revenue downstream.

But it weakens margin control. Nvidia's gross margin was 75% in the last two quarters. Guidance for the coming quarter drops to 74%. The company has already told its largest customers that AI server prices will rise more than 15% starting in early 2027, with memory costs as the reason.

Nvidia can raise prices because it has pricing power. But the moment it has to raise prices to maintain margins is the moment the market starts asking whether that pricing power will last. Hyperscalers are not charities. Amazon, Google, Microsoft, and Meta are themselves spending close to $700 billion this year on capital expenditures, and they have every incentive to push back, build custom chips, or negotiate harder.

Here is the tension in plain language: Nvidia is absorbing memory inflation to keep the supply chain running, then passing some of it through to customers who are already stretched thin on their own spending. The margin squeeze is not a one-time event. It is the structural result of being the center of the most capital-intensive buildout in tech history.

More importantly, the market's ambivalence after such strong earnings tells you that the old playbook does not apply.

Three years ago, Nvidia beat estimates and the stock soared. The assumption was simple: demand grows, margins hold, earnings explode, valuation expands. The whole cycle ran on that logic.

Now the cycle has a new dynamic. Revenue is growing faster than ever. Earnings are doubling year over year. But the margin trajectory is bending the other way, the company is sitting on $279 billion in pre-commitments that convert to cash flow only if memory suppliers deliver, and hyperscalers are simultaneously the customers funding this buildout and the competitors designing their own chips.

That does not mean the bull case is broken. It means the margin between the bull case and the base case has narrowed. The numbers need to keep working, not just for Nvidia, but for everyone down the chain.

What does Nvidia's valuation look like through this lens?

The stock trades at a trailing P/E of roughly 27. That is down sharply from a 12-month average around 42 and well below the five-year average near 69. The PEG ratio sits at 0.22, which implies the market is pricing growth at a steep discount to the actual rate.

Some investors call this "insanely cheap." The reasoning: at current earnings, Nvidia trades at 11 times consensus fiscal 2029 estimates, and that consensus may be too conservative given the demand trajectory.

But a low multiple relative to history is not proof of upside when the business has changed. The Nvidia that earned a 69x multiple had gross margins approaching 76% and was the undisputed gatekeeper of AI compute. This Nvidia still dominates, but it is negotiating its way through a memory bottleneck it did not create, raising prices to offset costs it cannot control, and operating in an environment where hyperscalers are simultaneously dependent on it and actively reducing that dependence.

The valuation is cheap relative to the past Nvidia. It may be fair relative to this one.

The supply chain tells a clearer story than the multiple.

Memory manufacturers SK hynix and Micron now have three-year demand floors from Nvidia alone. These companies trade at forward P/E multiples around 6x, which looks absurdly low if Nvidia's commitments materialize. But memory is cyclical, and no amount of commitment removes the risk of overcapacity down the road.

Advanced packaging at TSMC, networking through Broadcom and the Spectrum-X ecosystem, power and cooling infrastructure, land and shell capacity in the Ohio and Texas buildout corridors — all of these layers gain visibility when Nvidia locks in $279 billion. The order question moves upstream.

But the risk boundary moves with it. A $279 billion commitment across three years is a $279 billion counterparty risk. If hyperscaler capex decelerates, if custom silicon captures more inference work, or if Rubin Ultra does not deliver the efficiency gains Nvidia promises, the memory that was locked up today becomes stranded capacity tomorrow.

The checkpoint for investors is straightforward.

Watch next quarter's gross margin. Nvidia guided to 74%, but the 15% price increase has not fully flowed through yet, and memory costs are still climbing. If margins hold at or above 73-74% while revenue hits the $108 billion guide, the narrative stays intact: Nvidia is absorbing pressure and passing it through without losing the business.

If margins fall toward 70% or below, or if the revenue guide is missed, the story changes. A meaningful margin compression alongside slowing revenue growth would mean the bottleneck has shifted from a temporary constraint to a structural problem.

That is the line. Everything else is noise around it.