The debate landing on investment committee agendas this week is not whether Nvidia can raise prices. It is who ultimately pays for the increase, and what that cost does to the financial architecture of the AI buildout.
Contract server builders have told some of Nvidia’s largest customers that Vera Rubin and Grace Blackwell system prices will rise more than 15% in many cases, with increases tied to systems shipped in 2027. Rising prices and shortages for DRAM and high-bandwidth memory are increasing the cost of systems built around Nvidia’s newest accelerators. Companies that build servers under contract for large data center operators, including Microsoft, Google, and Oracle, have notified customers of the forthcoming increases.
Why Wall Street Cares
A 15% rise on rack-scale systems that sell for several million dollars each adds hundreds of thousands of dollars per rack across deployments that run to thousands of racks. Multiply that across the committed spending plans of the hyperscalers and the number is not trivial. The four hyperscalers are widely expected to spend about $725 billion in combined 2026 capex, up 77% from about $410 billion in 2025. A meaningful share of that is Nvidia-stack equipment hitting the books in 2027 at the higher price.
Given that AI assets typically depreciate over a short life, the hyperscalers were already set up for a sharp rise in annual depreciation expense, and these capex plans still do not capture the full extent of the AI build-out. Layer a 15% unit cost increase on top, and the depreciation math gets worse before it gets better.
The Bull Case
For Nvidia, passing the cost through is a show of pricing power, particularly given its very large share of the data-center AI accelerator market and gross margins around the mid-70s. Gross-profit dollars are protected. Revenue rises with the price. The unit economics of Nvidia’s dominant position look unchanged.
For the hyperscalers, there is a parallel argument. Cloud revenue accelerated in the same quarter capital expenditure peaked: AWS grew 37%, which Amazon described as its fastest growth in 18 quarters; Azure grew 43%; Google Cloud expanded 82%. Companies growing cloud revenue that fast can absorb higher depreciation if the revenue curve stays steep enough.
The Bear Case
Here is the tension that professional investors are actually arguing about: gross-margin percentage versus gross-margin dollars are not the same thing, and the market tends to price the percentage. Even if Nvidia passes through every incremental high-bandwidth memory dollar, gross-margin percentage can still drift lower if the pass-through carries little to no incremental gross profit. Gross-profit dollars are protected, but the percentage moves lower.
For the buyers, the choice is more binary. They can absorb the increase in depreciation, accept margin compression as amortized costs climb, or slow the pace of deployment and preserve cash. As recognized depreciation compounds, hyperscaler margins compress mechanically, and CFOs facing margin questions historically respond by moderating capex growth. Some investors also worry about individual balance sheets as capex rises relative to operating cash flow, but the crossover points vary by company and quarter.
The Evidence
TrendForce, citing GF Securities analysis, has put memory at about 29% of a Vera Rubin VR200 system’s roughly $2.1 million bill of materials. That share is climbing as memory demand grows faster than production capacity, a gap that points to continued tightness into 2027. Gartner has also warned that AI-driven memory shortages could persist into the second half of 2027.
Nvidia fell 2.9% on Monday, marking its seventh consecutive session of losses, the longest such streak since September 2022. Micron Technology fell 5.83% in the same session, while AMD dropped 3.49%, Intel fell 3.1%, and Broadcom declined 2.63%. The chip complex is pricing a risk, not just reacting to a headline.
What Investors Are Missing
The overlooked implication is what this does to the slower-moving hyperscalers. Oracle and Microsoft are the names most exposed to a deployment slowdown. Oracle fell 2.74% Monday. A 15% jump in server costs, arriving as capex intensity stays high, is not a rounding error for Oracle’s financing capacity.
Meanwhile, AMD stands as the quiet beneficiary if any hyperscaler tests alternative procurement. AMD said its Data Center segment expanded 107% year-over-year to $6.7 billion in Q2 2026. Any cost-driven motivation to diversify away from Nvidia improves AMD’s positioning in contracts currently being negotiated.
Stocks to Watch
NVDA: Gross-profit dollars are protected by the pass-through, but watch gross-margin percentage on Wednesday’s earnings release. The market prices the rate, not the level.
MU: SK hynix, Samsung, and Micron account for essentially all high-bandwidth memory supply at scale, and TrendForce has said server DRAM contract prices are expected to keep rising from the second half of 2026 through the second half of 2027. Micron is the most accessible U.S.-listed expression of that leverage.
ORCL: Most exposed to a deployment squeeze. A higher unit cost on the same build plan either slows rack additions or accelerates debt issuance.
MSFT, GOOGL: Both are large enough to negotiate long-term agreements that could limit exposure. The question is whether existing contracts are grandfathered or subject to the new pricing on 2027 shipments.
AMD: The structural alternative. If one large cloud customer tests non-Nvidia configurations at the margin, AMD’s data center trajectory becomes materially more interesting than current estimates reflect.
