Nvidia's Record Quarter Exposes the Tension Between Unmatched Demand and Rising Costs

Nvidia's record Q2 revenue of $96.22B misses market expectations as investors focus on margin squeeze and customer profitability concerns.

The numbers were, by any measure, extraordinary. Revenue of $96.22 billion for the second quarter of fiscal 2027, up 105.9 percent year over year and comfortably ahead of the $92.18 billion analysts had penciled in. Data-center sales surged 117 percent to $89 billion. It marked the thirteenth consecutive record quarter for the chipmaker.

And yet the shares finished the week in Germany at €187.78, down 4.1 percent on the day. The stock now sits 7.3 percent below its 52-week high of €202.50, even as it remains 3.6 percent above its 50-day moving average and 11 percent above the 200-day line. That gap between operational brilliance and market reaction tells its own story: investors are no longer prepared to reward headline numbers without interrogating what sits beneath them.

A Diversifying Customer Base Changes the Risk Calculus

What impressed most observers wasn't the aggregate growth but its composition. Hyperscaler revenue climbed 102 percent to $48.7 billion — strong, certainly, but almost expected at this point. The more striking figure came from the ACIE segment, which expanded 138 percent to $40.3 billion, outpacing the traditional cloud business. Nvidia is no longer tethered to the spending whims of Microsoft, Google and Amazon alone; the demand base is broadening in ways that undercut the concentration risk bears have cited for years.

The competitive moat remains formidable. In data-center sales, Nvidia booked more than thirteen times the revenue of AMD, its nearest rival. Yet AMD's shares have outperformed Nvidia's since 2023 — a reminder that operational dominance and stock-market performance do not always move in lockstep.

The Memory-Cost Squeeze

The counterweight to all this momentum is a cost structure under visible pressure. Nvidia has raised prices on its Vera Rubin and Grace Blackwell AI servers for deliveries from 2027 by 15 to 17 percent, a direct response to soaring memory costs. Server DRAM prices rose as much as 18 percent quarter over quarter in the third quarter, following an 80 to 90 percent jump in the first quarter. Memory now accounts for roughly a quarter of server costs, while the GPU share has fallen from around 80 percent to about 50 percent.

Chief financial officer Colette Kress has flagged that gross margin will slip to 71 to 72 percent over the next two quarters, down from 75 percent. New fab capacity isn't expected to come online until 2029 or 2030, meaning the squeeze has legs. The risk is that customers — Microsoft, Google, Oracle, Amazon, Meta — begin to throttle their investment pace if the total cost of AI infrastructure climbs too quickly.

The $279 Billion Question

On the demand side, the signals remain emphatically bullish. Supply commitments jumped from $119 billion to $279 billion within three months, a clear indication that orders are not merely holding up but accelerating. Amazon has reportedly ordered an additional two million GPUs for 2027 and 2028. A financing platform involving Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR is targeting more than $500 billion for AI infrastructure, with Nvidia-financed AI labs expected to contribute roughly a quarter of fiscal 2027 revenue.

But this is where the analysis gets more complicated. Mirae Asset Securities notes that the focus is shifting from raw GPU demand toward metrics like customer AI return on investment, margin stability and free cash flow. If clients such as Meta see improved ad clicks from AI models but can't translate massive data-center spending into commensurate returns, orders could eventually be trimmed despite full order books. Morgan Stanley estimates Nvidia's credit exposure could approach $200 billion by the end of 2028, with $170 billion off-balance-sheet — a structure that fuels growth in good times but would become a burden in a downturn.

Ambition at a Price

Nvidia's expansion into software and ecosystem plays adds another layer. The planned acquisition of Hugging Face for roughly $12.9 billion — about 86 times its estimated annual revenue of $150 million — looks ambitious, even aggressive, on valuation. It signals Jensen Huang's intent to control the entire AI stack, from data platforms to data centers, rather than just selling chips.

There are also competitive and political headwinds to weigh. Google, Amazon, Microsoft, Meta and OpenAI are all developing their own AI chips, with OpenAI's "Jalapeno" reportedly outperforming Nvidia in some tests. The US government is examining expanded semiconductor tariffs that could remove existing exemptions for data centers — a measure under review, not yet approved.

The Path Forward

For the third quarter, Nvidia has guided to $108 billion in revenue, and management has flagged 70 percent growth for fiscal 2028 — well above the roughly 45 percent the market had been expecting. If the company hits those marks, the current valuation, with a price-to-earnings ratio below its five-year average, starts to look reasonable. The stock remains 34 percent above its 52-week low and above all major moving averages.

The volatility — annualized at 45 percent over 30 days — is the market's way of pricing in genuine uncertainty. The bull case rests on supply commitments that keep rising and hyperscalers that keep ordering. The bear case rests on margin compression, customer profitability that has yet to be proven, and a credit structure that could amplify any downturn. The next test arrives with the third-quarter revenue figure: confirm the $108 billion and confidence in the growth story returns; miss it or cut the margin outlook, and the entire AI sector's valuation comes under scrutiny.