Nvidia’s Vera CPU Shifts from Ambition to Delivery — But the Market Is Watching Closely

Nvidia ships Vera-Rubin NVL72 to Google, Microsoft, Oracle, and CoreWeave, targeting Intel and AMD's server market with a $20B revenue goal and 10x AI efficiency gains.

Nvidia has spent years owning the AI chip conversation. Now it’s forcing a new one. The company has begun shipping its Vera-Rubin NVL72 platform to hyperscalers including Google, Microsoft, Oracle, and CoreWeave, marking the first serious push into a server-processor market long dominated by Intel and AMD. The move transforms what was a speculative narrative into a tangible revenue stream — and the stakes are enormous.

The stock closed at €186.16 on Wednesday, up 2.35% for the session, and traded at €187.36 on July 21 as production partner Wistron inaugurated a new $700 million AI factory in Fort Worth, Texas. Year-to-date, the shares have gained 16.15%, though they remain 8.07% below the 52-week high of €202.50 set on May 14. The 30-day annualized volatility sits at 34.58%, a reminder that this is still a high-beta name.

The CPU Bet That Could Redefine the Business

Nvidia’s Vera chip is not just another processor. It is the first designed specifically for “agentic AI” — systems that independently coordinate complex tasks, pushing traditional x86 architectures to their limits. Management expects roughly $20 billion in revenue from this business line this year alone. If Vera can capture meaningful share from incumbents, Nvidia gains a second growth engine just as the first wave of pure GPU data-center buildouts begins to mature.

Early performance data is encouraging. CoreWeave reported a tenfold improvement in AI compute per megawatt compared to the prior Grace-Blackwell generation — a critical metric given that power availability has become the biggest bottleneck for data-center expansion. Nvidia also claims its Vera CPU executes Python code up to 1.8 times faster than AMD’s competing Epyc Turin processor.

Bull Case: Efficiency Gains and a Capex Supercycle

The analyst consensus price target stands at €264.89, implying 42.3% upside from current levels. Wells Fargo estimates that cloud providers’ capital expenditures could reach $1.1 trillion by 2027 — 25% above current consensus — and Nvidia already holds $119 billion in delivery commitments. With a forward P/E of roughly 23.2, some analysts argue this is the cheapest valuation the stock has offered in seven years.

The efficiency story matters beyond the headline numbers. CoreWeave’s test of the DeepSeek-R1 model on the Vera-Rubin platform showed a tenfold improvement in token output per watt versus Grace Blackwell. That kind of gain could trigger a Jevons paradox: lower operating costs for AI workloads may drive total demand higher, not lower. Simplicity Wealth LLC added to its position in the first quarter, and Zacks Research upgraded the stock to “Strong Buy” on July 22, citing solid fundamentals.

Bear Case: Shadow Debt, Chinese Competition, and Margin Pressure

The bears see a different picture. LGT Capital Partners warns that the current “AI scarcity trade” could reverse once industry-wide production capacity comes online, turning shortage into glut. More troubling is the structural risk of off-balance-sheet commitments: large tech companies have accumulated nearly $1 trillion in purchase obligations outside their balance sheets. If AI investments fail to deliver measurable productivity gains quickly enough, those budgets could be slashed abruptly.

Competition from China remains a wildcard. Developments like Moonshot AI’s Kimi K3 have triggered sector-wide sell-offs in the past, and U.S. export controls prevent Nvidia from shipping its most advanced Vera-Rubin systems to China at all — locking out an entire market.

On the cost side, TSMC plans price increases of up to 10% starting in 2027, which would directly pressure Nvidia’s hardware margins. AMD is not standing still either: it is offering Helios rack systems to customers like Microsoft Azure and has secured a significant AI compute deal with Anthropic. If hyperscalers diversify their supplier base to reduce single-vendor dependency, Nvidia’s pricing power erodes.

Technically, the stock is testing support. A disappointing quarterly report could send it toward the 50-day moving average at €181.41 or even the 200-day average at €165.89.

The August 25 Report: What Matters Most

Nvidia reports its next quarterly results on August 25. Analysts expect revenue of roughly $91.79 billion. The key metrics to watch: any deceleration in the data-center segment, a gross margin dip below the current 75%, and — most critically — whether Vera CPU revenue is materializing at the $20 billion annual run rate management has signaled.

The stock’s path from here hinges on a single question: Can Vera actually take market share from Intel and AMD in the $200 billion server-CPU market? If the answer is yes, the €264.89 target is within reach. If not, the narrative shifts quickly — and the 52-week high of €202.50 may remain out of reach for some time.