There is a curious disconnect at the heart of Nvidia’s 2026 story. The company that powered the first wave of artificial intelligence is now delivering record financials while its stock struggles to regain momentum. First-quarter revenue hit $81.6 billion, up 85 percent year-on-year, and the full fiscal year closed at $215.9 billion. Yet the share price sits 13 percent below the May peak and has shed roughly nine percent over the past 30 days. Even the record quarter failed to excite — the stock fell after the report. CEO Jensen Huang has publicly acknowledged he cannot reconcile the slide with the fundamentals.
The explanation lies in a structural shift within the AI market itself. The initial boom was driven by training large language models, a task tailor-made for Nvidia’s GPUs. That is now giving way to inference and agentic AI, where memory, CPUs and cost per inference step matter more. Huang has declared 2026 “the year of inference” and expects the inference market to eventually surpass training in size. But whether this transition widens Nvidia’s lead or dilutes its dominance is precisely what investors are wrestling with.
Nvidia’s next-generation Vera Rubin platform, which entered mass production in June, is designed to address that shift. It combines Rubin GPUs with Vera CPUs and speech-processing units to deliver up to 35 times more throughput per watt than the preceding Blackwell architecture. An earlier bottleneck related to HBM4 memory has been resolved — the new memory delivers around 22 TB/s of bandwidth per GPU, triple the HBM3e in the previous generation. SemiAnalysis projects that Nvidia’s datacenter revenue could exceed consensus expectations by as much as 20 percent in the second half of the year.
One significant design change has been confirmed: the planned “Rubin Ultra” configuration with four chips has been dropped. It will be replaced by a more compact version with half the die size and proportionally adjusted performance. The long-term revenue implications of that simplification remain unclear.
Nevertheless, the forward order book provides a powerful vote of confidence. On the Bank of America Global Technology Conference in June, CFO Colette Kress put Nvidia’s supply commitments at roughly $124 billion. Companies do not secure nine-figure delivery volumes years in advance unless they are certain of demand. Huang’s team now expects cumulative demand for Blackwell and Vera Rubin systems to reach at least $1 trillion by the end of 2027 — double the previous year’s forecast.
Should investors sell immediately? Or is it worth buying Nvidia?
Kress also pushed back against the notion that custom chips are commoditising Nvidia’s hardware. She argued the opposite: agentic AI — software that autonomously plans, uses tools, makes decisions and runs continuously in the background — requires far more compute per task than simple query-and-answer models. If agentic AI scales as Nvidia expects, it could create a permanent additional layer of inference load on top of existing training demand. That is the bull case in its purest form.
Nvidia is also laying groundwork beyond hardware. The Vietnam Research and Development Center, opened in December 2024, is set to become a global AI engineering hub, according to vice president Steven Truong, who spoke at a Hanoi conference on June 30. The local business grew by roughly 130 percent between 2025 and 2026, while the engineering team expanded by 90 percent. Nvidia plans to build so-called “AI factories” in the region — infrastructure designed to reduce reliance on traditional cloud providers.
On the price front, the stock closed June at $200.09 on the New York exchange, up 2.63 percent for the day, after having lost about eight percent mid-month. The corresponding German listing stood at 175.22 euro. Technically, the market is neither panicked nor euphoric: the relative strength index sits at 46.4, shares are roughly seven percent above their 200-day moving average of 163.92 euro, and the consensus analyst target implies upside potential of 51 percent at 264.53 euro.
For the second half of 2026, the setup mirrors the conditions that drove the stock higher in 2025: accelerating revenue growth, stable margins, hyperscaler demand outstripping supply, and a new platform launch. Yet investors remain cautious, weighing fears of an AI bubble, the cyclical nature of industry financing, and rising geopolitical tensions. The inference era has arrived. Nvidia has the infrastructure, the order book and the next platform ready. Whether the market chooses to reward that logic before year-end depends less on the numbers — which are solid — and more on a collective willingness to believe them.
Ad
Nvidia Stock: Buy or Sell?! New Nvidia Analysis from July 1 delivers the answer:
The latest Nvidia figures speak for themselves: Urgent action needed for Nvidia investors. Is it worth buying or should you sell? Find out what to do now in the current free analysis from July 1.
Nvidia: Buy or sell? Read more here...
