NVIDIA Corporation (NVDA.US) conference call: Amazon.com, Inc. will deploy an additional 2 million GPUs, boosting the value per GW to $40 billion thanks to Vera Rubin.

Faced with a severely constrained supply chain and overwhelming demand for "Agentic AI," Nvidia issued a strong guidance predicting a 70% growth for the next fiscal year after reporting a record revenue of $96 billion for Q2. The company stated that this figure still does not fully reflect the true market demand. On August 26, 2026, local time, Nvidia held a conference call to discuss its Q2 earnings report for the fiscal year 2027. During this high-profile performance meeting, Nvidia's CEO Jensen Huang and CFO Colette Kress presented remarkable results to the market: - Q2 total revenue reached a record $96 billion, doubling year-over-year; - Revenue from data centers increased by 18% sequentially to $89 billion; - The company expects Q3 total revenue to reach $108 billion (with a 2% margin of error). In the earnings call, CEO Jensen Huang characterized this wave of AI computing investment as a "once-in-a-generation" platform transformation. A key variable in this call was the warning regarding profit margins. Nvidia expects the gross margin for the third quarter to be around 74%, and then to bottom out at 71% to 72% in the fourth quarter, citing that memory price increases are "far beyond previous expectations" and are still climbing. The company anticipates that price adjustments will take effect in the first quarter of fiscal year 2028, at which point the gross margin is expected to stabilize at 72% to 73%. Demand far exceeds capacity: "Without constraints, our growth would be much higher." One of the primary concerns for the market is the sustainability of Nvidia's future performance. In response, Nvidia provided a forward-looking guidance for the next year for the first time, expecting revenue growth of about 70% for fiscal year 2028. However, the management team repeatedly emphasized that this is a "discounted" figure. When answering analysts' questions, Huang candidly noted: "Although our demand is far beyond 70%, our supply gives us confidence in achieving 70% growth. If unconstrained, that number would be much higher. Our year-over-year growth this year has reached 100%, and current unconstrained demand is extraordinarily large, so we are just working diligently to secure more capacity." Describing the current tightness of the supply chain in a down-to-earth manner, Huang said: "The entire supply chain is facing challenges, and everyone is operating at full capacity... I need the entire supply chain to help me." Kress also pointed out that the growth momentum of non-hyperscale enterprises (including ACI & E divisions, such as sovereign AI, regional clouds, and enterprises) is extremely rapid. This sector achieved a revenue of $40 billion in Q2, a 25% increase quarter-over-quarter and a staggering 138% increase year-over-year. Huang remarked: "The computational demands in this part of the world could exceed what we currently see in the cloud as time goes by." The economics of computing power jump: value per gigawatt (GW) doubles to $40 billion. Nvidia's ability to encourage cloud providers and sovereign capital to willingly invest huge sums is largely due to its ever-rising "economics of computing power." During the meeting, Huang revealed some exciting unit economic model data for investors: "In the Hopper era, the revenue opportunity for our data centers was about $18 billion per gigawatt (GW); with Grace Blackwell, it's around $25 billion; and for Vera Rubin, that number is approximately $40 billion." This means that as whole cabinet systems (like NVLink 72) evolve, the value generated by the same scale of electricity and land is rising exponentially. Huang emphasized: "The ideal answer is actually that each gigawatt can generate limitless amounts of computing power. If we could fit one trillion dollars' worth of computing power on one gigawatt of electricity and land, it would be a wonderful outcome. As long as productivity, durability, and versatility continue to grow, people will be eager to invest in those assets that generate revenue, create profits, and recover costs at an incredible pace." Betting on frontier AI laboratories: "My only regret is not investing earlier and more." In response to market concerns about Nvidia's investments in frontier AI laboratories (such as OpenAI and Anthropic) and the potential for "circular financing," Nvidia's management team offered a strong and positive response. Nvidia has invested nearly $50 billion in frontier AI laboratories and, in partnership with capital giants like Blackstone and BlackRock, established a financing platform aiming to raise over $500 billion in third-party capital. Kress stated: "We know some people may call it circular financing, but we dont see it that way. This is a significant shift in computing platforms, and these companies are once-in-a-century opportunities. We expect they will become the largest tech companies in history." Huang openly expressed optimism about these investments: "Investing in these companies is a once-in-a-lifetime opportunity. My only regret is not investing earlier and more." He further stressed: "I am 100% confident that our technology will continue to provide them with extraordinary value, and they will be our customers and partners for a long time." Warning: Extreme inflation in memory prices will put short-term pressure on profit margins, but the AI engine is in full swing. While delivering impressive revenues, the management team proactively signaled expectations of pressured profit margins to the market. Kress noted that the memory market is experiencing "extreme pricing conditions," with price increases surpassing previous expectations. As a result, Nvidia expects Q3's gross margin to be 74%, and in Q4 to fall to a range of 71% to 72%, before stabilizing at 72% to 73% in fiscal year 2028. Kress objectively stated: "We want to make this point clear directly, rather than leaving it as an unresolved issue. Today's memory scarcity is largely driven by AI construction itself." Looking ahead, Huang focused on "Agentic AI." He predicted: "Today, the vast majority of AI is still human-prompted, but I believe last month marked a turning point, and the majority of AI now has agentic capabilities. In the future, every company will have throngs of AI agents... When you have completely agentified systems, you will let those agents run around the clock, working collaboratively with other 24/7 agents. They will continually operate in the background, improving your company and enhancing your life." This article is reproduced from "Wall Street Insights," GMTEight editor: Feng Qiuyi.