NVIDIA's fiscal second-quarter results surpassed expectations across the board, with revenue hitting $96.2 billion, EPS at $2.22, and gross margin reaching 75%. The company's third-quarter guidance of $108 billion, plus or minus 2%, also came in stronger than market consensus. More significantly, management made a rare move by providing an early outlook for fiscal 2028, projecting 70% revenue growth - substantially above the sell-side's 44%-45% estimate and higher than most buy-side forecasts. This could mark a pivotal moment that drives the AI semiconductor sector to new heights.
By breaking convention and offering this unprecedented forward guidance, NVIDIA has dramatically increased the visibility of compute demand. Management confirmed that sovereign AI, consumer internet, and enterprise demand now account for roughly 50% of data center revenue, taking over the growth baton from cloud providers and diversifying the demand structure. Optical modules, networking equipment, and ODM manufacturers stand to benefit from the upward revision in rack shipment forecasts.
The only "flaw" in the quarter was the third-quarter gross margin guidance of 74%, plus or minus 50 basis points, representing a one-percentage-point decline from the second quarter's 75%. Additionally, management did not provide full-year fiscal 2027 gross margin guidance as they did last quarter, reducing EPS visibility. However, viewed from another angle, this represents a trade-off of 1-2 percentage points of gross margin for approximately $100 billion in incremental revenue - a clear value proposition. Management explicitly stated that memory prices will continue to rise next year, making storage ASP trends more resilient, which is a positive signal for the memory industry. Both Micron and SanDisk rose more than 4% in after-hours trading. Short-term gross margin fluctuations do not alter the fundamental trajectory of compute demand continuing to exceed expectations.
Power has emerged as a critical bottleneck in AI deployment, with the industry chain's value extending downstream. Jensen Huang made it clear that NVIDIA is moving beyond simply selling compute power, now collaborating with power generation companies as well as land, facility, and cooling partners. This confirms that electricity is the key physical constraint for AI deployment and signals that the investment radius for AI infrastructure is expanding. Huang also offered strong endorsements for OpenAI and Anthropic, noting that both companies' sales are "soaring," profit margins are "excellent," and they are generating profitable tokens. He predicted they would become "the largest technology companies in history," alleviating market concerns about slowing ARR growth at frontier labs.
AI and semiconductor thematic funds can serve as effective tools for gaining exposure to this industry chain. 易方达AI (03489) covers 50 leading AI companies from China and the US, including NVIDIA, TSMC, and Microsoft, offering a dual-pole allocation of "US compute power plus Chinese applications" - a core instrument for capturing the sustained AI upcycle. Additionally, 易方达亚洲半导体 (03486) focuses on the complete Asian semiconductor industry chain, including SK Hynix and TSMC, directly benefiting from storage ASP resilience and the AI hardware cycle, making it suitable for investors bullish on the storage outlook.