跳到正文
The Next Platform·· 2 天前精選AI 評分83

Nvidia 2027 財年第三季收入與淨利創紀錄,上調 2028 財年增長預期

Nvidia Breaks Revenue And Profit Records, And Will Keep Doing It

AI 導讀

Nvidia 2027 財年第三季收入和淨利潤均創紀錄,分別達 $115.8 billion 和 $65.3 billion,並將 2028 財年增長預期上調至較 2027 財年增長 75%。公司預計 2027 財年第四季收入為 $130 billion,上下浮動 3%;雲端產業積壓訂單則升至 $2.5 trillion,三個月前為 $2 trillion。

推薦理由

文章把各業務收入、回購規模與財測放在同一脈絡,呈現數據中心需求及供應擴張如何支撐 Nvidia 的增長預期。

正文 · 原文

Nvidia’s first chairman of the board, Harvey Jones, who was hired in November 1993 to join the Nvidia board seven months after the company was co-founded by Jensen Huang, Chris Malachowsky, and Curtis Priem, wasted no time to quash any thought on the call with Wall Street analysts going over Nvidia's third quarter fiscal 2027 financial results concerning chatter about Huang returning to the chief executive officer position at Nvidia.

And I quote:

NEXTPLATFORM AD

“Nvidia is an exceptionally strong company. Its strength comes from its extraordinary people, groundbreaking technology and exemplary track record of execution. Our financial performance speaks for itself.”

“On November 12, our Board of Directors announced that after disclosure of personal behavior inconsistent with the company's code of conduct, standards and cultural norms, Jensen Huang voluntarily resigned his positions as president, CEO, and board member. The Nvidia board unanimously accepted his resignation and does not intend to reinstate Jensen as CEO. The board appointed Colette Kress, Nvidia’s longtime and well-respected CFO, as interim CEO. In addition, the Board announced it was initiating a search for a permanent chief executive, and we have since retained Heidrick & Struggles to lead the search.”

While this was an unexpected transition, the Board acted deliberately and in the best interest of the company's employees, customers, partners, and shareholders.”

So, that’s that. Unless shareholders vote to throw out the current board of directors and vote in a board that tries to reinstate Huang. Which seems unlikely.

NEXTPLATFORM AD

Nvidia is riding so high and its employees know full well how to ramp up “Vera-Rubin” systems, which they are doing now, as well as get the “Vera-Rubin Ultra” follow-on systems out the door next year and the “Feynman” GPUs and what we presume will be a new and accompanying CPU architecture out the door in 2028. We said as much on November 12 in The Nvidia Machine Is Self-Sustaining – And Was Already Facing A Tougher Future, the day Huang resigned.

Kress finished the call with Wall Street, where she walked through the numbers and answered question along with Jones, in this direct fashion: “Our momentum is real. Our demand is accelerating. Our supply is growing. Our customers and our partners are with us and employees are committed. We've got this.”

So, there you have it.

With that, let’s take a look at the Nvidia third quarter financial results and the upwardly revised forecast for fiscal 2028.

NEXTPLATFORM AD

In the quarter ended in October, Nvidia brought in a record-setting $115.8 billion in revenues, a little more than double the sales it had across all of its products a year ago. Net income was 56.4 percent of revenue, which works out to an astounding $65.3 billion, which also more than doubled year on year.

As I have said before: Just amazing.

Nvidia returned $28 billion back to shareholders during the quarter, with $22 billion in share buybacks on Wall Street and another $6 billion in dividends. Nvidia currently has $227 billion in authorization for share buybacks and just added another $150 billion as the third quarter came to a close. That $337 billion is expected to be spent on buybacks before the end of fiscal 2028. Which is just an astounding number. It took IBM decades to blow that much dough on share buybacks to engineer its earnings per share – the growth of which was how IBM rewarded Wall Street and its executive ranks with stock grants and other bonuses.

Nvidia ended Q3 F2027 with $97.7 billion in cash and investments, up 61.3 percent year on year.

NEXTPLATFORM AD

For the past several years, Nvidia has kept the books with two different divisions: Compute & Networking is one and Graphics is the other.

In the third quarter, Compute & Networking, which has a small portion of sales outside of the datacenter, accounted for just a hair under $106 billion, up 2.1X year on year and up 20 percent sequentially from Q2 F2027. The Graphics group, which sells graphics cards for PCs, laptops, gamers, and scientific and professional workstations, brought in a modest and yet record-breaking $9.85 billion in sales, up 61.5 percent year on year.

Beginning this year, with the datacenter business just utterly dwarfing its workstation, automotive, gaming, visualization, and OEM/IP divisions, Nvidia reclassified itself into different divisions, and this quarter the numbers were restated as a neocloud became a cloud as its business model changed and Nvidia back cast revised data as this company was moved from he Datacenter Hyperscale division to the Datacenter AI Clouds, Industrial, and Enterprise division.

The Datacenter Hyperscale division is what I call the hyperscalers (those that run vast consumer-facing applications) and the cloud builders (those who sell infrastructure and software as services). The AI model builders OpenAI and Anthropic are on their way to be hyperscalers as well, with the way they are spending on IT infrastructure to support their enormous AI training and inference workloads. In the third quarter, these customers spent $60.14 billion on Nvidia wares, up 71.1 percent year on year and up 23.5 percent sequentially.

The Datacenter AI Clouds, Industrial, and Enterprise division – often abbreviated ACIE by Nvidia employees – covers sales of Nvidia hardware to sovereign clouds, neoclouds, industrial companies, and enterprises. Governments and academic institutions are also in this division, although the name does not suggest this. The ACIE division had $47.47 billion in sales, up nearly 3X year on year and it probably will not be long before these ACIE customers, who are focused on inference, will do more spending than the hyperscalers, cloud builders, and AI model builders that have made Nvidia filthy rich.

Kress said on the call that Nvidia’s networking business brought in more than $21 billion in sales, and I have estimated how this break downs across the InfiniBand, Ethernet/Other product lines – other being NVSwitch memory fabrics. I estimate that InfiniBand switches, cables, and optics raked in $9.32 billion in sales, up 2.2X year on year, while Ethernet/Other brought in $11.97 billion, up more than 3X year on year. If you pushed me for a further breakdown for networking, I would estimate that Ethernet switching (which again includes cables and optics in addition to switching) drove $6.64 billion in sales, up 2.6X year on year, while NVSwitch drove $5.43 billion in revenues, up 3.8X year on year.

Datacenter compute, which means both CPUs and GPUs, accounted for $86.32 billion in sales in Q3 F2027, and more than doubled compared to Q2 F2026.

Looking ahead, Kress said that Nvidia was expecting $130 billion in sales, give or take 3 percent, in Q4 F2027, driven by the strong Vera-Rubin platform ramp, and that the company was raising its guidance for F2028, saying it would grow by 75 percent compared to F2027. Only a few quarters ago, Wall Street was expecting F2028 to grow maybe 45 percent, and last quarter Nvidia said it expected 70 percent growth in F2028. The demand is strong for Nvidia’s products, and the revenue increase is really about getting more supply that it can put into the field.

Kress also said on the call that the “cloud industry backlog” was now $2.5 trillion, up from $2 trillion only three months ago. I presume this means hyperscalers plus neoclouds and sovereign clouds, minus industrial, enterprise, government, and academic customers.

來源:The Next Platform · nextplatform.com