After the market rose 25%, Nvidia went "crazy"

    Last night, Nvidia released its financial report. Judging from the numbers, the performance of this chip giant is mediocre.

    The most valuable U.S. chipmaker said on Wednesday that revenue fell 13% to $7.2 billion in the latest quarter, topping expectations of analysts polled by FactSet. Net income rose 26% to $2 billion. The decline in sales was due to a sharp decline in the graphics chip business for gamers, who have cut back on spending after the pandemic slowed and are only just beginning to resume spending.

    Huang said big data center operators were adapting their computing infrastructure to better capture the opportunities offered by artificial intelligence, creating a surge in demand for their chips. Nvidia's data center revenue hit a record $4.28 billion in its most recent quarter, which Kress said reflected strong demand from consumer internet companies and cloud computing companies.

But after handing over the financial report, Nvidia's stock price rose by about 25% after hours. This is due in part to the company's strong expectations for the future.

01. Generative AI promotes a surge in chip sales, exceeding expectations by 50%

    Nvidia said it expects sales of about $11 billion in the current quarter, an increase of 2%, more than 50% above Wall Street estimates of $7.15 billion.

    Nvidia shares were up 109% so far in 2023 before after-hours trading, largely driven by optimism over the company's dominance in the artificial intelligence chip market. Nvidia CEO Jensen Huang said the company has seen a "surge in demand" for its data center products.

    As you all know, since late last year, the tech industry has been racing to develop ever larger AI models, leading some customers to privately worry about a shortage of the H100 chip, which only went on sale earlier this year. That helped the company's data center sales outperform even the most optimistic analyst forecasts, with the company saying sales of both the H100 and A100 chips were strong based on its previous chip architecture.

    Nvidia Chief Financial Officer Colette Kress pointed out that further surges in demand will go well beyond this quarter, but she said it is too early for the company to issue long-term financial guidance. Visibility into future demand has "extended for several quarters" and has resulted in Nvidia purchasing "significantly higher supply in the second half of the fiscal year," she said.

    Sales to the company's data center customers could double within three months, according to Nvidia's forecast. The growth comes from customers across the board, with consumer internet companies, cloud computing providers and enterprise customers eager to apply generative artificial intelligence to their businesses. in the business.

    Nvidia Chief Executive Jensen Huang told Reuters in an interview on Wednesday that the company began full production of its latest AI chips last August, giving chatbot apps some cushion of supply as they take off in popularity. "In January, there was a surge in new demand," Huang said. "We had to place additional orders and we sourced more supply for the second half of 2023."

    According to Huang, operators of big data centers are restructuring their computing infrastructure to better respond to the opportunities presented by artificial intelligence, creating a surge in demand for their chips.

    Jensen Huang further noted that the company is well-positioned to benefit from the AI opportunity, as it began producing a new wave of advanced equipment for data centers when interest in it exploded last year. Apple capitalized on the shift to smartphones about 16 years ago by releasing its advanced handsets. "All the technology coming together to help everyone realize what an amazing product it can be and what it can do."

    "As companies race to apply generative artificial intelligence to every product, service, and business process, the $1 trillion installed global data center infrastructure will transition from general-purpose computing to accelerated computing," Jensen Huang said.

    "Given that the generative AI gold rush is happening, this should boost demand for Nvidia chips for the rest of the year," said Edward Jones analyst Logan Purk.

    "The company is very optimistic about what it's seeing," said Adam Crisafulli, analyst and founder of Vital Knowledge media. "In addition to the huge revenue tailwind that AI is bringing to the company, Nvidia's gross margins have now largely recovered to "Nvidia's sharp surge has the potential to shed light on subdued sentiment across Asian markets," said IG Markets analyst Hebe Chen. "However, general pessimism over the debt ceiling crisis and monetary policy is likely to persist."

02. More manufacturers enter the game to grab GPUs, and the demand continues to soar

    Judging from recent news, it appears that the demand for Nvidia GPUs is just the beginning.

    Early last month, shortly after calling for a six-month moratorium on AI research, Tesla CEO Elon Musk was building an AI startup and secretly bought 10,000 GPUs, according to Musk. The saying at the time - even the dogs are robbing the GPU.

    “Graphics processing units (GPUs) are harder to come by than drugs,” Elon Musk told an interviewer earlier this week, according to Reuters.

    Meanwhile, cloud providers are building up armies of GPUs to provide more AI firepower.

    Nvidia said it is working with Amazon.com, Microsoft and Google units to develop generative AI, and with cloud computing companies to help small businesses use generative AI. The company said Tuesday that it is adding its artificial intelligence software to Microsoft's Azure cloud computing service, allowing enterprise customers to leverage its chips and software to accelerate large-scale generative artificial intelligence systems.

    At the annual Google I/O developer conference held earlier, Google also announced an AI supercomputer with 26,000 GPUs - A3. This supercomputer is Google and Microsoft's struggle for AI supremacy. Yet another evidence of an aggressive counter-offensive. The supercomputer reportedly has around 26,000 Nvidia H100 Hopper GPUs. For reference, Frontier, the world's fastest public supercomputer, has 37,000 AMD Instinct 250X GPUs.

    In addition, Meta, which had been clinging to the CPU before, is also speeding up the acquisition of GPU.

    Over the past decade or so, Meta has reportedly spent billions of dollars recruiting top data scientists and building new types of artificial intelligence, including now powering discovery engines, moderation filters, and ad recommenders across its apps and services. Powered by artificial intelligence. But the company has struggled to turn many of its ambitious AI research innovations into products, especially when it comes to generative AI.

    But until 2022, Meta will run its AI workloads primarily using a combination of CPUs and custom chips designed to accelerate AI algorithms -- CPUs tend to be less efficient for such tasks than GPUs. After taking stock of its own progress, Meta scrapped plans to roll out custom chips at scale in 2022 and instead ordered multibillion-dollar Nvidia GPUs, which would require a major redesign of several of its data centers.

    As foreign media reports have said, maybe one day, Meta will entrust most of its artificial intelligence workloads to the self-developed MTIA chip. But for now, the social network relies on GPUs in its research-focused supercomputer, the Research SuperCluster (RSC). To this end semianalysis believes that Meta is the only individual buyer of Nvidia H100 GPUs this year

    Debuting in January 2022, the RSC, assembled in partnership with Penguin Computing, Nvidia, and Pure Storage, has completed a second phase of expansion. Meta says it now contains a total of 2,000 Nvidia DGX A100 systems with 16,000 Nvidia A100 GPUs.

    As for China, although the latest A100 and H100 cannot be grabbed, the domestic enthusiasm for Nvidia GPUs has not diminished. According to foreign media reports, the artificial intelligence boom brought by ChatGPT has led to a surge in demand for high-computing GPUs, coupled with the United States banning the sale of high-end AI chips to China, for which large Chinese companies are buying Nvidia’s GPUs, which has led to A800 and H800 Both chips cost 40 percent more than their original MSRP,

    According to reports, the current price of the A800 is about 36,500 US dollars. The delivery time is also limited to three months, and the delivery time of the latest order is even more than six months.

03. Write at the end

    Although Nvidia is at the starting point of another rising cycle, the company's market value is far ahead of TSMC, Qualcomm, Intel and other semiconductor star companies. The company is also moving towards a market value of trillions. But Huang Renxun emphasized that the competition to supply AI chips to meet demand is fierce.

    "We face competition from all directions," he said, from established semiconductor companies to startups. So in order to better compete in the market, in addition to providing chips to customers, Nvidia has turned to selling entire AI supercomputing systems to large companies that want to own and are willing to pay Nvidia's prices and gross margins in dollars to get the same AI expertise that rivals Silicon Valley tech giants.

    "No company can build a state-of-the-art AI data center without (the cloud provider's) technology and all the software, but we have all these capabilities," Huang said. "Enterprise is a very, very different market."

    Referring to the Chinese market, he said Chinese companies started making their own chips to compete with Nvidia*'s gaming, graphics and artificial intelligence processors. Because if [China] can't buy from the US, they'll just build it themselves. So he thinks the US has to be careful. Because China is a very important market for the technology industry. "If we are deprived of the Chinese market, we have no contingency. There is no other China, there is only one China," Huang said. At the same time, he emphasized that if American companies cannot trade with Beijing, it will "cause great damage to American companies."

    Although Huang Renxun is prepared for danger in times of peace, as Bank of America analysts said last month, due to the surge in demand, sales of NVIDIA accelerator chips may surpass x86 CPU chips in 2023.

    Well-known analyst Lu Xingzhi also pointed out that Nvidia relied on AI GPU to defeat Intel's data center revenue in one fell swoop and become the leader in data centers and AI chips. In his view, Nvidia will replace Apple and become the new savior of TSMC.