We need more energy-efficient chips

Following Nvidia’s first-quarter earnings report, CEO Jensen Huang stressed that the company is facing overwhelming demand, not a downturn. This is because the company is transitioning from its Hopper AI platform to the more advanced Blackwell system. Huang dismissed concerns that demand might be slowing, saying,

"People want to deploy these data centers right away. They want to use our [graphics processing units] right away and start making money and start saving money. So the demand is very strong."

Nvidia reported excellent results in the first quarter. Adjusted earnings per share reached $6.12 on revenue of $26 billion, up 461% and 262% year-over-year. Non-GAAP operating income for the quarter was $18.1 billion.

Nvidia expects revenue of about $28 billion, plus or minus 2%, for the current quarter, beating analysts' expectations of $26.6 billion.

Nvidia's data center division, which is critical to AI and relies on high-performance server farms, generated $22.6 billion in revenue in the February-April 2024 period, accounting for 87% of its total revenue. Other divisions also saw growth, with gaming and visualization solutions increasing by 18% and 45%, respectively, compared to the first quarter of fiscal 2024. However, the data center division grew at an astonishing rate, increasing 427% year-over-year, as shown below.

Additionally, Nvidia announced a 10-for-1 stock split for shareholders effective June 7, which will take effect June 10 and increase the quarterly dividend from $0.04 per share to $0.10 per share. Nvidia shares rose 6% in after-hours trading following the earnings report.

Huang also stressed that Nvidia’s customer base for chips is growing, including Meta, Tesla and several pharmaceutical companies in addition to major cloud service providers. He specifically pointed out that the automotive industry is an important user of Nvidia’s data center chips.

Nearly 75% of global carbon emissions come from energy production and consumption, mainly from burning fossil fuels to generate electricity. Data centers currently consume 460 terawatt hours of electricity per year, accounting for about 2% of total global carbon emissions. However, as data centers continue to expand, this proportion is expected to nearly triple to 6% by 2030.

Improving data center energy efficiency is critical to reducing carbon footprint and mitigating environmental impact. By adopting more efficient technologies and practices, data centers can play a major role in lowering overall carbon emissions.

Nvidia aims to address growing concerns about the monetary cost and carbon footprint of AI by emphasizing Blackwell's energy efficiency.

A Microsoft expert said that the current deployment of Nvidia H100 will consume as much electricity as the entire city of Phoenix by the end of this year. The most striking thing about the new Blackwell GPU is its energy saving effect, which Nvidia is currently using as a major selling point.

Traditionally, more powerful chips also require more energy, and Nvidia has focused primarily on raw performance rather than energy efficiency. However, when announcing Blackwell, CEO Jensen Huang emphasized its superior processing speed, which significantly reduces power consumption during training compared to the H100 and earlier A100 chips.

Huang noted that using 2,000 Blackwell GPUs to train a very large AI model would consume 4 megawatts of electricity over 90 days, while using 8,000 older GPUs to perform the same task would consume 15 megawatts of electricity. This reduction is equivalent to the electricity use of 8,000 homes versus 30,000 homes.

It is undeniable that Nvidia is at the forefront of the surge in demand for AI applications, driven by major tech giants such as Tesla, Meta, Microsoft, and Alphabet. Recent management comments from these companies highlight the huge potential for Nvidia to expand its business in the AI ​​field.

Tesla’s ambitious plan to increase its use of Nvidia chips by 140% highlights the critical role Nvidia’s GPUs play in training AI models for fully self-driving features and its upcoming self-driving taxis. The massive investment represents a major endorsement of Nvidia’s technology by Tesla CEO Elon Musk.

Likewise, Meta has aggressively invested in strengthening its AI infrastructure, in line with CEO Mark Zuckerberg’s vision to make Meta the world’s leading AI company. As Meta continues to develop its Large Scale Language Model (LLaMA) and Meta AI chatbot, Nvidia’s chips remain an integral part of its AI training efforts.

Microsoft is also experiencing a surge in demand for AI that exceeds its available capacity. The tech giant is investing in its own AI development using OpenAI’s GPT model. However, it plans to increase spending to meet the growing demand, and Nvidia’s chips play a vital role in its cloud service offerings.

Finally, Alphabet’s large capital expenditures in the first quarter were primarily targeted at Google Cloud and advanced AI models, further demonstrating the importance of Nvidia’s technology in driving AI-driven initiatives. While Alphabet uses its chip designs for certain AI tasks, it still relies on Nvidia chips for the needs of its cloud customers.

While most AI runs on renewable energy, there are still concerns about water consumption for cooling data centers. As AI becomes more popular, demand for renewable energy could outstrip supply, prompting some to consider speeding up approvals for nuclear power plants, especially at Microsoft.

Overall, the huge demand for AI computing presents a major opportunity for Nvidia, as reflected in its strong financial results and soaring gross margins. The company's discussion of the energy efficiency of its Blackwell GPUs shows that the company is starting to think about the sustainability of AI.

END

If you like our content, please click "Reading" to share it with your friends.