The Nvidia wave is sweeping the tech sector

Jensen Huang, the Nvidia CEO, has been claiming for years that data centres around the world will have to undergo a complete overhaul to meet the needs of generative AI. He claims that $1tn will be needed over the next 4 to 5 years to train and run new AI models. This is essentially double the amount invested in digital infrastructure.

Nvidia has been the biggest beneficiary. The company’s stock price has risen at an incredible rate, making it the most valuable in the world.

A spate of announcements of earnings and AI-related deals in the past two weeks have also provided encouraging evidence of the spread of the boom that was sparked by OpenAI’s generative AI ChatGPT. There is no way to tell if the surge in spending will continue or be large enough for tech stocks to rise as much as they have. But it’s at least comforting the bulls.

Since Broadcom reported its latest AI-driven boost to its sales, its shares have increased by more than 20%. If it continues to rise at this rate, it will join the elite group of tech firms valued at over $1tn. That’s more than six-times what it was worth a half decade ago.

Broadcom’s AI accelerators – the chips it custom-designs to help customers like Google speed up AI calculations – have been a major factor in the growth. The rapid growth of Broadcom’s data centre business is also a sign that high-speed networks are now playing a more important role.

Due to the massive amount of data needed for AI training and running, it was necessary to have much faster connections both between processors as well as among machines in data centres. Broadcom CEO Hock Tan estimates that networking will make up 40% of Broadcom’s AI chips sales by the year’s end.

Oracle shares, which were late adopters of the cloud, rose 17 percent after announcing a deal with OpenAI to train its large language models using Oracle’s cloud infrastructure. Microsoft’s Azure cloud service will be brought to an Oracle data center, bringing together two of the most vehement tech industry enemies.

Wall Street has finally given Hewlett Packard Enterprise a break. It had appeared to be missing out on a booming demand for AI server that was boosting rivals Supermicro, and Dell. Investors reassessed the company’s position in the AI boom, and its shares rose by 24 percent after earnings.

As this news has fueled hopes that the AI boom will spread to more suppliers, a few things have become clear. The impact appears to be widespread and affecting many parts of the “tech-stack” – the hierarchy of components, from chips to software, that is required to run the complex IT systems of today.

Nvidia remains the clear leader. The majority of its sales come from entire servers that are often networked in racks. Nvidia’s proprietary technologies, such as networking, are used to optimize the performance of each system.

Nvidia’s largest customers want to reduce their dependence upon the company. They are pushing for new standards, from AI software to networking, that will allow other competitors to emerge. These initiatives will take some time.

The largest tech companies are expanding their direct involvement into more infrastructure components required for AI. Apple’s AI announcement from last week included news that the company is building its own servers based on its in-house chip design. Apple already controls most of the components that go into its phones. A similar move will likely be made in the data center as AI demands it to process more data in-house.

As a result, business models have changed. Companies like Broadcom are now playing a supporting role as customers gain more control. As the “hyperscalers”, or largest cloud companies, account for an increasing share of demand, this leads to a greater dependence on a smaller number of large customers. This will make suppliers more vulnerable to any economic downturn. Wall Street is focused on how many tech boats will rise with the tide of generative artificial intelligence.