Technology

Nvidia probably won’t be thrilled that it drew the US justice department card

Nvidia controls 80 per cent of the AI chip market and produces custom AI processors for market giants including Microsoft, Google and Amazon for their cloud services

Jensen Huang, left, co-founder and chief executive officer of Nvidia Corp during the Computex conference in Taiwan. Photograph: Annabelle Chih/Bloomberg

Karlin Lillington's face
Karlin Lillington
Thu Jun 13 2024 - 06:00

Earlier this year, I had an insightful conversation with our regular, 20-something DPD delivery driver.

When I explained the large box he’d just set down by door was a gaming PC – for me – his eyes widened and he grinned delightedly.

I’m pretty sure he hadn’t expected a woman older than his mother to be in the market for a gaming rig. His immediate response was (of course) to ask about the specs, particularly the graphics card, the element that handles the extremely demanding graphics in modern video games.

Gamer advice is always to get a PC with the best graphics card you can afford, as it makes the most difference to the quality of the gaming experience. My card? An Nvidia GeForce RTX 4060, not the top of the line but more than sufficient for gaming-novice me.

READ MORE

Then the discussion got particularly interesting. During the Covid lockdowns, he’d decided to build his own gaming PC, ordering the components online. “The problem was, I couldn’t get a decent graphics card,” he said.

All the best cards, such as those Nvidia makes, had been snapped up by people mining bitcoin and other cryptocurrencies, he said.

What? Why?

He shrugged. “They do really fast calculations.”

Ah, of course. The cards are little computing powerhouses, shouldering the heaviest gaming calculation tasks. Despite the name, “cards” are not small and thin, but substantial metal and plastic units ranging from card-deck to elongated videocassette size, and can exceed 30cm and 1.5kg.

[ Nvidia’s surge masks narrow US bull market ]

[ Nvidia rewrites the laws of growth ]

Each holds over a billion transistors and thousands of integrated circuits. Performing the complex calculations behind the visuals that deliver an immersive gaming experience also generates heat, so the cards have heatsinks, fans and sometimes even water cooling systems.

The best can cost over €1000. In any modern desktop or laptop, the graphics card is likely its largest and most expensive component.

Nvidia had been everywhere in the news as its value continued to skyrocket.

Last week, it even briefly overtook Apple to become the world’s most valuable company, surpassing $3 trillion in valuation.

What is going on? Has the entire world suddenly decided it needs to play Baldur’s Gate 3 on a top-end PC?

Think more along the lines of ChatGPT. Just as with mining cryptocurrencies, artificial intelligence applications demand hard-core computing power to grind through difficult calculations. That’s exactly the kind of task something like a graphics card can handle, though what Nvidia sells to the AI sector are the cards’ upmarket, pricey relatives, specially designed AI chips.

Right now, if a company can provide the white-hot AI sector with exactly what it needs, it pretty much has a license to print money (as well as mine Bitcoin).

Nvidia controls 80 per cent of the AI chip market and produces custom AI processors for market giants including Microsoft, Google and Amazon for their cloud services. Reuters notes that its market dominance gives Nvidia plump gross margins of around 70 to 80 per cent.

But Nvidia’s astronomical success is raising antitrust concerns with regulators in the US, EU and UK. Fears are growing because the AI sector is dominated by a tiny handful of powerful, hard to scrutinise companies, particularly Nvidia, Microsoft and OpenAI.

The single most powerful regulatory stick US regulators wield is antitrust law, which carries muscular punishments, including the breaking up of a company

Regulatory pressure already prompted Nvidia to drop its proposed $40 billion acquisition of dominant mobile chip designer Arm in 2022, though it still holds a company stake worth around $150 million.

Some initial US investigations and inquiries have begun into AI companies, including a Federal Trade Commission (FTC) probe into whether ChatGPT’s parent company OpenAI harmed consumers via its data collection tactics, and an examination of strategic partnerships between tech giants and AI companies, such as Microsoft with OpenAI, and Amazon and Google with Anthropic.

More weighty antitrust investigations had stalled for months over uncertainty as to which regulator had responsibility for which companies. Last week, though, the US justice department and the FTC came to an agreement that should clear the way to move forward against Nvidia, Microsoft and OpenAI.

The justice department will tackle Nvidia, while the FTC will take on Microsoft and OpenAI. That won’t be welcome news for Nvidia as it basks in its current valuation glory, nor the other companies – perhaps especially Microsoft, which has been through (and lost) such a toll-taking, time-consuming trial before.

The single most powerful regulatory stick US regulators wield is antitrust law, which carries muscular punishments, including the breaking up of a company.

Nvidia probably won’t be thrilled that it drew the justice department card, either. Due to the lax attitude of successive federal administrations, US antitrust regulation has lain dormant since that turn-of-the-millennium Microsoft prosecution.

The presiding judge then was often criticised for failing to break up Microsoft. Now, just two decades later, Microsoft is yet again the subject of a large antitrust investigation. That should have regulators, particularly the Justice Department, ruminating that splitting apart the three biggest companies in the already narrow AI space might save considerable headaches, and further costly prosecutions, in the AI future to come.

READ NEXT
LATEST STORIES