This analysis was first published in SvD Näringsliv, in Swedish, on May 8th, 2023. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.
Behind the hype — and the anxiety — surrounding next-generation AI lies a simple question. Who controls the essential hardware? The answer has geopolitical implications.
Should you mine for gold yourself, or sell pickaxes and shovels to those who do?
The question from the gold rush era keeps returning to California. What was once a literal business opportunity in 1848 has become a metaphor for a particular kind of business model — one that enables other companies to succeed.
Now the next gold rush appears to be starting: the explosion of artificial intelligence. And it has focused attention on a specific type of computer chip required to perform the most advanced calculations. Put simply, without the right chip, large-scale AI doesn’t happen.
At the center of this accelerating trend are companies that were previously consigned to being under the hood — invisible, but entirely essential. No company has benefited more from this moment than American chipmaker Nvidia.
Nvidia’s involvement in the AI segment started early. Geoffrey Hinton — often called the “godfather of AI” — built a product back in 2012 that performed advanced image recognition: software that can understand what a photograph depicts. The results were seen as a major breakthrough in AI, and underpin products used by millions of people today.
According to Hinton, it would not have been possible without Nvidia’s chips. Those chips weren’t originally designed for this purpose — they were primarily used for video games. But their architecture turned out to be particularly well suited to this new application as well.
Looking at the most discussed AI product today, ChatGPT, it is believed to have been trained using 10,000 GPU chips from Nvidia. When Sundar Pichai, CEO of Google’s parent company Alphabet, recently presented his quarterly results, he explicitly cited his company’s access to Nvidia chips as a competitive advantage.
The attention has translated into financial results. Nvidia’s share price has risen nearly 100 percent so far this year. Compare that to rivals Intel and Qualcomm, which are up 11 and 8 percent respectively.
The enormous demand for this type of chip is now creating complications.
Early in the pandemic, there was a major shortage of a different kind of chip — semiconductors — which led to delays in everything from cars to consumer electronics. That shortage has since been resolved, partly due to lower demand.
The constraints on AI-grade chips stem from the fact that their manufacture is extremely advanced, and only a handful of actors globally can produce them. Nvidia doesn’t fabricate its own chips; it uses suppliers for that. Among them is Taiwanese company TSMC — also the world’s largest semiconductor manufacturer.
Securing access to the right type of chips is becoming a geopolitical question.
Today, China spends more money importing various types of chips than it does importing oil, according to the book Chip War by Chris Miller, professor of history at Tufts University. China is also attempting to circumvent the American blockade on certain chips by renting access rather than importing them outright.
Both the US and EU have launched initiatives to bring more chip manufacturing to their respective regions. The European Chips Act is one such effort — committing roughly 490 billion kronor to increase Europe’s share of global chip production from 10 to 20 percent. The American equivalent, the CHIPS and Science Act, is similar in ambition and scale.
While manufacturing is concentrated among a handful of global players, chip design — how the chip actually functions — is becoming an increasingly important competitive differentiator. Apple is the clearest example of a company that has shifted strategy, moving away from off-the-shelf chips in favor of custom-built variants engineered for specific functionality. Apple’s new chips are notable in particular for their ability to perform AI calculations on-device.
Amid the explosion of AI products, there are more fundamental questions beneath the usual ones about ethics and existential risk, about who owns AI models and the concentration of power in the field.
Under all of that lies a very simple and practical question: who has access to the right physical chips? Because access to hardware may prove to be a decisive factor in how advanced an AI you can develop. It’s easy to understand why politicians, companies, and entire countries are starting to get nervous.
