Competition in the AI sector is heating up. Here's how it's impacting business and academia.

The Stanford report shows a nuanced picture of how the technology sector is changing in light of the AI boom. Overall global corporate investment in AI actually peaked in 2021 and has fallen the past two years, down to $235 billion in 2022 and $189 billion in 2023.

In part, this decline can be attributed to high interest rates. In 2022, central banks around the world started to combat inflation by raising interest rates. This in turn has made companies more hesitant to make as many large-scale investments as they otherwise would have.

But investors may have reason to be wary; many high-profile AI companies have struggled to stay afloat despite producing innovative technology. Take, for instance, Stability AI, the creators of a leading image-generation model. The company raised funds at a $1 billion valuation in 2022 but has struggled to turn a profit. According to reports by Reuters and The Information, the company recently installed a new CEO, raised $80 million, and settled about $400 million in debt.

Inflection AI, the startup behind the chatbot Pi, raised $1.3 billion in 2023 at a $4 billion valuation but has been unable to find a business model. In March 2024, Microsoft, which had previously invested in Inflection AI, all but acquired the company, according to Bloomberg. The Washington-based tech giant agreed to pay Inflection $650 million to license its technology and hire most of its employees.

Though overall investments in AI declined, generative AI is an exception to this downturn. Investment in generative systems that create text and images soared to $25 billion in 2023, up from roughly $3 billion a year earlier. Business formation is also trending upward. About 1,800 new AI companies received funding in 2023, an increase of about 40% from 2022.

The U.S. has by far the hottest AI market, accounting for $67 billion of investment in 2023, compared with just $7.8 billion in China and $3.8 billion in the United Kingdom. U.S.-based institutions were behind 61 notable AI models in 2023, compared with China (15), France (8), and Germany (5), according to the Stanford report. America's dominance is not without precedent: The country has led in private AI investment since 2013—a gap that has only widened in recent years.

The generative AI boom has significantly impacted the academic research landscape. According to the latest data, academic journal articles in the field surged to over 240,000 in 2022, marking a dramatic 175% increase from 2010. Machine learning studies have seen particularly explosive growth, with publications soaring from around 6,000 in 2010 to over 70,000 in 2022, underscoring the rapid advancement and growing interest in the sector.

The surge of interest in AI has also changed where academic research is done. Up until the mid-2010s, most leading machine learning models came from academia. However, university researchers now lag behind their private-sector counterparts because they have only a fraction of the resources in comparison.

Tech companies can offer researchers much higher salaries—and massive amounts of compute. GPT-4, the model behind ChatGPT, has been estimated to cost $78 million to train, while Google's Gemini Ultra has been estimated to cost $191 million. These numbers are high in the corporate world but would be considered astronomical in academia.

All of the commercial interest in AI has come with a cost. Historically, researchers have been keen to share their models and data with others. But industrial labs, which have invested billions in research and development, are much more cautious about collaboration. OpenAI, for instance, was founded as a nonprofit to share its work with the public out of fears that one leading tech company might monopolize the technology. It open-sourced GPT-2, an earlier chat model, but has since declined to share its models with the public, citing safety concerns.

Other researchers have followed suit. In February of last year, Google's head of artificial intelligence announced his staff should hold off on publishing their work, marking a sharp reversal of their policy pre-ChatGPT. Google, whose researchers made many of the discoveries driving current generative AI models, has so far failed to capitalize on its work, despite a generous head start.

Francois Chollet, a computer scientist at Google, recently lamented on a podcast that "frontier research was no longer being published," blaming OpenAI for creating the current competitive dynamics.

The AI sector is booming. But many tech leaders now worry that their companies might not come out on top.

Story editing by Alizah Salario. Additional editing by Kelly Glass. Copy editing by Tim Bruns.