Artificial intelligence is bringing together fishes big and small in technology. Microsoft Amazon.com Alphabet and Nvidia’s deals alone made up about one-third of the roughly $70 billion raised last year by data- and AI-related startups. Backing fledgling firms such as ChatGPT-maker OpenAI and rival Anthropic holds out the promise of future revenue. The concern is that mounting regulatory and financial risks will offset the gains.
For software developer Microsoft, e-commerce titan Amazon and internet search kingpin Alphabet, one clear rationale is that the funding targets are hungry for the computing power all three provide. In exchange for equity, the tech goliaths typically offer a combination of cash and free or discounted cloud services. When Amazon unveiled plans last September to inject $4 billion into Anthropic, the startup committed to spending the same amount on Amazon Web Services. In theory, Nvidia also benefits because it sells the high-performance processing chips powering AI.
Although the foursome helped write a combined $23 billion of AI funding checks in 2023, according to Altimeter Capital investor Apoorv Agrawal, using data from industry tracker PitchBook, the buzz is widespread. Salesforce which sells software used to manage customer relationships, last year started a $500 million fund dedicated to AI ventures. Even data analysis company Databricks, which has yet to go public itself, backed six AI firms in 2023, according to business information outfit Crunchbase.
Scrutiny is rapidly intensifying from multiple corners. The U.S. Federal Trade Commission’s trustbusters just opened an investigation into whether Microsoft, Amazon and Alphabet’s deals involving OpenAI and Anthropic are hindering competition. In addition, Bill Gurley, who led Benchmark Capital’s early investments in Uber Technologies Zillow and others, also has raised questions about whether companies doling out credits for their web services to startups should be able to recognize them as revenue, arguing that such codependency could lead to a “massive mess.”
For the startups themselves, which incur high costs to train and run AI models, their businesses may not prove sustainable once credits run out. There also may be legal bills from copyright infringement and, conversely, the expense of paying for information to feed computers trying to replicate human intelligence. Moreover, unlike traditional venture capitalists, bigger companies often have non-financial considerations, including the strategic knowledge they can access. These aspects could make them less price-sensitive, which probably helps explain why valuations for unprofitable AI startups are so high. Reuters