Google Lays Down $300 Million Investment in OpenAI Rival Anthropic

Staving off competitors, locking-in cloud hosting business, and more

Feb 06, 2023

The Financial Times reported last week that Google made a $300 million investment in generative AI pioneer and OpenAI competitor, Anthropic. What does Anthropic have that Google doesn’t already have internally? The answer may be nothing. Google may simply be hedging its bets and ensuring a rival tech giant doesn’t get a foothold in the company like Microsoft did with OpenAI.

And there is always the cloud hosting angle. Microsoft’s CEO Satya Nadella has made several statements about how the Azure cloud service has not only hosted OpenAI’s model training, it will be both the key hosting provider for OpenAI’s runtime services and for other companies looking to train large AI models.

Google would certainly like to be competitive in these cloud hosting segments as well. Notably, the company has not publicly revealed its investment in Anthropic but did announce a new Google Cloud partnership with the AI innovator.

Safe, Reliable, Interpretable, Steerable AI

Anthropic describes itself in terms of its mission and the technology it uses.

“Anthropic is an AI safety and research company that’s working to build reliable, interpretable, and steerable AI systems. Large, general systems of today can have significant benefits, but can also be unpredictable, unreliable, and opaque: our goal is to make progress on these issues. For now, we’re primarily focused on research towards these goals; down the road, we foresee many opportunities for our work to create value commercially and for public benefit.”

Some of Anthropic’s co-founders spent time at OpenAI before starting the company, which has raised $1 billion over the past year and has no products in the market. However, it does have a new ChatGPT rival named Claude, which is currently in a closed beta release and is developing new approaches to large language model training.

A Competent ChatGPT Rival?

Riley Goodside and Spencer Papay of Scale have early access to Claude and did an in-depth write-up comparing it to ChatGPT. Their conclusion:

Overall, Claude is a serious competitor to ChatGPT, with improvements in many areas. While conceived as a demonstration of "constitutional" principles, Claude is not only more inclined to refuse inappropriate requests, but is also more fun than ChatGPT. Claude’s writing is more verbose, but also more naturalistic. Its ability to write coherently about itself, its limitations, and its goals seem to also allow it to more naturally answer questions on other subjects.
For other tasks, like code generation or reasoning about code, Claude appears to be worse. Its code generations seem to contain more bugs and errors. For other tasks, like calculation and reasoning through logic problems, Claude and ChatGPT appear broadly similar.

Given this assessment and others, you can see why Google may be interested. Even if Google solutions such as LaMDA and PaLM are powerful, having another option in this market has clear value. Google also must consider what would happen if another well-resourced company started using Antropic technology to build competitive solutions. One way to limit competition is to acquire competitors, and another is to restrict access to the technology that competitors need.

Microsoft is about to introduce an upgraded Bing search engine that leverages the technology behind ChatGPT. That could represent the first significant threat to Google’s search dominance in 15 years. If Google had been the big early investor in OpenAI, it might have staved off this threat from Microsoft before it germinated.

Google’s Motivation to Act

Google may have become complacent about the pace of generative AI innovation and how quickly it needed to act. There is no company today that has as broad of a generative AI research and product portfolio as Google. Beyond the large language models like LaMDA, it has Sparrow from the Deepmind division, Imagen and Phenaki in AI-based image and video generation space, and much more. But where are all of the products and APIs to access these technologies?

Synthedia wrote last month about the company’s plans to introduce 20 new AI-based products in 2023. Some of those products are on the mundane side, but others may have promise as standalone offerings or could power features adding significant value to existing products like search.

We expect to see many of these products demonstrated at Google I/O in May. It is not clear that many of them would have seen much progress if Google CEO didn’t issue a “Code Red” and the “Green Lane” initiative in response to the widespread popularity of ChatGPT.

The Anthropic investment could have been a long time in the making, but that seems doubtful. Even if the companies were talking about this, ChatGPT is almost certainly the catalyst for this to happen now and at this scale.

Why Anthropic May Need Google

There may be a reason why this was important to Anthropic. The company raised $580 million in 2022 from former FTX CEO Sam Bankmen-Fried and Caroline Ellison, who led the company’s Alameda Research division. You are probably aware that FTX is in bankruptcy proceedings due to alleged fraud, and the courts are attempting to recover some of the funds.

It is unclear what impact that might have on Anthropic’s balance sheet, but it would be a motivation to take in more funding. The interest around ChatGPT may also have created a windfall in higher valuation as well, so the timing could be opportunistic. Regardless, training large language models is not inexpensive. OpenAI has required regular influxes in cash to fund its operations over the years. It will be interesting to see if the Google investment is merely replacing funding that originally came from the FTX founder or if it is additive.

There is speculation that Microsoft’s original investment in OpenAI was largely in terms of access to Azure cloud computing resources and that its recently rumored $10 billion investment may have been as well. Given the Google Cloud announcement about a “partnership” with Anthropic, it is likely that a substantial portion of the investment is in computing resources. While this may look on the surface like a battle of the AI wars, it is also a part of the larger cloud computing wars.