Cloud giants Amazon, Microsoft and Google ignite battle over AI

Indeed, Microsoft really kicked things off with its 2019 investment into a somewhat obscure generative AI startup called OpenAI, and followed up by investing again in 2021 and earlier this year with its own “multi-billion dollar,” multi-year commitment.

And Google, too, not to be left out of the party, announced a $300 million investment into Anthropic back in February, an amount that now seems paltry by comparison. The search giant also spread the love, backing video generative AI startup Runway ML.

Still, the news today of Amazon’s big investment signals to me a level of escalation in the competition between the three largest cloud providers (by market share) to partner with, benefit from, and offer their consumers the latest and greatest generative AI technologies.

In one swift move, Amazon has dethroned Google as Anthropic’s primary big tech backer, and gained a potentially hugely valuable ally as demand for, and interest in, generative AI continues to ratchet up across sectors.

Why is this moment qualitatively different than the to-date investments and announcements in generative AI? I can offer a few reasons:

After battling over the cloud, big tech sees AI as the next, most important frontier (and both are inexorably linked)

When most people think of the cloud giants, their mind likely goes to their popular consumer-facing services: Microsoft makes computers and software, Amazon sells stuff over the internet, Google lets you search and email.

And yet, the most recent — and arguably still most important competition — going on between Microsoft, Google, and Amazon prior to AI is the war over the cloud.

With so many of us individuals and our companies generating and relying on data, the cloud has become more important than ever as place to not only store this data, but run applications atop it, including, of course, generative AI apps.

According to market research firm Gartner, the global cloud services market is expected to grow from $491 billion last year to $597.3 billion this year, a 21.7% increase year-over-year, driven in no small part by the boom in generative AI.

“For example, generative AI is supported by large language models (LLMs), which require powerful and highly scalable computing capabilities to process data in real-time,” said Sid Nag, Vice President Analyst at Gartner, when the company released its cloud services forecasting report in April 2023. “Cloud offers the perfect solution and platform. It is no coincidence that the key players in the generative AI race are cloud hyperscalers.”

Current deficiencies could explain the rush to embrace AI

AWS, Amazon’s cloud business, is on its own accounted for nearly 70% of the entire company’s profit last quarter. And as the leading cloud services provider in the world with nearly 40% marketshare, according to Gartner, Amazon is the leader to beat in cloud.

At Microsoft, Satya Nadella rose to become CEO in 2014 after Steve Ballmer lost the mobile race to Apple and Google. Nadella was selected due to his success running Microsoft’s Cloud and Enterprise group, which was responsible for launching Microsoft Azure.

Today, Microsoft Azure is the second largest cloud provider behind AWS and closing in — a respectable place to be in, but you can be damned sure that Microsoft would prefer to be no. 1. If generative AI helps the company do that — by increasing demand from GenAI companies such as OpenAI — then all the better.

Amazon Alexa is not enough

Meanwhile, Amazon’s move to back Anthropic seems wise given the company only last week announced its new LLM-powered Alexa assistant.

The timing of that particular announcement was curious given Amazon’s role as an early leader to conversational AI assistants, launching Alexa as the voice of its first Amazon Echo device back in 2014. Shouldn’t Amazon already have all the in-house data and expertise necessary to field a competitive Gen AI model?

While Amazon’s Alexa is by some measures the most beloved voice assistant, the smart speaker market that it helped launch has begun to plateau, and the competition in that sector has only increased significantly in the last decade as Apple fielded its Homepod with Siri voice assistant and Google offered its Google Assistant on a variety of Nest and Google Home devices. New Amazon CEO Andy Jassy, who took over in 2021, cut employees from Amazon’s devices division and Reuters recently reported that the team there is said to have low morale and a weak development pipeline.

With the current devices strategy running into headwinds, and the first Alexa LLM announcement coming almost a year after OpenAI debuted ChatGPT — by all accounts, a much more capable and powerful AI assistant — it would seem Amazon recognizes it has work to do to catch up in AI, and sees Anthropic as a good shortcut for doing so, even if it risks undercutting AWS’s positioning as a “Switzerland,” i.e., neutral party, when it comes to running AI models and storing their training data.

Of course, undercutting its role as a neutral platform has never stopped Amazon before: the company’s Amazon Basics line of products compete with others from leading third-party vendors on its e-commerce marketplace, and its Amazon Prime Studios division makes movies and TV shows that compete for viewers’ attention with those of the leading film studios, whose titles can also be rented, viewed, or streamed for free through Amazon Prime (and FreeVee).

Google plays catchup

Finally, we get to Google. The company essentially kickstarted the Gen AI revolution of the last half-decade after a number of researchers at its AI subsidiary Google Brain published the seminal paper “Attention Is All You Need” in 2017, outlining the open-source methods to build the kind of transformer models that power OpenAI’s ChatGPT, Claude’s Anthropic, and basically all of the leading consumer and enterprise-facing generative AI models popular today.

But Google with its large and notoriously slow and political internal bureaucracy failed to keep many of these researchers around, and they’ve since gone on to found Gen AI startups that the company is now competing against directly, including Cohere and Sakana AI.

And as VentureBeat’s own editorial director Michael Nuñez covered in his review of the latest updates/Google Apps integrations to Google Bard, the search giant’s Gen AI assistant based on its LLM PaLM 2, Google’s Gen AI efforts have so far failed to impress and pale in comparison in functionality and output utility to OpenAI. Little wonder Google is reportedly moving fast to release a true competitor to OpenAI’s underlying GenAI model GPT-4 , which Google calls “Gemini.”

At the same time, Google is fighting to claw its way up in the cloud service provider rankings from a distant third behind AWS and Microsoft. Behind on both fronts, Google needs a big win: a powerful new Gen AI model could be the proverbial single stone to kill two birds.

Google may have planned to leverage some of Anthropic’s tech to assist in this quest, but with Amazon sweeping in as its new, higher-rolling backer, Google seems like it will have to go it alone in fielding a compelling new AI model and in making its cloud the go-to backend service for Gen AI models and apps.

Where this leaves us

Despite the looming one-year anniversary of ChatGPT’s launch in November, the Gen AI wars are really just beginning in earnest. If the competition a were baseball game, we’d be in the first few innings. And with the rate of competition — and spending — increasing so rapidly, this game may go on to extra innings.

Now, unlike a game, there’s not necessarily a “winner-takes-all” scenario in Gen AI. In fact, the potential applications for the technology and projections for its reach clearly do support a scenario in which there are multiple tech providers, some large and some small, some specialized and some generalized.

And we can’t of course discount the prominence of open source AI. Right now, Meta Platforms is leading the charge in terms of existing tech giants by open sourcing and licensing for commercial usage its Llama 2 LLM.

Llama 2 and other open source models like Falcon 180b and DeciLM 6B pose a way to “level the playing field,” allowing any enterprise access to powerful Gen AI tools to build applications atop of, or modify for their specific uses.

Yet, even in those cases, people deploying and fine-tuning open source AI models will still need cloud servers on which to store data and run inferences, which is why the three big cloud players — Amazon, Microsoft, and Google — are also motivated to deliver preferred AI offerings of their own, or through their partners.

If you’re an enterprise administrator singing up for Microsoft Azure or AWS, why not use one of their suggested AI models instead of trying to build or run your own? That seems to be the thinking motivating some of this partnering and gamesmanship in the Gen AI/cloud space.

No matter what happens, you can be sure the war will be a fierce and expensive one.