Wallstreetcn
2023.09.20 10:54
portai
I'm PortAI, I can summarize articles.

Five years ago, when Amazon and Microsoft encountered OpenAI one after another, the gears of destiny began to turn.

In hindsight, it can be seen that Amazon missed the opportunity to collaborate with OpenAI and instead handed it over to their competitor, Microsoft. This can be attributed to the constraints imposed by the typical problems faced by large corporations.

OpenAI+ Microsoft's alliance can be regarded as a model of cooperation and win-win in the current American technology industry. With a bold bet on OpenAI and generative AI technology, Microsoft CEO Satya Nadella and CTO Kevin Scott have become visionary figures admired by everyone in Silicon Valley.

However, the latest report from tech media outlet The Information reveals that it was not Microsoft, but the cloud computing market leader Amazon, that originally had the best chance to lead this wave of generative AI.

As early as 5 years ago, OpenAI had approached Amazon with a cooperation proposal, hoping to obtain support from AWS's cloud computing resources, but it was rejected, and OpenAI turned to Microsoft instead.

The Curse of Being a Leader

In hindsight, Amazon's failure to seize the opportunity to collaborate with OpenAI and instead hand it over to its competitor Microsoft was a result of the typical problems faced by large companies.

As the world's largest cloud computing service provider, Amazon had actually been involved in AI technology for a long time. It even made a donation to support the establishment of OpenAI.

According to The Information, in 2018, OpenAI extended an olive branch to Amazon, hoping to obtain hundreds of millions of dollars worth of computing power from AWS for free, without giving Amazon any equity.

Insiders said that AWS was surprised by OpenAI's bold proposal, but after careful consideration, they chose to refuse. The reason was that they were concerned that allocating a large amount of computing power to OpenAI would affect other customers. Moreover, this collaboration, apart from the reputation of supporting innovation, did not offer any substantial benefits to Amazon.

More importantly, at that time, Amazon was developing its own machine learning platform, SageMaker. The company's top management believed that Amazon had the capability to keep AI technology in-house and saw no need to collaborate with external startups.

Even in 2021, Amazon remained very confident in its own AI capabilities. The Information's report also mentioned that in 2021, Cohere and Anthropic had also proposed similar collaboration invitations, hoping for equity investments from AWS and access to cloud computing resources to integrate their large models into Amazon's products. However, AWS CEO Adam Selipsky and Group CEO Andy Jassy also chose to refuse.

In contrast, Microsoft, despite having a large AI research division at the time, CEO Nadella and CTO Scott chose to go against the tide and invest billions of dollars in OpenAI, a small-scale organization with unverified technology, and provided substantial computing power support.

Especially CTO Scott.

This Silicon Valley veteran, known for his fiery temper and occasional use of profanity during meetings, was a pioneer in transforming Microsoft's outdated and conservative culture during the Redmond era under Nadella's leadership. At that time, Microsoft was in a downturn due to missing opportunities such as the mobile internet revolution. Satya Nadella, who had just taken over as CEO, noticed that Microsoft had a closed corporate culture that favored internal creativity in resource allocation, and employees were resistant to managers entering the company through acquisitions. Therefore, Nadella decided to appoint Scott, who came from LinkedIn, as Microsoft's CTO, responsible for overseeing Microsoft's research department. He also set up a small team for the CTO, allowing them to collaborate with various departments of the company and help guide their technical strategies to counter Microsoft's conservative atmosphere.

According to The Wall Street Journal, Scott quickly realized that although Microsoft had been building large language models for many years, it was falling behind competitors such as Google and Facebook in terms of progress.

So when OpenAI approached, he quickly convinced Nadella to bet on external small companies rather than Microsoft's AI research department. In 2019, Microsoft invested $1 billion in OpenAI, and later increased the investment amount to tens of billions of dollars.

OpenAI CEO Sam Altman later stated that he knew Scott when he was working at Google, and without Scott, he would not have made this deal. He believed that Scott could balance OpenAI's AGI vision with Microsoft's need to incorporate AI technology into profitable products.

For critical computing resources, Scott established a "Capability Committee" to foster competition among different project teams. Every week, representatives from different teams had to present their reasons for needing computing support to Microsoft executives. An employee interviewed by The Wall Street Journal likened these meetings to kindergarten children fighting over toys.

The result of the competition was evident - Microsoft handed over the most advanced GPUs to OpenAI. Other teams within the company had to make do with what was left. Many employees at Microsoft Research Institute were particularly angry, as they found their AI projects sidelined and resources reduced, leading many to leave out of frustration.

It is worth mentioning that AWS, which was collaborating with OpenAI and Microsoft at the time, did not pay much attention to this matter. According to The Information, AWS internally did not regard Microsoft Azure as a real competitor. They believed that Google Cloud was the true competitor to AWS.

However, not long after, Amazon and AWS executives were no longer laughing. The popularity of OpenAI's GPT models far exceeded expectations and attracted a large number of Amazon's customers. Informed sources revealed that from February to August 2023, Intuit, a long-term major customer of AWS and a financial software service provider, increased its monthly spending on Microsoft Azure from thousands of dollars to millions of dollars, mainly to use the GPT models.

Who will dominate the future of cloud computing?

Many analysts believe that the wave of generative AI has the potential to reshape the cloud computing market. Can Microsoft, with the help of OpenAI, replace Amazon's leading position? It is still too early to make a judgment. According to Statista's data, AWS has a market share of about 32% in the field of cloud computing, significantly ahead of Microsoft Azure (22%). Some AWS executives believe that the cloud computing AI service market is just beginning, and AWS has enough time to catch up.

In the first half of the year, Amazon launched its own AI services, Bedrock and Titan, but mainly targeted enterprise customers.

However, according to The Information, AWS's competitiveness in self-developed AI products is very limited. An insider who has tested Titan's large models said that Titan is far behind other AI startups' chatbots, and it is difficult for these self-developed services to attract customers. Therefore, Amazon also announced the opening of the AWS ecosystem, allowing customers to use AI chatbots such as Claude through AWS.

Victor Raymond, CEO of AWS partner Triumph Technology Solutions, believes that although OpenAI is the first company to enter the market in the search revolution, he believes that Amazon's cloud business development model is more outstanding. He believes that the key to the cloud computing market is to allow enterprises to use their own data.

He explained:

"I think the real winning strategy for AWS now and in the future is to leverage its own data to utilize these large language models. OpenAI doesn't give you that opportunity... That's where AWS shines, it can fine-tune what you want to do. It is highly customizable, while some other solutions from our competitors - Azure, Google - are far behind in terms of customization compared to what AWS currently offers."

Amazon firmly believes in this model and has developed a $100 million plan to help customers develop and use generative AI tools on their own data. AWS AI experts also provide free advisory services.