Microsoft's ambition goes beyond "AI assistants"! Github Copilot becomes a chip to compete for cloud computing customers.

Microsoft, among the seven giants of the US stock market, is widely recognized as the leading technology company in AI applications, being the first to utilize AI to drive business growth.

A typical example is the code storage platform GitHub. Leveraging OpenAI's technology, Microsoft's GitHub platform has integrated the popular AI programming assistant GitHub Copilot.

However, Microsoft's ambition goes beyond creating a popular product. According to reports, Microsoft is actively promoting the use of GitHub Copilot among its customers and encouraging them to rent Azure cloud servers. This business can generate much more revenue compared to a standalone product.

GitHub Copilot doesn't make money, but Azure does

In 2018, Microsoft acquired GitHub for $7.5 billion. Since the acquisition, there has been an expectation that GitHub would bring in more revenue for Azure.

However, Microsoft has been cautious about monetizing GitHub and has not previously required companies using GitHub to create Azure accounts.

Now, thanks to the excellent product capabilities of GitHub Copilot, Microsoft is finally taking a bold step to promote Azure to GitHub customers.

According to a survey conducted by software company JetBrains in July 2023, among over 30,000 software developers, 42% of respondents stated that they use GitHub for code review, up from 36% a year ago.

Swami Chandrasekaran, Head of Artificial Intelligence and Digital Innovation at KPMG, stated to the media that last year, they had 500 developers from the tax services department try out GitHub Copilot, resulting in an average 10% increase in coding efficiency, equivalent to saving approximately 180 hours per developer per year. KPMG plans to expand the use of Copilot to other departments.

However, despite the impressive performance of GitHub Copilot, it does not actually generate profits. GitHub Copilot charges $10 per person per month, but there are reports that this feature causes a loss of over $20 per user per month (some users even incur losses exceeding $80 per month).

In other words, for every $10 in revenue generated by GitHub Copilot, it incurs a loss of $20.

The root cause of the loss lies in the high costs of developing and operating AI models. Unlike conventional software development and operation, which benefit from economies of scale where costs decrease as the number of users increases, AI models may require intensive new calculations for each user's demand. However, Microsoft has its own ways of making money.

According to some Microsoft salespeople, as more and more companies migrate their code to GitHub to use GitHub Copilot, Microsoft has started actively promoting security tools and other advanced services that require an Azure account to customers. They believe that Copilot has become an important driving force for Azure promotion.

For example, insiders have revealed to the media that Goldman Sachs has been using a mix of code storage platforms such as GitHub and GitLab for a long time. However, recently, Goldman Sachs has significantly increased its spending on GitHub and has purchased Copilot assistants for over 10,000 software developers, with an annual subscription fee of approximately $2 million.

Insiders also revealed that in the second half of last year, Goldman Sachs' spending on Azure increased by more than 20%, and it is expected to exceed $10 million annually on Azure.

Considering that Amazon AWS is currently Goldman Sachs' main cloud service provider, Goldman Sachs' practice of voting with money undoubtedly confirms that Microsoft is successfully encroaching on the territory of its biggest competitor.

According to research firm IDC, in the first half of 2023, Azure's share of global cloud server sales will reach 20%, reaching $24 billion, higher than 19% in 2022. AWS's market share has declined from 33.6% to 31.5% during the same period. Google Cloud's market share remains relatively stable at 6.6%. Another research firm, Canalys, also reported similar market share trends in the first half of 2023.

Google and Amazon Join Forces to Counter Microsoft

Faced with Microsoft's strong offensive, Google and Amazon, the two major competitors, naturally cannot sit idly by.

Although the luck of these two companies is not as good as Microsoft's, as they did not invest in unicorns like OpenAI. However, Google and Amazon are intensifying their investments in competitors of the Microsoft-OpenAI alliance, such as another major code storage platform, GitLab.

As the second-largest code storage platform after GitHub, GitLab's revenue growth has slowed from 75% as of April 30, 2022, to 32% as of October 31, 2023, due to the downturn in the enterprise software market over the past two years. Analysts estimate that GitHub's revenue is roughly twice that of GitLab.

Although it is smaller in scale, it still has competitiveness. In order to cope with GitHub's rapid progress, Google and Amazon are increasing their support for GitLab. Executives from both companies have revealed to the media that Google is providing its own AI model support for some of GitLab's new code generation tools, and GitLab is in early negotiations with Amazon to make it easier for software developers to use both companies' AI products simultaneously. Some analysts believe that GitLab may be an attractive acquisition target for cloud service providers, as it is a handy tool to counter Microsoft GitHub. The most likely candidate is Google, which owns about 2.5% of GitLab. According to insiders familiar with GitLab's business, Google Cloud is now one of the most important cloud service providers for GitLab.

Ashley Kramer, Chief Marketing and Strategy Officer of GitLab, said that the company's own AI code assistant is expected to be adopted rapidly, and another tool to track the efficiency of different coding teams within the company will also be launched.

She also mentioned that some of GitLab's customers have switched from GitHub because they found that the latter's software running and testing platforms are inefficient.