"The smartest $1 billion investment!" Microsoft made a lot of money

Bill Gates has changed.

Since leaving Microsoft, Gates has been appearing in public as a philanthropist. He is committed to using his wealth to fight poverty, disease and climate change.

As a result, people gradually forget that in the distant 80s and 90s, Gates was actually a grumpy boss.

At the time, Microsoft employees would describe Gates as a "grumpy bully": he liked to curse in meetings, email employees late at night, and call his work "the dumbest code ever."

What's even more "perverted" is that he even goes to the company's parking lot on weekends to see who works overtime at the company on weekends - this is what he admitted in the interview:

I know everyone's license plate, so I can look at the parking lot to see when they come and when they go.

The bad reviews died down after Gates left Microsoft in 2008. Although he still works part-time as a technical consultant at Microsoft, his focus is no longer on Microsoft.

This changed in later years.

At the invitation of Microsoft's current CEO, Satya Nadella, Gates re-engages "more" in Microsoft's affairs. On average, he spends nine hours a quarter in meetings with Microsoft employees, mostly around Microsoft Office, which in recent years has become artificial intelligence.

Sometimes employees make a "pilgrimage" to Gates' elegant wooden private office in Kirkland, Washington, according to people who attended the meeting, and sometimes Gates personally visits Microsoft's headquarters in Redmond.

A former Microsoft manager who met Gates several times said:

The old people are either afraid or shrugged. Newcomers will be very excited about this opportunity.

But over time, it became clear that Gates' temper had eased (although there would still be strength, but it was already rare).

When confronted with something he couldn't understand, he no longer sharply sarcastically or denied it.

For example, he remains skeptical about AI, but Microsoft formed a partnership with OpenAI in 2019 and invested $1 billion in the startup.

According to people interviewed by The Information, Gates was involved in Microsoft's discussions about a partnership with OpenAI, and he was "very engaged" in scrutinizing OpenAI's technology and expressed some skepticism about the startup:

Gates is skeptical of natural language understanding in other contexts because there has been little meaningful progress in the field over the past five years.

Of course, Gates' position at Microsoft is just an advisor, so decisions about investment and development are not part of his purview.

At that time, neither Gates himself, Nadella nor the Microsoft employees involved in this project could have imagined that this investment would become the most colorful in the history of Microsoft and the history of artificial intelligence technology development in the future.

At this year's Microsoft Developer Conference Build, Microsoft announced that almost all of its main products will be embedded in Copilot, integrating generative AI functions, and working with OpenAI to build a ChatGPT plugin ecosystem.

For now, OpenAI seems to be Microsoft's "core competency" in the next decade.

Betting the future of $2.5 trillion on startups that once spent only a billion dollars on is a big bet.

OpenAI: Nonprofit "Broken Dreams"

Going back to Microsoft's acquisition of OpenAI starts with that key shift in the startup.

Most people who have heard of OpenAI know that the original OpenAI was a non-profit organization.

This is also the origin of the name OpenAI: Open, that is, open source.

In 2015, OpenAI wrote in its self-introduction:

Our goal is to advance digital intelligence in ways that are most likely to benefit all of humanity, unconstrained by the need to generate financial returns. Since our research has no financial obligations, we can better focus on the positive impact on humanity.

With this announcement, OpenAI has been faithfully acting as a nonprofit research institution for the next three years, constantly synchronizing the latest research results to the public.

Then, suddenly, in 2019, OpenAI announced an architecture reform and established a for-profit entity on the basis of the original non-profit OpenAI: OpenAI LP.

According to OpenAI, OpenAI LP is a company with a "profit cap" that limits returns to investors to 100 times and is governed by the board of directors of a nonprofit entity (OpenAI Inc.).

In OpenAI's organizational structure, OpenAI Inc. is the operating entity of OpenAI (i.e., a non-profit entity), the board of directors manages the decisions of OpenAI Inc., and also serves as the general manager (GP) of OpenAI LP, which manages the fundraising and use of funds of OpenAI LP.

In addition, OpenAI said OpenAI Inc will retain control.

Although OpenAI's explanation for the shift was that it was "expanding much faster than it was founded," the change quickly drew accusations that it was "going against its mission."

In the forum Hacker News, a post states that OpenAI sets a profit cap of 100 times, meaning that only returns above this amount will go to the original OpenAI nonprofit entity, meaning that unless the company grows to FAANG size, it is "basically unlimited":

Even Google's early investors only made about 20 times their return on capital. Do you bet you will have a company with a rate of return that is orders of magnitude higher than Google?

The commotion is also happening within OpenAI, and some employees are uneasy about these shifts, and to quell the turmoil, OpenAI's leadership has written a FAQ that mentions:

Can I trust OpenAI? The answer is, yes.

According to an exclusive report from MIT Research Review, pressure to commercialize OpenAI is mounting, and when privately sharing his vision for the lab with employees, Altman's message is clear: OpenAI needs to make money to do research.

OpenAI's leadership says [commercialization] is a difficult but necessary trade-off due to the lack of wealthy philanthropic donors.

OpenAI faces this trade-off not only because it's not rich, but also because it makes a strategic choice: trying to reach AGI (General Artificial Intelligence) before anyone else.

OpenAI sees the strategy of massive computation as the fastest path to AGI, which makes it face computing costs and research needs that far exceed the amount of money it can raise as a nonprofit.

It is this pressure that forces OpenAI to make decisions that seem to be moving further and further away from its original intention.

At the same time, Microsoft CEO Satya Nadella has always hoped to catch up with competitors in the field of artificial intelligence, and has also tried to hire some well-known artificial intelligence scientists, but all have failed.

To advance its efforts on language models, Microsoft has also built a huge and expensive cluster of dedicated chips — exactly what OpenAI needs.

In an effort to secure Microsoft's investment, Altman reportedly flew to Seattle several times to negotiate deals with Nadella to show him OpenAI's AI model.

Finally, in 2019, when OpenAI turned into a for-profit institution, it announced the introduction of Microsoft as an investor.

trade

In fact, the 2019 deal initially largely came in the form of OpenAI's free use of Azure, Microsoft's cloud platform.

In exchange, Microsoft became the exclusive cloud provider for OpenAI. Prior to that, OpenAI was one of Google Cloud's top five customers — in 2019 and 2020, OpenAI paid Google more than $120 million in cloud computing fees.

According to a set of internal data from Microsoft, OpenAI spent less than $1 million on Microsoft products and services in the 12 months before Microsoft's investment. In the two and a half years since the investment, OpenAI spent less than $230,000 on Microsoft.

In addition, OpenAI will license some of the technology to Microsoft for commercialization. OpenAI will standardize all of its AI applications on the Azure platform. This puts Azure ahead of Amazon's AWS and Google's Cloud.

Compared to the initial $1 billion investment, Microsoft's upgrades to meet OpenAI's need to use Azure can be said to be numerous.

In early meetings, Altman told Nadella that they needed incredible computing power to power their AI models at scale.

Afterward, Nadella told Azure head Scott Guthrie that Microsoft needed to build AI supercomputing clusters in Azure data centers.

The collaboration allowed Microsoft to tilt its computing power toward OpenAI, and naturally, other customers were affected — Azure's biggest customer at the time was Facebook, which had not yet changed its name to Meta, which was also using Azure to train its AI models.

According to Microsoft insiders, Azure leaders have warned Meta that since the former engineers are still setting up large clusters of graphics processing units for OpenAI, it is expected that the servers needed by Meta will have reliability problems.

(Xiaoza: I thank you)

However, this tilt did not damage Microsoft's relationship with Meta, but first caused anger within Microsoft.

Product leaders within the Office 365 organization reportedly complained to Rajesh Jha, the company's executive vice president of experience and devices, that they had to request and get approval to use a small amount of server space for product development and testing. In contrast, OpenAI can ask for as much compute as needed, and Azure will do it all.

In response, Jha said there was nothing he could do and suggested that their efforts to Open AI be seen as an investment in Office:

This technology will ultimately enhance the results of your work.

Investments in outside companies have also raised suspicions about Microsoft's own internal AI research department — OpenAI has just over 200 people, but Microsoft AI Research has 1,500 people and has been around for more than a decade.

This disparity led Nadella to "soul torture" of Peter Lee of the AI Institute last year:

OpenAI built GPT-4 with 250 people. So why do we have Microsoft Research?

Of course, the efforts of Microsoft AI Research are not without results, and OpenAI's research is based on the Transformer architecture developed by Microsoft.

But to Lee's dismay, OpenAI is investing less and achieving more than Microsoft Research.

However, after being "PUA," Lee received a private call from Nadella thanking him for his work on understanding and implementing OpenAI technology, and letting the institute act as Microsoft's "gatekeeper" to OpenAI, telling him "whether the emperor was wearing clothes or not."

As an established technology company, Microsoft has put aside its "pride".

The "smartest ever" $1 billion

Meanwhile, investment in OpenAI continues.

In early 2023, Microsoft invested another $10 billion in OpenAI. According to Fortune magazine, in 2021, Microsoft quietly invested another $2 billion in OpenAI.

According to a person familiar with the terms to The Information, after OpenAI repays its first investors, Microsoft will receive 75% of its profits until its principal investment is repaid, and then 49% of profits until it reaches the theoretical upper limit of 100x.

At that time, Microsoft's shares in OpenAI will be returned to OpenAI's non-profit entity.

But Microsoft gets much more than the financial benefits of these investments.

With Microsoft's support, OpenAI built large-scale generative models such as GPT-4 and DALL-E, and completed a business model accordingly:

OpenAI provides API access for businesses that want to build applications on top of it, can be customized with proprietary data and additional AI capabilities, and can also plug models directly into the product, and OpenAI has launched a paid version of ChatGPT Plus.

Microsoft, on the other hand, could also commercialize products that embed AI technology through a partnership with OpenAI.

This was evident in the rise of New Bing, the company's first real attempt to undermine Google's search dominance since its founding.

In 2021, the search business generated $150 billion in revenue for Google, while Google's share of the global search market was about 90%, and Bing accounted for only 3%.

According to Insider's estimates, if ChatGPT were integrated into Bing, Microsoft's market share in this area would increase to 10% — and Microsoft CFO Amy Hood once told analysts that one percentage point of market share in the search market is worth $2 billion.

This means that just by gradually approaching Google, Microsoft can reap billions of dollars in revenue.

Investment bank D.A. Davidson also noted in a report:

In the long run, we think integrating ChatGPT functionality into Bing could provide Microsoft with a once-in-a-decade opportunity to replace Google's search dominance.

In addition to these, there is an even more critical, perhaps the most critical: Azure.

As we mentioned earlier, Microsoft had to upgrade its Azure data center to meet the training needs of OpenAI. In turn, the success of OpenAI has cemented Azure's dominance in cloud services.

As Eric Boyd, vice president of Microsoft AI Research, puts it, meeting the need to train and run LLMs for OpenAI drives innovation that benefits all Azure customers — and it's easier to train and run large AI models on Azure in the future.

This is Microsoft's advantage and ambition: the number one infrastructure service provider in the AI era.

In fact, cloud service providers such as Microsoft and Amazon have enjoyed high profit margins over the past few years due to a lack of competition and huge infrastructure.

But in the next decade, as artificial intelligence sweeps the world, the entire industry's cooperation with cloud service providers will also be based on Microsoft and OpenAI.

By then, Microsoft's AI Azure supercomputing cluster could become Microsoft's most valuable infrastructure for the next decade.

On the other hand, it is precisely under the impetus of OpenAI that an era of artificial intelligence that could mean the fourth industrial revolution is accelerating.

And it all started with a billion dollars.

This article is from Wall Street News, welcome to download the APP to see more