Microsoft Rises Again, Google Begins to Decline

Let's talk about the different development trajectories of two of the world's top companies - one is Microsoft's resurgence, and the other is Google's decline.

First, let's talk about the resurgence of Microsoft.

On January 24, 2024, Microsoft's stock price exceeded $405, with a market value reaching $3 trillion. This number also surpassed last year's GDP of France at $2.8 trillion. Saying Microsoft is rich enough to rival countries is not an exaggeration, as it even exceeded the GDP of a current major developed country. Soon after, Apple's market value also surpassed $3 trillion. Microsoft and Apple became the only two companies in the history of the global financial market with a market value over $3 trillion.

Since its founding in 1975, Microsoft has been growing rapidly. The successive generations of Windows and Office completed under Bill Gates' leadership are Microsoft's core business. The growth inertia comes not only from the twenty-year PC era's dividends but also from its own excellent product design. The company's main products peaked with Windows XP in 2001. This version of Windows had the longest lifespan, from its release in 2001 until support ended in 2019.

After 2001, Microsoft missed out on many era dividends, such as the search engine and mobile internet dividends. Several giants quickly rose in these fields, such as Google, Amazon, and Apple. However, because Microsoft's operating system and office software have always been in a monopoly position, Microsoft still maintained its foundation and did not collapse.

If Microsoft continued to develop at this pace, the continuous depletion of innovation ability would eventually drag it down. Since 2011, Microsoft has adjusted its direction, making cloud computing a future focus, and truly developed it into a main product that surpassed traditional software business in ten years.

Moreover, the growth of this business has slowly brought Microsoft and the emerging AI closer together. In July 2019, they established an exclusive cooperation with OpenAI, providing massive computing power as an investment for the newly restructured OpenAI. Afterwards, Microsoft integrated OpenAI's technology into Microsoft Azure cloud services and developed its own series of AI products, including Azure AI's range of speech recognition and generation, visual image analysis, and translation API services.

In addition, there are a series of Copilot assistants, including GitHub Copilot programming assistant and Office365 office software's Copilot assistant, among others. The release of a series of products by OpenAI in 2023 also prompted Microsoft's stock price to begin growing rapidly for the first time in 14 years. The stock price rose by about 60% in 2023. Moreover, Microsoft still has an ace up its sleeve that has not been used yet, which is Windows 12.

For a long time, while cloud services were the main business, Microsoft's innovation in Windows was severely lacking. Many users commented that the innovation of Win8 was mainly hiding the "Start" menu that everyone had been used to for decades; the innovation of Win10 was restoring the hidden "Start" menu; the improvement of Win11 was requiring motherboards to support a strange thing called TPM2.0, which means your computer could not install Win11 unless it was bought within the last year. Setting aside the innovation of new versions, even system updates would occasionally bug out, leading to blue screens and startup failures.

Just when many thought Microsoft would continue to maintain Windows with a just-enough effort attitude, the killer feature, Win12, is coming, expected to be released in 2025.

It is called a killer feature because it deeply integrates the operating system with AI, planning to incorporate all the currently useful AI functions into the operating system. For example, the previously almost unused Paint will be combined with DALL-E to automatically generate images; a dedicated AI key on the keyboard to call out the Copilot assistant; and native support for large models in the operating system. If previously, AI applications were only of interest to a minority of tech enthusiasts or those with a strong ability to adopt new things, when it is embedded into the operating system, AI's popularity will increase by at least an order of magnitude.

However, the upgrade to Win12 is also likely to be the most demanding of all Windows versions, even more demanding than the TPM2.0 required by Win11.

Firstly, the minimum memory requirement is 16GB. By comparison, Win10's minimum memory requirement was 2GB, and Win11 only required 4GB. If 16GB is the minimum threshold, then 64GB may be the minimum for a smooth experience with Win12, and 128GB might be the minimum capacity for handling slightly complex tasks smoothly.

You might think this is not too big a problem, as you are already using 32GB of memory today, and by 2025, 64GB-256GB of memory should be a mainstream range. But there's a minimum requirement even more demanding than memory, which is that the processor used with Win12 must have a dedicated AI processing unit, and the processing power must be greater than 40 TOPS.

Currently, among the two monopolies of processors, Intel's newly released Core Ultra has a separate AI processing unit but only a processing capability of 34 TOPS; AMD's desktop ZEN4 processor also has an AI processing unit, but its processing capability is only 16 TOPS. The only one that truly meets the requirements is the Qualcomm Snapdragon X ELITE, set to be released in June 2024, with 45 TOPS, which is just slightly above the minimum threshold of 40 TOPS.

Therefore, if Microsoft wants to dominate the next era's mainstream growth point—AI—through the operating system, it would have to start from the next generation of processors, with the earliest point being around 2026. Before this, various AI startups still have many opportunities to make their products good enough so that users lose the incentive to switch to Microsoft's AI suite before 2026.

This scenario is not impossible. A similar situation once happened in the web browser market. But if this 2-year window is not seized, the AI market will be all Microsoft's in the future.

Overall, Microsoft first bet correctly on cloud services, then naturally took the fast train to AI, and its future development is also long-term positive.

On the other hand, looking at Google's development, the outlook is bleak.

The earliest signs noticed by the outside world were Google's two AI product launches in early and late 2023. The first was hastily organized in Paris, a location far from the world's tech center, on February 8, just two months after OpenAI's ChatGPT went viral. It seemed like a last-minute arrangement in a café, clearing chairs for the event. Most of the audience could only watch through an online live stream.

The only on-site demonstration was answering a question for a 9-year-old child about the latest discoveries by the James Webb Space Telescope. The picture provided by AI was pointed out by a professor from Harvard University's Center for Astrophysics on the same day as not being taken by the James Webb Telescope, but by a European ground-based telescope.

Although the AI chat tool Bard launched later was not entirely unusable, Google, which used to be at the level of Shaolin and Wudang in the AI world, showed a level that was only equivalent to a lesser-known sect.

Unsatisfied, Google launched a similar product called Gemini 10 months later. This time, it was only a PPT presentation, with many achievements claimed to be better than GPT-4, but a closer look at the report reveals that Gemini's victory came through unfair testing methods. If this were a startup company's product, it might have secured tens of millions of dollars in investment based on this PPT. But this is a product made by Google, aimed at making a strong comeback, and it only reached the level of a first-tier follower, which is somewhat embarrassing.

In the past decade, Google has created an unmatched image in the AI field through a series of breakthroughs and excellent products. Today, all commercially successful AI technologies in the market can be traced back to Google's R&D team—either created by companies founded by former Google employees or built using Google's open technologies. While having Google Brain internally, Google further expanded its AI advantage by acquiring DeepMind for $500 million in 2014. However, the situation of two AI product launches within a year was quite dismal.

Many employees who left Google have leaked information, revealing Google's big company disease:

With 170,000 employees, management has become increasingly cumbersome, and a new project can no longer be driven by passion and technological breakthroughs. The decisive factor lies in approval processes, legal reviews, performance evaluations, and meetings at various levels. Employees' motivation to join Google has shifted to job promotions, bonuses, nearly free delicious food, and a semi-casual work environment. A senior manager in a team has to make plans six times a year, each taking one month.

In addition, there is a month off for holidays, a month for performance evaluations, leaving only three months a year to do actual work. It's easy to imagine how difficult it is to push forward any new projects. But it's no problem not to profit from new projects, as Google's bonuses are still distributed, covered by the ever-growing advertising revenue.

Therefore, Google has launched quite a few failed products in recent years. For example, Google Glass, Google Play Music, Google Spaces, Google+, Google Phone, Google Fiber, space elevators, and autonomous cars, all ended without success. Google's heavily invested DeepMind has also never had the ability to profit independently.

After falling far behind Microsoft and OpenAI in the AI era, knowledgeable people inside Google began to leave and start their own businesses. The Transformer large language model was published by 8 Google scientists in 2018, and just a few years later, all 8 people left Google. At the beginning of 2023, Google carried out the largest layoff in its history—12,000 people.

In January 2024, Google was again brewing a layoff plan. By the end of January, it had already laid off more than 1,000 people across various teams, including advertising sales, hardware, search, shopping, maps, and YouTube. It is said that in 2024, the number of layoffs will be roughly comparable to that of 2023. Nowadays, Google's headquarters occasionally sees employee protests.

Recently, Google's senior software engineer Diane Hirsh Theriault posted a lengthy message bluntly pointing out Google's current problems. She has been working at Google for 8 years, witnessing both its most glorious times and its current issues, and crucially, she is still an employee of Google.

She said that the team she led was cut by three-quarters last year, and Google's layoffs seemed random. Over the years, Google has not launched a single successful product. Vice Presidents attempt to issue orders, saying we need a new Demo, and then the matter is dropped. If a new product doesn't gain 100 million users within six months, they shut down the project.

All middle managers are guessing what might appeal to the Vice Presidents, using this method to protect their teams. If you are a team leader and guess wrong, both you and your team will be laid off. Many people continue to stay only because they are unwilling to give up their salaries, and no one works late anymore. By 4:30 PM, the office is half empty. Those who were once willing to work extra hours in the evenings and weekends are gone. That atmosphere has disappeared.

Indeed, this problem is not unique to Google. The technology giants that rose with the internet in the 2000s now all face issues with stagnant growth, creativity stifled by various obstacles, and the most valuable, motivated young people leaving to start their own ventures.

Microsoft managed to emerge from the mire earlier because it made the right bet on AI with OpenAI, while Google, Amazon, Apple, and Meta all have to face this round of tests for a new blood.

Alright, that's the content for today. We'll see you tomorrow.