As 2025 nears its end, Google emerges as one of the top-performing major tech firms in Silicon Valley. Its stock price has seen a cumulative surge of 63%, reaching a peak increase of 70%. This performance significantly outpaces the S&P 500 and Nasdaq indices. Google also came tantalizingly close to hitting the $4 trillion market cap milestone, nearly joining Microsoft, Nvidia, and Apple as the fourth company in human history to break through this barrier.

The past year and a half has been an extraordinary success story of Google's "resurgence from the brink," a tale worthy of being adapted into a film. By May 2024, amidst the rise of generative AI, Google seemed to be teetering on the edge of collapse. The Gemini large model underperformed, TPU had no external customers, and the TensorFlow platform was completely supplanted by PyTorch. There were serious doubts about whether large models would replace search engines as the primary means for users to obtain information in the next decade. Although Google's core advertising business remained robust, Wall Street viewed it as a temporary reprieve. The key issue was Google's low organizational effectiveness in AI, failing to respond effectively in both technology and product development, possibly hinting at flaws in its decision-making mechanisms and corporate culture.

Then came Google's stunning counterattack. The release of Gemini 1.5 Flash in May 2024 marked Google's catch-up in foundational large models. The launch of Gemini 2.0 in January 2025 solidified Google's position among the leaders in large model technology. The recently released Gemini 3.0 proved that Google's large model technology could now rival that of OpenAI. Gemini 3.0, entirely trained on TPUs, further piqued external customers' interest in purchasing TPUs. By 2026 - 2027, TPUs are expected to make significant inroads into the external market. While Nvidia will remain the leader in the AI chip market, the second place will likely go to Google, not AMD.

More importantly, Google is rapidly "rebuilding" its entire core business with generative AI. Hundreds of millions of users now use the "AI mode" of the search engine daily. The Android operating system has deeply integrated with the Gemini chatbot, giving it an edge over iOS. New generative AI-based commercialization models, such as AI shopping guides, are under development. The accuracy of ad targeting is slowly but steadily improving. Google's released AI creation tools have had a positive impact on YouTube creators.

If anyone doubts that AI is a bubble or incapable of changing the work and lives of most users, they should take a closer look at Google. This isn't a daunting task. If someone is too lazy to study Google or doesn't know how to, they might not be well-suited to study the internet industry (or any tech-related field).

On the surface, OpenAI still holds the initiative in generative AI. As of the fourth quarter of 2025, ChatGPT has around 800 million weekly active users, while Gemini has only 650 million monthly active users. However, in the past year, Gemini's user growth has far outpaced that of ChatGPT. Moreover, over 2 billion Google search engine users can access Gemini's features through "AI summaries." If Gemini's user base surpasses ChatGPT's at some point in 2026, it wouldn't be surprising. I believe that in terms of enterprise API calls, Gemini is also rapidly catching up.

After ChatGPT's sudden emergence in late 2022, Google issued a "red alert" across the company, acknowledging that its core business was under serious threat. Three years later, with the release of Gemini 3.0, it was OpenAI's turn to sound the "red alert." Sam Altman later claimed that the impact of Gemini 3.0 on GPT was not as significant as imagined and that the "red alert" would be lifted by early 2026. Regardless, from being severely threatened by OpenAI to posing a serious threat to OpenAI, Google has largely succeeded in protecting its core territory over the past three years.

How did all this happen? Why did the seemingly shaky Google empire in the first quarter of 2024 become so vibrant by the fourth quarter of 2025? I believe the logic is straightforward—there's no magic involved.

Google has deep accumulations in AI technology (not limited to generative AI). However, before 2023, it didn't realize the importance of generative AI and underinvested in this area.

Google's organizational scale was too large, leading to certain management issues. For example, the two AI R&D institutions, Google Brain and DeepMind, appeared to be in internal competition but were actually chaotic. After their integration in 2023, it took a long time to get back on track.

However, after recognizing the importance of generative AI and seeing the path validated by OpenAI, Google's technological accumulations and resource advantages were fully unleashed. As long as it didn't continue to make mistakes, catching up was only a matter of time.

After the severe turmoil at OpenAI in late 2023, the loss of core R&D personnel and the obvious slowdown in technological progress meant that OpenAI was no longer a vibrant tech startup, creating external conditions for Google's pursuit.

Sergey Brin's return as co-founder and leadership of Google AI research may have played a significant role, especially in making up for Sundar Pichai's shortcomings. However, we don't have concrete evidence on this point.

In 2025, the capital markets realized that Google is the only company globally with so-called "full-stack AI capabilities" (Note: Some argue that Alibaba also qualifies, but we won't evaluate that here). From AI computing power at the top of the supply chain to cloud computing and development environments, to foundational large models, and finally to consumer-facing applications, Google not only possesses all these capabilities but also ranks first or at least in the top three in most segments. If Google has any obvious weak link in the AI value chain, I believe it's open-source large models. Google abandoned the open-source route several years ago and hasn't truly returned to it, leaving the open-source community largely dominated by Meta, xAI, Alibaba, and DeepSeek's large models.

Now is the window of opportunity to fill this gap. Meta's LLaMA-4 has already disappointed, and if it can't release a significantly improved new version next year, Meta will almost be out of the foundational large model race. xAI, being a startup, may see its embrace of the open-source community as a temporary measure, and its technology is not yet stable. If Google truly intends to dominate the open-source arena again, the cost may not be high. Of course, it may not have this intention, as promoting open-source technology has both advantages and disadvantages.

I would further argue that Google's existence proves that any so-called "AI bubble," if it exists, is limited. As a company that consistently delivers double-digit revenue and profit growth each quarter, with abundant cash flow and vast imagination space in AI, Google's current forward P/E ratio is only 29. Pessimists may claim that there is a huge bubble among the "Neocloud" platforms and AI application companies; they may even argue that there is a massive bubble in the computing power value chain (including but not limited to Nvidia and Broadcom). However, by any standard, Google probably doesn't have any substantial bubble.

Of course, if the capital markets turn fully pessimistic, Google will undoubtedly take a hit, as will all major Silicon Valley tech companies. Some will be exposed as bubbles, while others will shed their bubble layers and shrink significantly. Companies like Google, however, will prove that their businesses remain unaffected and can deliver on their potential in any market environment. It's precisely such companies that give me confidence in the long-term prospects of generative AI. When the tide recedes, some will be found swimming naked; those who aren't will swim even better when the next tide arrives.

This article has not received any funding or endorsement from Google or its competitors.