On June 5th, Google CEO Sundar Pichai was interviewed at the Bloomberg Technology Summit. The conversation covered the deep integration of AI with Google's search business, 75 billion USD in capital expenditures, AI's specific impact on the future of work, as well as topics such as corporate power, user trust, and technological boundaries.
Sundar Pichai stated that Google's core strategy is not to start fresh with AI, but to deeply integrate cutting-edge models like Gemini 2.5 Pro into existing products to enhance their capabilities and enable them to handle more complex user needs.
Sundar Pichai also pointed out that the current pace of AI progress is astonishing, and this speed will continue in the future. However, technology itself may have limitations, and we are still far from true general-purpose technology. It's entirely possible that AGI may never be achieved.
01
The Symbiosis of AI and Search: Core Business Restructuring
As chatbots become mainstream information portals, will Google Search be eroded or even replaced?
Pichai believes that this is not a zero-sum, mutually exclusive relationship. He noted that while AI chat applications have reached hundreds of millions of users, Google's search query volume is also growing. He used the example of TikTok and YouTube both achieving good growth during the same period to illustrate that the emergence of new models does not necessarily lead to the demise of old ones. "I firmly believe that search remains exceptionally good at its core function," Pichai said. "People value the utility it provides and demonstrate this by using it more frequently."
The foundation of this confidence is Google's core strategy of deeply embedding AI into its products. Pichai defined the Gemini 2.5 Pro model as a "true breakthrough," tasked with fully injecting this top-tier capability into Google's product line. This is not a simple feature addition, but a deep restructuring of the core business, aiming to transform Google from an information indexer into an AI-driven engine capable of providing comprehensive answers and services.
02
AI Investment and the Value Dynamics of the Open Web
Implementing this AI strategy requires substantial capital support. Pichai revealed that last year, Google spent over 50 billion USD on R&D, making it one of the world's highest R&D investors. Google engages in long-term investments; products like Chrome have seen nearly two decades of investment. For Waymo, investment has also exceeded a decade; in quantum computing, they have also toiled for over ten years. He believes that such large-scale R&D and innovation are only meaningful when focused on long-term development. Additionally, Google is aggressively expanding its AI infrastructure, with current capital expenditures reaching 75 billion USD, compared to 20 billion USD a few years ago. "We are dramatically scaling up our infrastructure."
Pichai also mentioned that new growth areas have emerged, such as Google Cloud's Vertex AI platform, whose token-based usage has grown 40-fold in the past 12 months alone. Furthermore, newly launched subscription services like Google AI Pro and Ultra have received positive market feedback. Pichai stated, "The opportunities before us are greater than any we've ever had."
However, the rise of AI has also intensified external anxieties about the open web ecosystem. On one hand, there's the proliferation of AI-generated content. Pichai views this as an opportunity for Google to leverage its core capabilities: "Our strength is in discernment, 'finding a needle in a haystack' from vast amounts of information." He revealed that Google is using Gemini to help improve YouTube's recommendation algorithms.
On the other hand, there are concerns about the health of the open web. Pichai responded that these doubts have existed for one or two decades, but data shows that in the past two years alone, the number of web pages indexed by Google has increased by 45%. He believes that the barrier to content creation is lowering, and thus the opportunity space for creators is expanding. Among these, the most talked-about is the impact of the AI Overviews feature on content publishers. Regarding publishers calling it "catastrophic" criticism, Pichai said, "Google's value proposition is that users come here, and sometimes they may get direct answers, just as we launched featured snippets many years ago. But people will come back again because humans are curious. They will expand their use cases, and they do indeed seek information sources online. We will continue to prioritize solutions that achieve this goal. With the introduction of AI Overviews, the quality of answers and the contextual information we provide to users are improving. Data shows that AI Overviews are directing users to a more diverse range of websites, and the average click-through time is longer."
03
Real-world Impact from Personal Devices to Enterprise Productivity
When discussing Google's newly released smart glasses prototype, Pichai believes the key is to create products that are "delightful and surprising" and impose no additional burden. He used an example to illustrate the future: a friend wearing AR glasses missed three basketball shots, and the glasses immediately prompted, "That was a terrible three-point shot," to which the friend's first reaction was, "How can I do better?" In Pichai's view, future AI devices will become human "partners" and "coaches."
In terms of enterprise productivity, Pichai confirmed that a significant proportion of Google's code is now AI-generated, citing Microsoft's comparable data of up to 35%. But does this mean a large-scale reduction in engineering teams? Pichai believes that, at least in the short term, the size of engineering teams will continue to grow.
"I simply see AI as a tool that can greatly enhance engineers' productivity," he explained. "It can free engineers from many mundane daily tasks, allowing them to focus on more high-value assignments. This in itself means AI is an accelerator." Because more can be done, the boundaries of opportunity are expanding, and companies will achieve greater things, thus requiring more talent.
Regarding concerns that AI might have a broader impact on social employment, Pichai stated, "These concerns must be taken extremely seriously." He acknowledged that society needs to consider how to help people reskill and build new social safety nets, but he reserved judgment on specific unemployment rate predictions. He cited MIT economist David Autor's research—"60% of today's jobs didn't exist in 1940"—and admitted, "I cannot linearly extrapolate what profound impacts this will bring."
04
AI Governance and the Possibility That AGI May Never Be Achieved
As AI capabilities continue to deepen, issues concerning corporate power, user trust, and information authenticity have become more prominent.
On antitrust issues, Pichai reiterated his stance of disagreeing with the ruling and intending to appeal. He used the success of ChatGPT as an example to show that market competition and user choice still exist. Regarding extreme rectification measures proposed by outsiders, such as "breaking up the company," he considered them "overkill" and emphasized that Google is a company focused on long-term development. He listed: "Last year, we spent over 50 billion USD on R&D. Products like Chrome have had nearly two decades of investment; on Waymo, our investment has exceeded a decade; in quantum computing, we have also toiled for over ten years."
When asked why users should trust Google with more personal information, he replied that trust must be earned through continuous actions, such as responsibly handling user emails for years, protecting data from hackers, and resisting unreasonable censorship demands. He revealed that the biggest current demand from users for Gemini in Gmail is "Why can't it write more like me?", which is precisely what Google is working to meet.
Regarding AI's application in children's products, Pichai drew an analogy to the popularization of online dating, believing that people will eventually adapt to natural interactions with AI, and he has already seen users treating AI as a "partner" for questions like interview preparation. However, he explicitly stated that versions for children, like YouTube Kids, will have "a completely different set of protection mechanisms," strictly limiting functions and experiences to appropriate ranges.
Finally, regarding the "truth" challenge posed by video generation technologies like VEO, Pichai introduced Google's solutions. On the technical front, through built-in SynthID watermark technology, users can detect whether content is generated by Google models using the "About this image" feature. You can upload any image to Google, use the "About this image" feature, and the system will tell you if it was generated by our VEO model. On the governance front, he believes society needs to gradually establish new social norms to punish malicious deepfake content, much like establishing financial fraud regulations.
At the end of the interview, when asked if AGI might never be achieved, Pichai replied, "It's entirely possible." As for whether we are on a clear path to AGI, he stated, "I don't think anyone can give a definitive answer. I can only say that the current pace of progress is astonishing, and looking ahead, I anticipate this speed will continue. But technology itself may have limitations. This technology currently feels like it's progressing rapidly, but in some areas, it cannot even do some obvious things."
He used a straightforward comparison to explain his point of view: teaching a teenager to drive takes only about 20 hours, while Waymo, despite investing vast resources, has yet to fully solve this problem. So, while this technology is admirable, we are still far from true general-purpose technology.
When asked who would lead Google at 50 years old, he replied, "I do believe that whoever is at the helm will have an outstanding AI partner assisting them."
[AI Technology and Application Exchange Group | Invitation Only]
An exclusive circle for TOP-tier professionals in the AI computing power sector.
√ In-depth dialogue with leading computing power companies
√ In-depth dialogue with AI upstream and downstream companies
√ Obtain first-hand global AI and computing power industry information
√ Obtain exclusive information on AI hotspots and cutting-edge industries
√ Stay updated with the latest views and full transcripts of global AI executives
√ Opportunity to participate in AI-themed industry exchange events
Scan QR code to verify identity (please note name/company/position)
More than DeepSeek, it's the future of the AI industry!
| Article Source: Digital Kaiwu
• END •
[Column] Featured Rereads
The largest application scenario for large models has appeared | 2024 Digital Kaiwu Conference successfully held
AI and data centers going overseas to Southeast Asia face two major challenges | Ten-thousand-word roundtable transcript
Full 20,000-word speech transcript: Strongest TPU chip, new AI Agent arrived | Google Cloud Next conference
Jensen Huang's 30,000-word complete refined transcript: "Thinking tokens" erupt, AI infra on the verge of massive change
He Baohong: The first half of large language models has entered "garbage time"
In-depth | "Awakening" Tesla, China's advancing Robotaxi