Alphabet CEO Pichai: Google AI helps search upgrade, quantum computers and robots are the key to the explosion

As generative AIs such as ChatGPT, Meta AI, and MuskxAI sweep the world, the outside world is also questioning whether Google, the leader of the search advertising empire, will be replaced in this wave of AI revolution? Not only did it suffer a significant loss in advertising revenue, but it was even so consumed by internal friction that it dared not touch its most profitable business.

In this regard, Alphabet CEO Sundar Pichai was interviewed by the well-known technology channel "All-In Podcast" on May 16. He generously stated that AI will allow Google search to "evolve" and that the company's transformation began as early as ten years ago. The interview also discussed Google's future blueprint for quantum computers, robots and AR. From the entire conversation, we can understand that Google will become an important driving force for future AI innovation infrastructure.

AI helps Google search upgrade, Pichai says Google's transformation started 10 years ago

Pichai was the first to say that since he took over as CEO of Google in 2015, he has clearly focused on "AI First".

It had previously launched an AI research project (Google Brain) in 2012 and acquired DeepMind in 2014, all to lay the foundation for AI transformation.

Pichai emphasized: "Search will always be at the core of Google, and AI will upgrade our search."

Google restructures search mode, AI Overviews and AI Mode are officially launched

Pichai also highlighted two major upgrades:

• AI Overviews: Helps users integrate information from multiple websites into a key summary. It has now expanded to more than 150 countries and 1.5 billion users.

• AI Mode: Make Google search like chatting with an AI Agent.

He pointed out that these new features not only did not reduce usage, but instead brought new traffic and novel search categories, becoming one of the important foundations of transformation.

Will AI make Google search unprofitable? Pichai says its own TPU reduces costs and advertising revenue is stable

In response to concerns that generative AI responses would stop users from clicking on ads, thereby impacting $200 billion in search ad revenue, Pichai said:

• At present, AI search advertising revenue is on par with traditional search

• Google's self-built TPU chip has significantly reduced the cost per AI response, eliminating the need to worry about cost issues.

• Optimizing latency is the key

Gemini App has fewer users than ChatGPT, but Overviews has been integrated into its own products

Although Google's Gemini App is not as popular as ChatGPT, Pichai pointed out

“AI Overviews on Google Search is actually the most commonly used AI product in the world.”

He emphasized that Google's strategy is not to focus on a single application, but to deeply integrate AI into all products such as search, YouTube, cloud, Google Docs, etc. This is the selling point.

Where does Google's cost advantage come from? It turns out it comes from 20 years of infrastructure advantages

Pichai also talked about Google's AI cost advantage, which comes from years of accumulated infrastructure investment:

• The self-developed chip TPU has evolved to the seventh generation. The latest generation "Ironwood" can achieve 4 × 10²⁰ floating-point operations.

• Self-built data center, power grid, submarine optical cable, and AI ecosystem

• Capital expenditures reach $75 billion in 2025, half of which will be used to expand Google Cloud

Is AI model innovation stuck? Pichai says he hasn't seen the end of the road yet

The outside world is worried about the bottleneck in the development of large language models (LLMs), but Pichai has a different view:

“We are accelerating in the direction of AI agents, multimodality, and visual-language fusion. Whenever you think you are stuck, there will suddenly be a new breakthrough.”

He also emphasized that Google not only does LLM, but also goes deep into:

• Image Generation (Diffusion)
• Vision AI
• Robotics AI

Good cooperation relationship with Nvidia, dual-track development of TPU and GPU

Although Google is vigorously developing TPU, Pichai does not deny:

• Gemini training mainly relies on TPU, but deployment and commercial use still use a large number of Nvidia GPUs

• I respect Jensen Huang very much and we have a good working relationship.

In the future, we will continue to develop TPU and GPU on a dual track to maintain flexibility and supply chain advantages.

Admiring China's DeepSeek, the world re-evaluates China's AI strength

He pointed out that Google's internal model Flash was no less powerful than DeepSeek, but when DeepSeek appeared, even Google began to re-evaluate China's AI capabilities:

“It’s really impressive that they were able to create such an AI model under the constraints of hardware.”

Energy has become a bottleneck for development, the United States must speed up its efforts to solve it

Pichai admitted that Google Cloud is currently facing a power bottleneck due to the simultaneous expansion of global data centers:

• Project delays

• Not enough technicians, shortage of power workers

• Nuclear energy, solar energy, and power grid upgrades are urgent

Quantum computers, robots and AR are the key to Google's next wave of success

Pichai also revealed that Google is actively investing in the following areas:

• Quantum computers: The current technology is like AI in 2015, and disruptive breakthroughs are expected in the next 3 to 5 years.

• Robotics: The Gemini team has begun developing AI models that can integrate vision, language, and action.

• AR glasses: will be the ultimate form of human-computer interaction in the future, allowing AI to appear in front of you at any time and actively assist you with daily life.

In the next 10 years, Google will become an important promoter of global innovation infrastructure

Pichai said that AI is a wave bigger than all previous technologies, and the key to success is not who defeats whom, but who can continue to innovate and integrate AI into every corner of global life.

He also said that in the future, Google will not only be an advertising search company, but also an important driving force behind global innovation infrastructure.

The Google I/O conference was held on May 21, during which CEO Sundar Pichai also took the opportunity to announce several major upgrades to the company's AI. Not only did it launch the Ironwood TPU, which is 10 times faster than the previous generation, and the Google Beam, which focuses on 3D immersive calling experience, but also the Gemini App's agent mode, which can directly help with booking, viewing, arranging itineraries and other novel products and functions. It can be seen that Google really wants to move towards creating a "universal AI assistant" and integrate it into the daily lives of all mankind.

Ironwood TPU is now available, 10 times faster than previous generation

Pichai first launched his seventh-generation TPU "Ironwood", featuring:

• 10 times faster performance than previous generation
• A full set of TPU Pods can perform 42.5 teraflops of computation per second.

It will be available to Google Cloud users by the end of the year.

Google Beam, an AI-powered 3D video device, is now available

Featured AI-driven 3D video devices:

• Consists of 6 lenses
• After shooting, 3D light field images can be synthesized
• To make remote video chat like face-to-face chat

The first batch of devices were co-developed by HP and will be available to the first wave of users this year.

Real-time translation and screen sharing are now available, marking the great evolution of Gemini

As a Gemini Live AI model actively built by Google, its functions have been greatly evolved:

• Real-time voice translation : Currently supports English and Spanish, other languages ​​will be launched in due course

• Supports screen sharing and image analysis : It can analyze the user's screen in real time. If you mistake a street lamp for a person and follow you, Gemini will reply "That's your shadow!"

• Available for Android and iOS users from 5/21

AI multitasking agent Project Mariner launched, available to developers via Gemini API

Pichai also announced that Google will soon open up Project Mariner, a multitasking agent that will:

• Process 10 tasks at a time

• Learn and replicate task flows

• Available to developers through the Gemini API

Gemini App's new Agent Mode can help you find a room and arrange your trip at the same time

As Google's own AI signature application, Gemini App has evolved its functions quite significantly this time:

• Add AI agent mode to automatically search for properties and arrange house viewings
• We can also help you make phone calls and book itineraries.

The Gemini App has added an "agent mode" function that will automatically search for listings on platforms such as Zillow, arrange house viewings, and even help you make phone calls and book trips, and supports MCP integration with other services.

MCP helps Gemini connect to various websites, apps, and service systems, upgrading it from being just a talker to an agent who can help with work.

Gmail starts to import Gemini to automatically reply to users' emails

Gmail, a typical email service, has also started to import Gemini, including:

• Gemini reads the user's past writing styles, documents and calendars
• Automatically generate reply content
• Available to subscribers on Gmail this summer

Gemini Flash and 2.5 Pro are upgraded, AI program assistant Jules helps to modify the code

Gemini's new Flash model is also faster and more powerful than before, not only:

• Introducing the 2.5 Pro "Deep Think" mode, which can handle difficult math problems and long tasks
• Officially launched in June
• Supports 24 languages, natural intonation conversion, bilingual mode, all integrated into Gemini API
• After taking a screenshot of the code, feed it to 2.5 Pro, and Jules, the developer-specific AI assistant, will help you modify the code. The public beta will be open on 5/21.

New AI models for music and video are released one after another, and AI video creation platform Project Flow is launched

• Imagine 4 : A new generation of AI image generation model, more accurate text processing, 10 times faster generation speed, and can also handle typesetting

• Veo 3 : A new video generation model that can integrate narration and ambient sound

• Lyria 2 : Generates AI music models to produce high-quality music

• Project Flow: A new AI video creation platform that allows users to freely generate or upload characters and scenes, and then use text commands to let AI create the screen.

Fully integrated with Chrome, Wear, and TV, allowing Search AI to truly assist

Search AI evolves: “AI mode” becomes a true assistant

• AI Mode : You can ask complex questions and answer them using charts, tables and summary reports

• Search Live : Search and interact like making a video call

• Try-on function : upload photos to simulate and compare clothes trying-on conditions

• One-click checkout : price change reminder, add to cart, automatic checkout, all done

• Gemini in Chrome : can directly read the page content to give the answer

• Deep Research + Canvas : Upload your report and turn it into a webpage, podcast, or quiz

• Gemini Live is integrating with Keep, Maps, and Calendar

Gemini enters XR field and creates AI glasses and head-mounted displays with Samsung

Google is also working with Samsung to create XR smart glasses Project Muhan, which is expected to be launched this year.

Project Muhan supports voice, visual search, translation, navigation, instant response and other functions, and cooperates with Warby Parker to develop styles.

Ultra Subscription Plan and Global Layout

• Google AI Pro / Ultra : Pro offers a higher usage limit, while Ultra offers early access to new features, YouTube Premium, and large cloud storage.

• 2.5 Pro Deep Think, Veo 3, Flow and other features will be given priority to Ultra subscribers

As CEO Pichai said at the end, Gemini is evolving from a multimodal model into an "AI world model," and Google's vision is to create a truly "universal AI agent" that can accompany humans in writing letters, solving problems, editing videos, dressing up, auditioning, and even walking to find a coffee shop, fully integrating into human daily life.