From: allin
Google, now Alphabet, is positioned as a deep computer science company that applies fundamental research to build products impacting daily life [02:21:55]. This approach has driven significant financial growth, with revenue growing from 100 billion per quarter, and the stock increasing 4.5 times to a $2 trillion market cap under CEO Sundar Pichai’s tenure [01:53:05].
AI-First Philosophy and Core Research
Google adopted an “AI-first” approach almost a decade ago [04:11:00]. Key foundational investments include:
- Establishing Google Brain in 2012 [04:18:00].
- Acquiring DeepMind in 2014 [04:20:00].
- Pichai declared Google “AI first” upon becoming CEO in 2015, believing AI would drive the biggest progress in search and beyond [04:23:00].
Evolution of Search with AI
Despite concerns about AI’s potential disruption to its core search business, Google views AI as an “extraordinary opportunity” for search [04:42:00]. The company aims to make information more accessible to everyone [04:50:00].
Key AI advancements in search include:
- Transformers (BERT and MUM): These models significantly improved search quality over the last couple of years [05:09:00].
- AI Overviews: Launched about a year ago, these overviews are used by over 1.5 billion users in more than 150 countries [05:16:00]. They expand the types of queries people can use and have led to empirical query growth [05:28:00].
- AI Mode: A new dedicated AI experience coming to search, announced at Google I/O [05:51:00]. This mode allows for full-on AI experiences, including follow-on conversational queries, using Google’s cutting-edge models [05:59:00]. Users are typing much longer, paragraph-length queries, two to three times the average length from two years ago [06:18:00].
Google’s strategy is to “follow the user” [08:06:00], adapting to new formats and user experiences even if monetization follows later. This approach was evident with YouTube Shorts, which initially didn’t monetize as well as long-form content but was prioritized for user experience [07:43:00].
Infrastructure as a Core Advantage
Google’s long-standing investment in infrastructure is a significant competitive advantage [15:56:00]. This includes shipping container data centers from 20 years ago and continued investments [15:37:00].
TPUs and Custom Chips
Google literally operates on the “Pareto frontier” of performance and cost, delivering the best models at the most cost-effective price points [16:40:00]. This is partly due to training and serving models on its own infrastructure, including Tensor Processing Units (TPUs) [17:05:05].
- Google is in its seventh generation of TPUs, with the first version built in 2017 [17:10:00].
- The latest TPU in the series, Ironwood, boasts a single part exceeding 40 exaflops [17:43:00].
- TPUs are especially good at inference, helping manage the cost per query in search at scale [17:39:00].
Google continues to work with NVIDIA, using both TPUs and GPUs for Gemini traffic, providing customer choice and leveraging NVIDIA’s world-class R&D and software stack [20:12:00].
Cost and Latency Considerations
While there were concerns about the cost of serving AI-driven queries, Google has seen the cost per query fall dramatically in an 18-month timeframe [11:27:00]. The primary constraint is latency, as search has historically been near-instantaneous [11:33:00]. Ad revenue per AI query with AI Overviews is already at baseline compared to non-AI search, with potential for improvement [12:05:00].
Investment Strategy
In 2025, Google plans to invest $70 billion in capital expenditure [18:37:00]. The majority goes into servers and data centers [18:48:00]. Half of the compute spend in 2025 is directed towards its Google Cloud business [19:04:00]. This investment also powers innovations from Google DeepMind, pushing the frontier across various AI dimensions, including large language models, text, images, video, and world models [19:17:00].
Broader AI and “Other Bets” Innovations
Google’s AI investment and strategy extends beyond search, with AI horizontally impacting all aspects of its business, including YouTube and Cloud [15:06:00].
Gemini Ecosystem
Google is actively developing the Gemini app, which has seen an “uptick in engagement and usage growth” with the introduction of Gemini 2.5 Pro [09:20:00]. Recent features include deep research, updated canvas audio overviews, V2 video generation, and Gemini Live for screen sharing and interaction on Android phones [09:32:00].
Quantum Computing
Google has been making a patient investment in quantum computing for some time, working on it “out of conviction on the long-term trends” [41:52:00]. Quantum computing is seen as fundamental to conducting large-scale simulations that truly represent nature [42:10:00]. Pichai believes that within a five-year timeframe, a “really useful practical computation” will be performed quantum-style, far superior to classical computers, marking an “aha moment” for the industry [42:27:00]. Google aims to provide access to quantum computing through its cloud services [44:59:00].
Robotics
Google has one of the most advanced frontier R&D teams in robotics, with Gemini robotics efforts in vision, language, and action models being “world-class” [45:53:00]. While past attempts to enter the application layer were too early (e.g., selling Boston Dynamics), the current combination of AI plus robotics presents a new “sweet spot” [46:15:00]. Pichai predicts a “magical moment in robotics” within two to three years [47:18:00]. Google’s intrinsic team is developing models like “Android for robotics” to support robotics manufacturers [47:27:00].
Waymo
Waymo, Google’s self-driving car project, is highlighted as an example of “mindblowing persistence and patience” [41:18:00]. It is now on track to be a hundred-billion-dollar business, demonstrating the success of Google’s long-term, conviction-based approach to innovation [41:12:00].
Culture of Innovation
Google’s culture emphasizes investing in and empowering employees to foster a “positive, optimistic, innovation mindset” [48:48:00]. Perks like free food were intended to encourage cross-pollination of ideas [49:06:00]. Google believes this culture attracts higher-caliber talent who feel they have the agency to innovate [49:43:00].
Pichai notes that the intense and exciting nature of the current AI moment reminds him of early Google, with engineers working with passion and intensity [51:47:00]. The company has also reintroduced “labs” to foster innovation from small, 10-person teams [52:39:00].
Future Outlook
Google sees AI as a much bigger opportunity landscape than all previous technologies combined [33:15:00]. The future of human-computer interaction, in 5 to 10 years, will involve less human adaptation and more computing that “works for you” [25:50:00]. This seamless interaction could be driven by advancements like AR glasses, which Pichai believes will “wow people” and represent the “next leap” in user experience [26:11:00].