From: allin
The landscape of search is undergoing a significant transformation with the emergence of artificial intelligence (AI) tools, leading to an intense AI competition between long-standing rivals Google and Microsoft. This battle centers on integrating generative AI into search engines, promising to redefine how users find information and how search companies monetize their services.
The New Battlefield: AI in Search
Traditional search engines, exemplified by Google, operate on an “information retrieval” model where a query results in a list of ranked links from an index [00:36:13]. Over time, Google evolved to provide more direct answers in “one-box” results for specific queries, like movie showtimes or local businesses [00:36:51]. However, generative AI, such as OpenAI’s ChatGPT, offers a fundamentally different approach: natural language responsiveness that can provide a direct answer, potentially eliminating the need to click through multiple links [00:38:16].
Microsoft’s Aggressive Stance
Microsoft, through its significant investment in OpenAI (the developer of ChatGPT), has positioned itself as an aggressive challenger in this shifting search landscape [00:32:25, 00:32:55]. Microsoft CEO Satya Nadella has openly declared his intention to make Google “dance,” aiming to disrupt Google’s dominant 93% market share in search [00:33:38, 00:42:39]. Bing, Microsoft’s search engine, is integrating ChatGPT into its search experience, aiming to get ahead of Google in this new wave [00:32:25].
This strategy involves potentially “scorching the Earth” by making AI tools pervasive, forcing Google to adapt in ways that could degrade its profitable business model [00:42:42, 00:42:50].
Google’s Challenges and AI Strategy
Google has long possessed significant AI competency, particularly since acquiring DeepMind [00:35:01]. This AI investment has primarily been focused on internal optimizations, such as improving data center energy efficiency, ad optimization, and YouTube’s video suggestion algorithms [00:35:15]. However, the public emergence of ChatGPT has forced Google to bring its AI search capabilities to the forefront [00:35:51].
Google’s initial AI demo for its chatbot Bard was poorly received, causing its stock to drop by 12% [00:34:50, 00:34:54]. Bard notably provided an incorrect answer during the demo, stating it took the first pictures of a planet outside our solar system, which was false [00:34:12, 00:34:16]. This public stumble highlighted Google’s AI challenges and the difficulty of deploying these complex models accurately [00:34:04].
One proposed counter-measure for Google to maintain its dominance is to significantly increase its Traffic Acquisition Costs (TAC) [00:45:16]. By doubling TAC, Google could secure exclusivity on search traffic and compel publishers to restrict AI agents from crawling their content, thus depriving competitors of valuable data [00:46:27, 00:46:50].
The Cost of AI Search
A significant challenge for generative AI in search is the computational cost. A back-of-the-envelope analysis suggests that running a GPT-3 model for each search result costs approximately 30 cents in compute, an order of magnitude higher than Google’s traditional search cost of about 2.5 cents per search [00:39:21, 00:39:31, 00:39:38]. Scaling this to Google’s volume would cost an estimated 8-20 billion [00:40:20, 00:40:35]. This high cost necessitates a 10x reduction in compute costs for widespread economic viability [00:39:57].
However, Microsoft’s investment in OpenAI includes investing in Azure infrastructure, which is expected to drive down these costs through software, data, chip, and cloud optimization [00:41:14]. The increasing efficiency of specialized silicon and cheaper energy costs are expected to lead to a 10x reduction in model running costs over the next few years [00:41:30, 00:41:45].
Impact on Business Models and Monetization
The shift to AI-driven search poses a direct threat to Google’s highly profitable ad-based business model [00:43:34]. If AI provides direct answers, users may not click on sponsored links, which currently generate significant revenue [00:49:28, 00:49:53]. Google’s model thrives on a delicate balance where paid ads can sometimes offer a better user experience by leading directly to purchase pages, reducing “bounce back” rates (users returning to search after clicking a link) [00:50:03, 00:51:11].
The integration of AI could significantly impact Google’s business model by degrading its core profit [00:43:09, 00:43:29]. Even a small loss of market share (e.g., 5-6%) could create “enormous headwinds and pressure” for Google’s valuation [00:44:44, 00:44:58].
For content publishers, a major concern is how they will be compensated if AI models synthesize their information without direct clicks or attribution that leads to revenue [00:47:29, 00:47:40]. While some AI demos show citations, the question remains whether this translates to fair compensation or if it interferes with the original content owner’s ability to profit [00:59:49, 01:00:01]. Legal challenges, similar to those faced by YouTube in its early days, are anticipated regarding copyright and fair use [01:03:07, 01:28:01]. Content creators, such as Getty Images, are already suing AI companies for using copyrighted material for training [01:28:01].
Broader Implications of AI Productivity
Beyond search, AI’s impact on productivity is seen as transformative across various industries:
- Application-Specific Assistants: AI could be integrated into nearly every application, allowing users to interact via voice or text to accomplish complex tasks, such as generating Excel formulas [01:06:50, 01:06:58].
- Code Generation: AI tools like GitHub Copilot can assist engineers in writing code, potentially increasing developer productivity by 50% and enabling smaller teams to build complex software like Stripe [01:13:53, 01:14:02, 01:08:08]. This could lead to “margin destruction” in middleman businesses [01:09:54].
- Content Creation: AI can democratize content creation, from generating user interfaces (e.g., Galileo AI for app designs) to writing blog posts or creating music mashups with famous voices [01:11:20, 01:24:44, 01:15:48]. This could lead to an explosion of creation due to ease of use, though it raises questions about compensation for original content providers [01:25:34, 01:25:51].
- Market Competition: Increased ease of building products through AI could lead to more niche startups and greater competition across all industries, potentially driving down prices and profits in the aggregate [01:13:16, 01:31:31].
Ultimately, technology is seen as fundamentally deflationary, lowering costs but also potentially aggregate revenue and profit [01:23:31]. The current AI and its impact on Google’s business model may force Google to cannibalize its own business before others do [01:23:55]. This shift in power dynamics also has implications for antitrust regulation, as increased competition, even from a larger entity like Microsoft, could make it harder for regulatory bodies to argue against monopolies [01:29:19, 01:29:40].