This past week, Microsoft announced the new AI-powered Bing search engine. With that, the first salvo of the new AI war was thrown.
For the past several years, Microsoft has heavily invested in OpenAI, the non-profit AI startup and creator of Chat GPT and Dall-E. The total investments have exceeded $10 billion. As a result, Microsoft is the exclusive cloud provider for OpenAI. In exchange, Microsoft will infuse many of its products with OpenAI’s Generative AI. One of the first outcomes of Microsoft’s and OpenAI collaboration was Copilot, an AI pair programmer.
This partnership seems to have refreshed Microsoft’s image. The company has traditionally been seen as the stodgy incumbent and has taken the backseat to Google, with its many bets, on the innovation front. After all, the new era of generative AI was birthed by Google, whose team wrote the 2017 seminal paper on Transformers (the network architecture that underpins these new generative AIs). However, Google has been slow to release AI-powered products. It has the most to lose, so it moves more slowly and carefully. Meanwhile, OpenAI has been heavily recruiting AI experts from Google.
This past week, the stake got a lot larger. Microsoft, in partnership with OpenAI, announced that it is revamping Bing, its search engine, using AI. This strikes the heart of Google’s core business.
To better understand the significance of this move, let’s first discuss the business of search.
The business of search
The search advertising business is arguably one of the most profitable businesses in the world. Google’s search business generated $162 billion in 2022 (or 55% of total revenue, growing 43% and 9% in 2021 and 2022, respectively). Currently, Google is in a very dominant position. It has roughly 93% of the search market share, with Microsoft’s Bing at 3%.
At a high level, of 10 billion search queries per day, there are three types of search. Each has different levels of monetization and profitability:
- The first is navigational search. This is where you are trying to find something. For example: “where is the post office?”
- The second is informational. For example: “what is the weather outside?” or “what was the score from the game last night?”
- The third is a catch-all search, everything else. These are typically deeper research in nature. For example, trip planning (“I’m visiting Rome in June for 3 days, what are the top sites I should visit?”, complex shopping (“I plan on buying an IKEA Micke desk, will it fit my car?”), etc.
This third category of search is the trickiest one to solve and the one where the user needs to string together multiple queries to find answers. This category is also likely harder to monetize.
According to Microsoft, about 40% of searches resulted in users clicking the back button immediately, indicating that they did not find the answer they were looking for. Over the years, search evolves and, in some areas, becomes more verticalized:
- Amazon is the dominant shopping search engine. About 60% of US shoppers start their product search on Amazon
- Reddit is gaining traction as an advice search engine. When looking for advice, many users now append “Reddit” on their search queries to get answers posted by Reddit users rather than SEO articles
- YouTube is the dominant video search engine
- There is also Yelp, which specializes in SMB vertical searches for restaurants, dentists, hair salons, and many travel-related search sites
The past few years have also seen a plethora ($) of Generative AI startups. It is Microsoft’s goal (and Google’s as well) to combine these two and create a new search engine experience.
The new AI-powered Bing will mostly help answer the ‘everything else category’ search mentioned above. In its announcement, Microsoft demoed how the new AI-powered query works. One example was the query, “what can I substitute egg for in a cake recipe, give me exact amounts”. Not only did Bing give the potential answers, but it was also able to highlight the pros and cons of each option (see Figure 2 below).
Another cool example from the demo is the ability of Bing to synthesize text, in this case, earnings release, and compare the key metrics from the earnings document with that of another company. As someone who reads many of these earnings releases, this looks amazing (see Figure 3 below).
Undoubtedly, the integration of ChatGPT into Microsoft’s products won’t end here. It has already announced that it will provide API access to other companies to build their version of ChatGPT using their data. Microsoft will soon demo ChatGPT integration into the Office suite.
As we mentioned above, the third category of search will be the initial focus to be tackled by AI. But serving these queries via ChatGPT is not cheap. The cost of one ChatGPT chat session can be looked at through single-digit cents. Assuming that it’s $0.05/chat, back-of-the-envelope math tells us that the variable cost of serving 40% of the 10 billion queries per day is $200 million per day or $73 billion per year. This does not even include the billions that must be spent on capex. Undoubtedly that cost would go down over time, but this transition would likely be margin destructive for Google.
But for Microsoft, which only has about 3% search market share, the impact of the higher cost would be much less. In fact, AI-powered search does not have to be profitable for Microsoft, as it can meaningfully drive adoption of Azure cloud services, which can be beneficial for branding and therefore recruiting, and can drive the company’s SaaS sales. As Satya Nadella, CEO of Microsoft, stated:
“From now on, the [gross margin] of search is going to drop forever,” Satya Nadella, CEO of Microsoft, in an interview with the Financial Times.
On the revenue side, Microsoft expects that for every percent of market share gained in search, its annual advertising revenue will grow by $2 billion (this translates to 17% of its advertising revenue in 2022).
To dominate what is arguably one of the most profitable segments in the world for over 20 years is no small feat. Google has become a verb. Its products are ubiquitous. Everyone adapts their sites to cater to Google’s search algorithm, so much so that in the US alone, Google has spawned a $75 billion industry where consultants and marketers sell their services to help you rank well within Google’s algorithm (SEO). Google’s dominance seems to be unquestionable.
That is until now.?
One day after Microsoft’s announcement, Google cobbled together a hastily arranged presentation. During this live broadcast, Google’s new generative chat AI, Bard, gave an incorrect answer. This triggered a 9% share price slide (erasing a $100 billion market cap). This flubbed presentation wiped out more than what Google gained a few weeks ago when its share price jumped 5% after it announced the layoff of 12,000 people.
I doubt we’ve seen a product demo that caused such a large decline in share price for such a large company before. Did the market overreact? Probably.
The ubiquity of Google search is predicated on its control of demand:
- Google pays Apple between $18 billion to $20 billion in 2022 (a figure that is increasing year-over-year) to be the default search engine on Apple devices
- The company also spends billions of dollars developing Android, YouTube, Gmail, its Google Doc suite, Google Maps, and Chrome. These are some of the world’s most used applications. Chrome has a 65% global market share, while Microsoft’s Edge is at 4.5%.
The strength of Google’s search moat will take years to be undone. In the meantime, the company is not sitting idle. Despite the botched presentation, Google still has one of the best AI talent collections in the business. It owns DeepMind, a pioneer in cutting-edge AI research. It also recently invested $400 million in Anthropic (a ChatGPT competitor). And it builds its own custom AI chips (Tensor Processing Units or TPUs) that may work better for these new AI models.
The definite losers
For many companies, especially small businesses (and startups), who do not have large marketing budgets to drive awareness, generating organic traffic via SEO has been the go-to way to acquire customers cheaply. Many of these SEO articles are ‘how to’, specifically written for the third type of search. In the not-too-distant future, this content will be crawled and summarized by AI bots to be served to users, potentially rendering many SEO strategies for organic traffic acquisition obsolete. This will disintermediate websites from organic traffic (on the flip side, there will be fewer spammy SEO articles on the internet).
While we are only in the early stages of the battle, the first salvo has been thrown. Microsoft showcased a very compelling integration of ChatGPT in its search engine, which was followed by an effective PR blitz that drove the narrative home (its share price rose 4% after the announcement). Even though the war has just begun, Nadella and his team at Microsoft seem to have gotten the first laugh.
“And I hope that, with our innovation, [Google] will definitely want to come out and show that they can dance. And I want people to know that we made them dance, and I think that’ll be a great day.” — Satya Nadella, CEO of Microsoft