Microsoft, Meta, or Google: Who Owns the Next Tech Era?
Smart World News |
Microsoft, Meta, and Google each have their strengths in the field of artificial intelligence, but competition in the future technological era will depend on the speed of technological iteration, market strategies, and the depth of industry applications.
Microsoft: Core Advantages in Enterprise AI
Microsoft holds a dominant position in the enterprise AI market through its partnership with OpenAI (an investment exceeding $13 billion) and the integration of Azure cloud services. Its GPT-4 Turbo technology covers 90% of the world's top 500 companies, and its Azure cloud services are growing at 30%, particularly excelling in generative AI and enterprise applications. Microsoft appears to be in a highly advantageous position. Like AWS, it also provides GPU cloud services and is the exclusive cloud provider for OpenAI. While the price is very high, considering that OpenAI seems poised to become another top technology company in the AI era, this means Microsoft is investing in the infrastructure for that era.
Bing is like the Mac before the iPhone: it contributes a considerable amount of profit, but only a dominant small portion, a relatively insignificant figure within Microsoft's overall picture. Integrating ChatGPT-style search results into Bing might risk gaining massive market share at the expense of its business model, but it's a worthwhile gamble.
A recent treport from The Information suggests that GPT will eventually be integrated into Microsoft's productivity apps. The trick is to mimic the AI coding tool GitHub Copilot (built on GPT), aiming to be helpful rather than cumbersome (e.g., not Clippy).
Crucially, adding new features—perhaps for a fee—perfectly aligns with Microsoft's subscription business model. Microsoft, once considered a prime example of a victim of disruptive change, has not only been born from it but is also in a favorable position to reach even greater heights.
There's much more to write about the potential impact of AI, but this article is already quite long. From the perspective of new companies, OpenAI is clearly the most interesting: OpenAI could become the platform for all other AI companies, meaning the economic value of AI outside of OpenAI could become rather limited.
Beyond image generation, open-source models in text generation are also likely to proliferate. In this field, AI has become a commodity: its impact on the world may be most significant, yet paradoxically, its economic impact on individual companies is the least.
In fact, the biggest winners may be Nvidia and TSMC. Nvidia's investment in the CUDA ecosystem means the company not only has the best AI chips, but also the best AI ecosystem, and is investing in expanding it. However, this has already spurred, and will continue to spur, competition, especially in internal chips like Google's TPU. Nevertheless, at least for the foreseeable future, all companies will be manufacturing chips at TSMC.
However, the biggest impact may be entirely beyond our predictions. Nat Friedman told me in an interview that Riffusion uses Stable Diffusion to generate music based on text. I wonder what other possibilities might emerge when images truly become a commodity. Currently, text is the universal interface because it has been the foundation of information transmission since the invention of writing. However, humans are visual beings, and the availability of AI in image creation and interpretation could fundamentally change the meaning of information transmission in ways that are unpredictable.
Meta: AI Revolution in Content Ecosystem
Meta, leveraging its Threads app (100 million registered users) and WhatsApp Business's advertising capabilities, drove a 7% increase in user time spent through its AI-powered recommendation system. While facing user growth bottlenecks, its AI-powered content recommendations in the Facebook Feed became the fastest-growing category, with advertising revenue increasing by 12% year-over-year. Meta boasts massive data centers, but these are primarily focused on CPU computing, as this is essential for Meta's services. CPU computing is also necessary for Meta's deterministic advertising models and content recommendation algorithms.
However, the long-term solution for ATT (App Tracking Transparency) is to build probabilistic models that not only determine who should be targeted by ads but also understand ad conversion rates. These probabilistic models will utilize a significant number of GPUs; for example, NVIDIA's A100 GPUs can cost five figures. This might be too expensive in a world where deterministic advertising is more effective, but Meta is no longer in that world, and it would be foolish not to invest in better targeting and metrics.
Furthermore, the same approach is crucial for Reels App's continued growth: recommending content across the entire network is far more difficult than recommending it only within a circle of friends and family, especially since Meta plans to recommend not only videos but all types of media, weaving them in with content you care about. AI models are also critical in this regard, and the equipment used to build these models is extremely expensive.
However, these investments should pay off in the long run. First, better targeting and recommendations are beneficial, which should restart revenue growth. Second, once these AI data centers are built, the cost of maintaining and upgrading them should be significantly lower than the initial cost of building them. Third, no company other than Google can make such a massive investment (and Google's capital expenditures will also increase).
This last point is perhaps the most important: AT&T has hurt Meta more than any other company, as AT&T already has the largest and most sophisticated advertising business to date, but in the long run, it should deepen Meta's moat. This level of investment is simply not feasible for Snap, Twitter, or other laggards in the digital advertising space (even though Snap relies on cloud providers rather than its own data centers). When you combine the fact that Meta's ad targeting may be moving away from this area with the fact that Reels has led to a significant increase in inventory (which will lower prices), you can't help but wonder why advertisers would bother going elsewhere.
Google: Continued Power from Technological Foundation
Google maintains its lead in AI chips, knowledge graphs, and lifelong learning systems, with Project Ellmann reshaping the trajectory of education through AI. Its cloud business is growing at 31.5%, and advertising revenue accounts for 77.93% of total revenue, but its over-reliance on search advertising may be facing a growth ceiling. The Innovator's Dilemma was published in 1997, the same year Kodak's stock price reached its peak of $94.25, and there seems to be a reason for that: technologically, Kodak was in an excellent position. The company not only dominated film technology at the time but also spearheaded the next wave: digital cameras.
This problem boils down to a business model issue: Kodak made a lot of money by supplying silver halide film, while digital cameras, being digital, meant they didn't need film at all. Therefore, Kodak's management was so confident that digital cameras would forever remain only for amateurs, and only when they became very cheap—a process that would certainly take a long time.
In fact, Kodak's management was right: it took 25 years for digital cameras to surpass film camera sales since their inception; and even longer for them to be adopted in professional settings. During this period, Kodak made a fortune and paid out billions of dollars in dividends. Although Kodak went bankrupt in 2012, it was because consumers had better products: first digital cameras, then camera-equipped smartphones.
To be sure, the idea that this was a happy ending is a contrarian perspective: most people see Kodak as a failure because we expect companies to last forever. From this perspective, Kodak's case serves as a warning—how an innovative company can let its business model lead it to its ultimate doom, even if that doom is caused by consumers having better products.
This then leads us to Google and AI. Google invented Translator, a key technology underpinning the latest AI models. Google reportedly possesses a chat product far superior to ChatGPT. Google claims its image generation capabilities surpass those of Dall-E or any other company on the market. However, this is just a claim, as no actual product is seen on the market.
This is not surprising: Google has long used machine learning to improve its search and other consumer-facing products (and offers these technologies as a service through Google Cloud). However, search has always relied on humans as the ultimate arbitrator: Google provides links, and users click to decide which is correct. This extends to advertising: Google's service is revolutionary because it doesn't charge advertisers based on impressions (the value of which was difficult to determine, especially 20 years ago), but rather on clicks, where the target audience of the ad decides whether their ad is good enough.
