The 2025 AI Race: Accelerating Elimination
Google, Microsoft, and OpenAI gear up for 2025, focusing on AI advancements, monetization, and consumer value amid intense generative AI competition.
“From a historical perspective, you don’t always need to be the first, but you must perform well and truly become the best in your category.”
In December, Google CEO Sundar Pichai remarked during an internal strategy meeting, “This will be the theme of 2025.”
It was a “call-to-action” meeting for the entire Google team.
According to an audio recording of the meeting obtained by CNBC, several top Google executives participated, while some employees joined remotely.
Sundar Pichai repeatedly urged employees to stay motivated and emphasized the critical importance of 2025.
Powered by Gemini, Google performed commendably in the AI sector last year, delivering standout products like NotebookLM. However, the CEO believes 2025 will be pivotal, and Google still has ground to cover.
“We must recognize the urgency of this moment and act faster as a company,” Pichai said. “In 2025, we need to relentlessly focus on harnessing the benefits of this technology and solving real user problems.”
As the intense competition in AI enters its third year, the race is approaching its second half. Not only Google but also other major players have set ambitious goals for 2025.
Halftime Review: Established Giants Focus on Future Monetization
Facing increasing regulatory scrutiny due to a series of antitrust lawsuits, Google has been under mounting pressure.
Last summer, the U.S. Department of Justice called for Chrome browser to be separated from Google and filed antitrust lawsuits targeting Google’s core advertising business.
Meanwhile, Google’s lifeline, its search business, faces competition from OpenAI and Perplexity.
To respond, Pichai emphasized the need to build large new businesses, including the AI-driven Gemini program.
Demis Hassabis, founder of DeepMind and leader of Gemini, stated that Gemini-backed products aim for significant growth within the next year or two.
During the meeting, updates were presented on several consumer-facing products, including NotebookLM.
Similarly, Microsoft is doubling down on bringing AI closer to consumers in a more aggressive manner.
Some users told The Wall Street Journal that subscription fees for Microsoft Office products quietly increased after upgrading to the Copilot version.
By integrating Copilot into its suite of Office tools, Microsoft allows users to write emails, summarize meetings, and create PowerPoint presentations more effectively.
The premium consumer version of Copilot launched in January last year in the U.S., priced at $20 per month, in addition to the $7 monthly 365 personal subscription fee.
Enterprise software costs $30 per person. During Microsoft’s recent earnings call, the company revealed that its AI business is on track to generate $10 billion in annual revenue.
Integrating Copilot into 365 proved to be the best way to elevate its position. Without this, as a standalone chatbot product, Copilot had only 37 million app downloads — nowhere near ChatGPT’s 433 million.
However, Copilot’s text editing, polishing, and writing features have not created a strong enough competitive edge. For its subscription price, its value is relatively low.
In other words, in 2025, Microsoft needs to prove the value of AI in office scenarios — delivering not just minor improvements but value worthy of subscription fees.
The generative AI race has entered its second half.
In the first two years, users were intrigued by text length and reasoning capabilities. But now, no matter how much a model’s context window expands or how fast its response time gets, it’s hard to excite users.
After all, for consumers, only one thing matters: What’s truly useful?
This will be the core question Google and Microsoft aim to answer next year. Whoever delivers the best answers will gain an advantage in the AI competition.
New Players Carve Their Paths
Microsoft and Google stand out as “older giants” in this arena, having built vast product networks over previous generations.
For them, it’s time to move from R&D and infrastructure to testing real-world revenue potential.
For younger tech players, the situation is different.
For example, during Christmas 2024, Sam Altman publicly asked netizens for ideas on what OpenAI should achieve by 2025. On December 31, he unveiled a series of vague keywords.
These hints suggest that 2025 could be the year OpenAI fully transforms into a true company.
In 2024, OpenAI began transitioning its organizational structure. Previously a nonprofit aiming to achieve AGI, it realized that achieving its goals was extraordinarily costly. Its nonprofit structure hindered its ability to compete effectively in the fiercely competitive AI market.
“Investors want to support us, but as capital scales up, we need a traditional equity structure to reduce customization in our corporate setup,” OpenAI stated in a recent blog post.
Upgrades to its models and iterations of its products are closely tied to this transformation. Only by embracing commercialization can OpenAI continue its R&D breakthroughs; conversely, R&D requires a company structure to attract more funding.
Meta, who grew up in the social media era, has taken a slightly different route. While it carries historical baggage, it’s less burdened than Microsoft and Google.
Unlike OpenAI, which is still transitioning, Meta has been operating stably as a commercial entity.
Meta even benefited from past failures. After the social media boom ended, its bet on the metaverse turned out to provide valuable technical reserves, such as partnerships with chip manufacturers for custom chip development — key to training Llama.
Llama, being open-source, doesn’t generate much revenue. Other AI features on Facebook and Instagram aren’t lucrative either. Zuckerberg opted for hardware as the path forward.
Last September, Meta unveiled the Project Orion AR glasses at its Connect event, positioning them as a standout smart wearable device.
Three months later, the smart glasses co-developed by Ray-Ban and Meta received three major upgrades, with Meta AI playing a crucial role. The glasses not only “see” what users see but can also interact with users about the content.
Developing smart glasses has been Meta’s ongoing effort. In 2023, media reports revealed its AR/VR roadmap, with high-end products codenamed Orion, expected to undergo internal testing in 2024 and launch by 2027.
Meanwhile, Meta is already training Llama4, aiming for deeper integration between its models and glasses.
Despite differing approaches, one future goal unites everyone: agents.
Sam Altman’s list ranks agents just below AGI; Gemini has launched Project Astra; Meta has Shazam; and Copilot is Microsoft’s bet on agent form factors. With its upcoming launch in the European market next spring, Apple Intelligence is poised to reach more users.
As the second half unfolds, the competitive landscape is stabilizing. Major tech companies continue to dominate due to the high costs of this race.
The real disruptions will come from practical applications of generative models and agents. These signify AI’s shift to becoming more reliable and providing tangible value in the real world.
