Google Reinvents Its AI Strategy

(Bloomberg) — Since the introduction of ChatGPT three years ago, there have been concerns from various analysts and technologists, including former Google executives, that the tech giant was falling behind in the crucial AI landscape. Those concerns, however, might be over.

Google has recently unveiled new AI tools and formed partnerships, notably with Anthropic PBC on chip technology, which has reassured investors about its competitiveness against OpenAI and others. The launch of Gemini 3, a versatile AI model, has garnered acclaim for its coding and problem-solving abilities, especially in areas where AI chatbots typically struggle. The cloud segment of Google is also thriving, benefiting from the global AI boom and the heightened demand for computing resources.

Evidence of increasing interest in Google’s specialized AI chips is surfacing, as they present a notable alternative to Nvidia’s products. Following news that Meta Platforms Inc. is contemplating the use of Google chips, Alphabet Inc.’s shares rose. The company’s overall market value has jumped by nearly $1 trillion since mid-October, largely fueled by Warren Buffett’s $4.9 billion investment last quarter and excitement from Wall Street regarding its AI initiatives.

On Tuesday, Alphabet’s stock climbed 1.5% to $323.44 in New York, pushing the company’s market cap to almost $4 trillion. In contrast, SoftBank Group, a major investor in OpenAI, saw its shares drop 10% amid worries about competition posed by Google’s Gemini. Nvidia’s stock also suffered, decreasing 2.6% and leading to a loss of $115 billion in market capitalization.

According to analyst Neil Shah from Counterpoint Research, “Google has definitely always been the dark horse in this AI race. It is a ‘sleeping giant fully awake.’”

Google executives have maintained that significant investment in research is key to staying competitive and retaining their position as a search engine leader while also shaping future computing technologies. The emergence of ChatGPT presented a genuine challenge to Google Search—something it hadn’t faced in years—despite the fact that Google was one of the pioneers in the technology employed by OpenAI. Nevertheless, Google possesses advantages that OpenAI lacks, such as abundant training data, profitable operations, and robust computing infrastructure.

Google’s CEO, Sundar Pichai, recently emphasized, “We’ve taken a complete, deep, full-stack approach to AI, and it actually works,” expressing confidence in their ongoing efforts. Regulatory concerns that might have held Google back are diminishing, especially after the company managed to avoid the most severe consequences of a U.S. antitrust lawsuit due to the recognition of threats from emerging AI competitors. Furthermore, Google has been making strides to diversify its business. For instance, Waymo, its self-driving car division, has expanded into new cities and introduced highway driving features to its taxi service—an achievement supported by extensive research and investment.

Part of Google’s edge lies in its comprehensive economic model. It is one of the few tech firms that develops a complete suite of computing technology—from AI applications and software models to cloud architectures and the chips themselves, including the popular Nano Banana image generator. With access to vast amounts of data from its services like search indexes, Android, and YouTube, Google has more control over its AI developments and can operate independently of external suppliers.

Several other tech players, including Microsoft and OpenAI, are exploring ways to create their semiconductors or build partnerships to lessen their reliance on Nvidia’s popular products. Historically, Google was the primary customer for its own processors, known as tensor processing units (TPUs). Initially designed to optimize search results, these processors are now adapted for complex AI tasks. In October, AI startup Anthropic announced it would utilize up to 1 million TPUs in a landmark deal worth billions.

Recent reports suggest that Meta plans to incorporate Google’s chips into its data centers by 2027. Google has refrained from confirming specifics but acknowledged a surge in demand for both custom TPUs and Nvidia GPUs. “As we have for years, we remain committed to supporting both,” a spokesperson stated.

While some in the industry recognize Google’s achievements, others see potential limitations. Meryem Alik, CEO of AI startup Doubleword, indicated that Google’s TPUs mainly attract a narrow segment of high-financial clients, such as Meta and Anthropic, while some experts believe companies might hesitate to commit to a single supplier. Developers of AI can only use Google’s chips via its cloud service, which contrasts with the flexibility that Nvidia’s GPUs offer. “Using TPU locks you into the Google cloud environment,” explains one analyst.

Thomas Hasson from Forrester summed it up, noting, “It’s safe to say Google is back in the game with Gemini 3. Reports of Google’s demise have been greatly exaggerated.”