Gemini API call volume doubles, Google AI licensing business reaches a profit turning point

Last year, Google’s Gemini model experienced rapid growth in the licensing business, becoming a new engine driving the company’s overall revenue increase. Several insiders close to Google’s sales team revealed that the licensing scale of the Gemini series has shifted from niche to mainstream, with market demand so strong that it even exceeded Google’s supply capacity expectations.

This growth is backed by Google’s massive investments in AI beginning to pay off. Last fall, Google disclosed an ambitious capital expenditure plan: investing $91 billion to $93 billion in infrastructure and AI capabilities, nearly doubling the $52.5 billion spent two years prior, demonstrating Google’s strong commitment to the AI race. Investors are closely watching when these huge investments will translate into profits.

Gemini API Business Turns from Loss to Profit

The most direct growth data comes from the volume of Gemini model API calls. Since the launch of Gemini 2.5 in March last year, developer enthusiasm has been fully ignited. API calls surged from around 35 billion initially to about 85 billion by August this year, doubling in size. This reflects widespread recognition of the model’s performance among developers.

To support this, Google increased API licensing delivery efficiency. With well-known developer tools like Cursor and GitHub Copilot integrating Gemini APIs, the surge in user demand forced Google to optimize delivery methods, freeing up more computing resources to meet the increased demand. This created a virtuous cycle: performance attracts developers, and developer usage provides more data and optimization opportunities for Google.

The shift in profitability is also noteworthy. According to internal data from sources familiar with the matter, early versions of Gemini (1.0 and 1.5) operated at a loss due to aggressive promotional discount strategies, resulting in negative profit margins. Gemini 2 improved but remained unstable, sometimes profitable, sometimes not. The real turning point came with Gemini 2.5—thanks to significant performance upgrades, Google could move away from price wars and compete based on performance advantages, achieving its first profit.

However, this profit only covered the token service costs of the model itself, excluding other R&D expenses. By mid-last year, the overall profit margin of the Gemini series had just turned positive, but it was still far below Google’s cloud business average. Nonetheless, this milestone indicates that Google’s AI strategy has shifted from burning money for market share to seeking sustainable profitability.

Multiplier Effect of Cloud Service Ecosystem

The true value of Gemini licensing growth may far exceed direct API sales revenue. Internal estimates suggest that customer spending on API calls often correlates with increased expenditure on other Google Cloud products like storage and databases. This means Gemini acts as a traffic gateway, attracting customers to deeply experience the Google Cloud ecosystem through high-quality AI capabilities.

A Google spokesperson stated: “Our cloud business is experiencing strong growth across the board, with AI applications performing particularly well.” Implicitly, AI is becoming a key weapon in Google’s cloud market share competition.

Google is also accelerating talent recruitment. Google Cloud and DeepMind are actively hiring AI specialists to prepare for the next phase of technological breakthroughs. This shows that Google is not relaxing despite recent growth but is instead increasing investment.

Challenges in Enterprise Application

Compared to the rapid growth of API business, the commercialization of Gemini at the application level faces more difficulties. The Gemini Enterprise Edition is a key product—integrating Gemini chatbot, enterprise data retrieval, and AI agent development platform, aimed at helping companies quickly deploy customized AI solutions.

In terms of user base, the results are promising. A Google spokesperson disclosed that the Gemini Enterprise Edition has 8 million users across 1,500 companies, with over 1 million online registrations, reflecting initial market penetration. Google plans to highlight this growth in its upcoming earnings report.

However, customer feedback is more complex. Sada, a consulting firm specializing in Google Cloud for enterprises, revealed an interesting phenomenon: among its clients, more than half hold positive views of Gemini Enterprise Edition, but overall, “satisfaction and dissatisfaction are nearly evenly split.” This indicates that while there are no major negative waves, the product has yet to achieve the industry-wide acclaim expected.

KPMG’s head of Google Cloud, Mark Shank, is more optimistic. Internal surveys show that 83% of employees are satisfied with Gemini Enterprise Edition, and procurement decisions are not difficult. This suggests acceptance varies between large enterprises and small to medium-sized businesses.

Chirag Meta, chief analyst at Constellation Research, pointed out the product’s capability limits. Feedback from clients indicates that Gemini Enterprise performs well in answering general questions and code generation but falls short on highly specific business problems or complex tasks. One consulting firm even tried to develop an email summarization agent using Gemini Enterprise—an officially recommended use case—but ultimately failed.

Meta’s assessment is that customers are still observing. “Everyone’s attitude is ‘we’re willing to give it another chance,’” he said, “but no one has completely abandoned it like they did with Copilot.”

Google’s Cloud Positioning Dilemma

Margo Lis, head of Google Cloud, raised a deeper question: Google’s cloud has always been positioned as a “developer platform” rather than a “ready-made software purchase platform.” This difference in positioning is affecting Gemini Enterprise sales. Enterprise clients face a choice: buy Google’s packaged Gemini Enterprise Suite or use Google’s tools to develop customized AI agents independently.

Ironically, Google’s ease of customization is too high. Many enterprise clients find that developing tailored solutions based on Gemini APIs is more efficient and better suited to their needs than purchasing the enterprise version. This means the success of Gemini APIs, to some extent, constrains the sales growth of the full enterprise software product—posing a new challenge for profit margin optimization.

Harvey’s application research director, Nico Gruben, exemplifies this client need: he extensively uses Gemini Enterprise’s data integration for research, documentation, and image generation but has little interest in developing intelligent agents, as they have the capability and willingness to build their own.

This situation clearly signals that convincing enterprise clients to pay for high-end software development is key to improving AI business profitability. Google needs to better define the value proposition of the enterprise edition; otherwise, it risks turning into a “showcase” for Gemini APIs rather than a profit contributor.

Outlook and Challenges

Google’s AI investments are approaching a harvest phase. The rapid growth of the Gemini API business indicates the technology direction is correct and market demand is real, with the massive capital expenditure initially validated. The jump from $52.5 billion to $91–$93 billion in investment reflects Google’s continued commitment—this time with doubled stakes.

However, the other side of commercialization—the profitability of application-level software—still requires time and product adjustments. Google must maintain API growth momentum while addressing issues in enterprise positioning and capabilities. Balancing these will be crucial for the future of Google Cloud.

In any case, the Gemini story is far from over. As market recognition of the model increases and Google continues to optimize product experience, the next chapter will unfold gradually through this year’s earnings and subsequent product iterations.