Here's the latest scorecard in the AI chatbot wars, and it looks like OpenAI's ChatGPT is still wearing the crown. According to new data from BNP Paribas, ChatGPT held its lead in both web traffic and mobile app usage in July, and actually took back some ground from Alphabet Inc.'s (GOOGL) Gemini.

But the real story might not be who's winning the popularity contest. BNP Paribas also flagged a significant shift in how AI money is being spent, moving from building models to actually using them. And there's a growing push to fund the infrastructure that makes all this possible.

ChatGPT Regains Ground Against Gemini

Let's break down the numbers. ChatGPT remained the clear leader in web traffic and mobile app usage in July, according to BNP Paribas. It also regained share across both visits and daily active users, or DAUs.

Gemini's share of worldwide visits slipped to 28.8% in July from 30.3% in June. Its DAU share held steady at 19.8%.

Meanwhile, Claude, the chatbot from Anthropic, kept climbing. Its share of visits rose to 10.4% from 10%, while its DAU share stayed flat. Grok, on the other hand, lost some ground. Meta Platforms Inc.'s (META) Meta AI remained stable or improved slightly.

BNP Paribas noted that Grok and Meta AI are still trailing the big three: ChatGPT, Gemini, and Claude.

Overall, AI chatbot web traffic dipped about 1% from June. But here's a silver lining: average monthly mobile DAUs and monthly active users each ticked up about 1%. So while fewer people might be clicking around on desktop, mobile engagement is actually growing.

AI Spending Shifts Toward Real-World Use

Now, let's talk about where the money is going. BNP Paribas highlighted a shift in AI spending toward inference, which is the process of using trained models to run real-world applications. Think of it as the difference between baking a cake (training) and actually eating it (inference).

The firm cited Gartner estimates showing global inference spending will hit $23.3 billion in 2026, surpassing the $19 billion expected for AI model training. That's a big deal. It suggests the industry is moving beyond just building frontier models and starting to deploy them in practical ways.

BNP Paribas said this trend underscores a shift toward deployment and real-world usage as AI adoption expands.

AI Infrastructure Demand Builds

Of course, all this AI activity needs serious hardware. BNP Paribas also highlighted NVIDIA Corporation's (NVDA) partnerships with major private equity and asset management firms, including Apollo Global Management, Inc. (APO), BlackRock, Inc. (BLK), Blackstone Inc. (BX), and Brookfield Asset Management Ltd. (BAM). The goal? To mobilize more than $500 billion in third-party capital for AI infrastructure.

That's a massive number, and BNP Paribas said it underscores significant and growing demand for AI infrastructure.

The firm also pointed to surging demand for high-bandwidth memory, or HBM, which is putting pressure on memory supply and pricing. If you're wondering why memory stocks have been acting up, this is part of the reason.

And here's another interesting development: CME Group plans to introduce Compute futures with Silicon Data on Oct. 5. BNP Paribas said this launch highlights AI compute's emergence as a commodity, as companies look for ways to manage pricing and capacity. It's like trading oil or wheat, but for computing power.

As for the market's reaction, Alphabet shares were down 0.34% at $345.18 at the time of publication on Friday, according to market data.

So, what's the takeaway? ChatGPT is still the chatbot to beat, but the real action is happening behind the scenes. AI spending is shifting from training to inference, and the infrastructure buildout is attracting serious capital. It's a sign that AI is moving from the lab to the real world, and investors are paying attention.