What is the story about?
For a brief moment earlier this year, Grok looked like it had cracked the internet’s attention economy. But the momentum appears to be fading just as the AI race becomes more ruthless.
New market data suggests Grok is falling behind competitors including OpenAI’s ChatGPT, Anthropic’s Claude and Google’s Gemini, both in consumer adoption and enterprise use. The slowdown comes at a difficult moment for Musk, who is simultaneously trying to expand his AI ambitions while preparing SpaceX for a long-anticipated public offering.
The contrast is striking. Musk once described Grok as a “maximally truth-seeking” alternative to what he considered overly restricted or “woke” AI systems. Yet two years after launch, Grok’s biggest bursts of attention appear increasingly tied to controversy rather than sustained growth.
According to data from analytics firm AppMagic, Grok downloads dropped to around 8.3 million in April, down sharply from a peak of more than 20 million in January.
Part of that earlier spike came after Grok introduced tools that allowed users to virtually undress people in photographs. The feature spread rapidly online, with users generating sexually explicit images of celebrities, influencers and ordinary individuals. Reports that the system was also being used on images of minors triggered criticism from lawmakers, regulators and child safety groups.
Several countries temporarily restricted or banned access to Grok’s image-generation features after concerns emerged over non-consensual deepfake content involving women and children. Under mounting pressure, xAI eventually limited access to parts of the toolset.
The controversy may have driven engagement, but it did not translate into lasting growth.
Research firm Recon Analytics found that Grok’s paid adoption remained almost unchanged over the past year. In a survey of more than 260,000 American AI users and workers, only 0.174 per cent said they paid for Grok in the second quarter of 2026, barely different from 0.173 per cent a year earlier. By comparison, more than 6 per cent reported paying for ChatGPT.
Grok is also struggling to gain traction inside businesses, where AI coding assistants and workplace productivity tools are becoming the industry’s biggest battleground.
Enterprise Technology Research found only 7 per cent of surveyed companies said they were actively using and planning to continue using Grok. Claude and Gemini, meanwhile, showed far stronger adoption growth across enterprise teams.
As Grok struggles to keep pace, Musk has taken a surprising turn behind the scenes.
In early May,
SpaceX reportedly signed an agreement to provide Anthropic with massive computing capacity at one of Musk’s primary AI data centres. The deal gives Anthropic access to the full capacity of the Colossus 1 facility near Memphis, Tennessee, potentially generating billions of dollars annually for SpaceX.
The partnership marks a dramatic shift in tone from Musk’s earlier attacks on Anthropic. Only months ago, he described the company’s AI as “misanthropic and evil” in a post on X.
Yet the economics of the AI race are rapidly reshaping alliances. Companies including Anthropic and OpenAI are scrambling for computing power as demand for advanced AI models continues to outstrip available infrastructure.
New market data suggests Grok is falling behind competitors including OpenAI’s ChatGPT, Anthropic’s Claude and Google’s Gemini, both in consumer adoption and enterprise use. The slowdown comes at a difficult moment for Musk, who is simultaneously trying to expand his AI ambitions while preparing SpaceX for a long-anticipated public offering.
The contrast is striking. Musk once described Grok as a “maximally truth-seeking” alternative to what he considered overly restricted or “woke” AI systems. Yet two years after launch, Grok’s biggest bursts of attention appear increasingly tied to controversy rather than sustained growth.
Grok downloads fell after early viral success
According to data from analytics firm AppMagic, Grok downloads dropped to around 8.3 million in April, down sharply from a peak of more than 20 million in January.
Part of that earlier spike came after Grok introduced tools that allowed users to virtually undress people in photographs. The feature spread rapidly online, with users generating sexually explicit images of celebrities, influencers and ordinary individuals. Reports that the system was also being used on images of minors triggered criticism from lawmakers, regulators and child safety groups.
Several countries temporarily restricted or banned access to Grok’s image-generation features after concerns emerged over non-consensual deepfake content involving women and children. Under mounting pressure, xAI eventually limited access to parts of the toolset.
The controversy may have driven engagement, but it did not translate into lasting growth.
Research firm Recon Analytics found that Grok’s paid adoption remained almost unchanged over the past year. In a survey of more than 260,000 American AI users and workers, only 0.174 per cent said they paid for Grok in the second quarter of 2026, barely different from 0.173 per cent a year earlier. By comparison, more than 6 per cent reported paying for ChatGPT.
Grok is also struggling to gain traction inside businesses, where AI coding assistants and workplace productivity tools are becoming the industry’s biggest battleground.
Enterprise Technology Research found only 7 per cent of surveyed companies said they were actively using and planning to continue using Grok. Claude and Gemini, meanwhile, showed far stronger adoption growth across enterprise teams.
Elon Musk signed a deal with Anthropic after calling it ‘evil’
As Grok struggles to keep pace, Musk has taken a surprising turn behind the scenes.
In early May,
The partnership marks a dramatic shift in tone from Musk’s earlier attacks on Anthropic. Only months ago, he described the company’s AI as “misanthropic and evil” in a post on X.
Yet the economics of the AI race are rapidly reshaping alliances. Companies including Anthropic and OpenAI are scrambling for computing power as demand for advanced AI models continues to outstrip available infrastructure.
