Cybersecurity giant Palo Alto Networks’ CEO Nikesh Arora said that AI tokens need to become 10 times more efficient, and that AI companies need to lower prices for enterprises in order for large-scale AI adoption to occur.
“The demand for AI consumption is still infinite. We do not have enough compute in the world to satisfy the demands of both the consumer side and the enterprise side,” he said in a June 9 interview with CNBC. “The conversation will shift over the next six to 12 months on—not how smart the model is [or] how intelligent the model is—[but] how quickly can we deploy the models.”
Arora said the efficiency of tokens—the units that AI models use to process texts and how they are priced for enterprises—needs to improve by five to eight times within the next 12 months, and by 10 times the year after.
“We need to see the pricing for AI come down. I think the pricing today even makes it very hard for enterprises to take the bet,” he said. “We don’t need friction introduced by [AI] model companies from a pricing perspective and a deployment perspective.”
Palo Alto has been one of the biggest beneficiaries of AI adoption, according to Arora, as the company’s stock has jumped almost 77 percent since the start of the year.
The company, in its latest earnings report, published on June 2, said revenue increased 31 percent year-over-year to approximately $3 billion, including $388 million from CyberArk and Chronosphere, two companies that Palo Alto acquired last year.
“The latest advancements at the AI frontier have increased the level of urgency around cybersecurity, and redefined the shape of the industry for the coming years,” Arora said in the earnings report’s press release.
Other large cybersecurity companies have also had major gains, such as Crowdstrike, in which shares have jumped nearly 60 percent since the start of the year.
On July 8, xAI, now known as SpaceXAI, launched Grok 4.5, the newest model of its chatbot, and touted its token price at $2 per million input tokens and $6 per million output tokens. Its previous model, Grok 4.3, costs $1.25 per million input tokens and $2.50 per million output tokens.
OpenAI’s ChatGPT priced its latest GPT-5.6 model between $1 and $10 per million input tokens and between $6 and $45 per million output tokens, depending on the model type.
Anthropic’s Claude has its latest Fable 5 model priced at $10 per million input tokens and $50 per million output tokens, though some of its older models are at least half the price.
