Australian businesses increase AI spending despite cheaper tokens
Australian companies are increasing spending on artificial intelligence, although the price of tokens — units of data that AI models process to perform tasks — is falling significantly. Businesses find it difficult to assess the technology’s actual returns and forecast future costs, ABC News Australia reports.
Anna Volkova, head of people and corporate culture at HR software developer HiBob, said her company sees benefits from AI, including faster work. At the same time, rapid changes in the technology’s capabilities mean businesses must continually review work processes, making it difficult for now to accurately assess the effectiveness of spending. In her view, AI’s impact on financial results may become noticeable in two to three years.
Andrew Charlton, Assistant Minister for Technology and the Digital Economy of Australia, said the country already spends between $5 billion and $8 billion on artificial intelligence annually. According to him, the overwhelming majority of these payments go abroad, and the government considers it possible for annual AI spending to rise to $20 billion-$40 billion over the decade.
Costs rise because of more complex tasks
According to the Epoch research institute, the cost of one million tokens for large language models at the level of GPT-4 Turbo or more powerful fell from $15 in November 2023 to 17.5 US cents in February 2025. Competition, more efficient microchips, the expansion of data centres and technological improvements contributed to the decline.
More current news is available on the UA.News Telegram channel Telegram.
At the same time, companies are finding new tasks for AI and increasingly using it for complex computations that require more tokens per request. Stu Scotis, AI lead at Deloitte, noted that a digital worker is not necessarily cheaper than a human: the cost depends on the complexity of the task and the duration of the model’s work.
Companies control model use
The most common way of paying for AI is by the number of tokens, although some companies pay for access for individual employees, including to Microsoft Copilot tools. Elastic representative Jeremy Pell noted that businesses are increasingly linking artificial intelligence initiatives to measurable returns.
Companies also distribute tasks among different models, using more expensive advanced models only where necessary. Some set token limits for employees for a certain period. AI expert John Whittle warned that unsystematic implementation of the technology complicates cost control: according to him, studies showed that estimates of the number of tokens for the same request using the same technology can differ by a factor of 30.
