By Eoin Higgins
• 3 min read
“As an industry, we’re very much supply-constrained rather than demand-constrained,” AI exec tells IT Brew.
Getty Images
Cash rules everything around me, Method Man once rapped—in the IT sector, it’s AI that gets the money.
As AI becomes cheaper and more efficient, its ubiquity is exerting downward pressure on what companies can charge for their tools based on the technology. For organizations getting more value out of their AI tools, that could lead to questions about what to do with their surplus IT budget—including spending more on machine learning and language models.
On Nov. 17, OpenAI debuted its GPT-6.2 model, undercutting arch-rival Anthropic’s pricing during the latter’s IPO week. Monzy Merza, co-founder and CEO of AI security company Crogl, said that the OpenAI–Anthropic back-and-forth could be the beginning of prices falling.
It’s part of the “traditional technology curve,” Merza told IT Brew.
“If the prior technology inflection points are any indicator, then prices will continue to go down,” Merza said.
Line go up. Your AI spend depends on your specific needs. As Hao Yang, Splunk’s head of AI, told IT Brew, concerns over the instability of tokenization and general AI spend could drive down prices.
How to deploy excess budget capital is a question with changing parameters for IT teams, including whether to double down on cheaper tokens and AI tools. What’s undeniable is that overall AI spend is currently increasing, Yang said.
Off kilter? Instability impacts bottom lines. Without a holistic picture, there’s no way to know how to allocate surpluses, or even if you’re going to face a future deficit. Vico De Santiago, VP of innovation at Arionkoder, told IT Brew that a fundamental challenge for budgeting is outcome assessment.
“Every single month, that AI bill changes—in some months, it may justify accelerating a specific release for a client,” De Santiago said. “In others, you may be scratching your head and saying, ‘Hey, where did all this money go?’” Making sure that companies are investing time and money on how they monitor and track that is something that they shouldn’t underestimate.”
There are some tells. As Yang noted, more experimentally minded enterprises will see higher rates of AI use and spend. Meanwhile, the tokenmaxxing trend that once swept the industry has largely disappeared as a concern.
Unfortunately, that only adds to the unpredictability.
“As an industry, we’re very much supply-constrained rather than demand-constrained,” Yang said. “That also means that overall, collectively…their infrastructure spending on AI is actually going up pretty rapidly, not going down.”