Report by Bain & Co. It turns out that the AI sector is facing $800 billion problems.
Consulting Company's 6th Annual Global Technology ReportReleased on Tuesday (September 23) I said it is take Funding $2 trillion in annual revenue Computing Power Required To meet the expected AI demand by 2030. Even with AI-related savings, the world is still short at $800 billion If you're going to catch up request.
By 2030, the global incremental AI calculation requirements could reach 200 gigawatts, and Build it up Half of Power, Report I said. Even as US companies move all their on-premises IT budgets to the cloud and reinvest their savings as they apply AI to different aspects of their business in new data centers, the computational demand for AI is more than doubled. Moore's Law.
“If current scaling methods are maintained, AI will become increasingly burdened by supply chains globally.” David CrawfordChairman Bain's Global Technology Practicessaid on Tuesday News Release. “By 2030, technology executives will face the challenge of deploying approximately $500 billion. Capital expenditure I found about $2 trillion in new revenue To be beneficial to meet the needs. ”
Meanwhile, the computational demand for AI is moving faster than semiconductor efficiency can keep up, so this trend requires a “dramatic” rise in power supplies in grids that have not added capacity for decades. Added With release.
“It adds the dynamics of weapons racing between the nation and the major providers. And the possibilities of overbuilding and underbuilding have never been more challenging to navigate,” he said in the release. “To navigate the next few years, it is important to work through innovation, infrastructure, supply shortages and the potential benefits of algorithms.”
Advertisement: Scroll to continue
Meanwhile, this month, Pymnts explored it. The importance of reasoningthe stage in which the AI model is actually used to provide prediction, response, or insight.
As far as demand is concerned, the shift from research on generative AI to mainstream use has created billions inference Every day event. As of July of this year, Openai I said that Processing 2.5 billion prompts every dayIncludes 330 million users from US users Brookfield's predictions show that Three-quarters of all AI calculation demand Intention By 2030 it will come from inference.
“Unlike training, reasoning is in the production stage,” Pymnts wrote on Monday (September 22nd). “The delay, cost, scale, energy use, and location of deployment all determine whether the AI service works or fails.”
Subscribe daily for all PYMNTS AI coverage AI Newsletter.
