AI’s Cost Curve Has Big Tech Losing Money

Artificial intelligence (AI) has captured the imagination of some of the world’s largest companies.

It has also captured a sizable portion of their spending — without giving much back.

Big Tech companies, including Microsoft, Google and others, are struggling to profitably monetize their generative AI products due to the costs that the models require to produce, develop and train, The Wall Street Journal (WSJ) reported Monday (Oct. 9).

AI represents a new phase shift in technology, one that breaks away from the known and familiar economies of scale of existing software products.

“A lot of the customers I’ve talked to are unhappy about the cost that they are seeing for running some of these models,” Amazon Web Services (AWS) CEO Adam Selipsky told WSJ, speaking of the industry broadly.

A study published Tuesday (Oct. 10) titled “The growing energy footprint of artificial intelligence” found that by 2027, AI servers could be responsible for as much energy use as the entire country of Sweden, or around 0.5% of the entire world’s electricity.

And that represents a middle-ground estimate.

Managing for AI’s costs will require the development of new business models that allow firms to pass operating costs and cost-savings to their end-users through disruptive pricing strategies.

Read also: Costs Are Top of Mind as LLMs Upgrade Chatbots for Industry-Specific Use Cases

AI Is a Money Loser Because It Is So Expensive to Run

While the AI economy is still in the initial development stages, where companies are seeking to generate public interest in their products and capture customers for their platforms, the technology is entering the end of its first full year of commercialization without having cracked the profitability nut.

The high cost of AI, driven primarily by the computing power an AI model requires — which grows in step with the number of customers using the product — is an uncomfortable and expensive reality that businesses need to adapt to in order to remain competitive.

Analysts estimate that Microsoft’s Bing AI chatbot, which is powered by OpenAI, needs at least $4 billion of infrastructure just to do its job. OpenAI spends up to $700,000 a day maintaining its underlying infrastructure and server costs, and the company recorded total losses of $540 million in 2022.

See also: Peeking Under the Hood of AI’s High-Octane Technical Needs

A single AI query can cost nearly 1,000 times that of the same question asked of a normal Google search, making the margins for AI applications smaller than other Software-as-a-Service (SaaS) solutions.

Even when they aren’t being used by customers, AI models need to be constantly re-trained and fine-tuned to stay relevant and safe, a computationally heavy process that isn’t necessarily easy on a firm’s war chest.

Despite all this, the generative AI industry itself is expected to grow to $1.3 trillion by 2032.

Companies are also drawing solace from the fact that many technical innovations tend to debut with sky-high costs before coming back down to Earth as they hit their stride in the marketplace and find cost and production efficiencies over time, as subsequent ecosystem innovations drive down costs.

Moving Beyond Excitement to Adoption

Crucial to reaping the benefits of AI without overpaying for it is identifying and acting upon the right implementations.

For example, as AI-powered scams become a greater risk for financial institutions, AI-powered defenses are becoming a must.

“[When it comes to fraud prevention, some] organizations are still debating how to leverage AI — but we’re beyond ‘how’ at this point,” Karen Postma, managing vice president of risk analytics and fraud services at PSCU, told PYMNTS. “Firms need to leverage AI now; they need to have done it yesterday.”

PYMNTS has also been tracking how AI solutions can revolutionize finance, accounting and procurement workflows.

“You can delight customers and capture more customers when underwriting is seamless, the credit process is seamless, and how money flows is seamless and with less error,” Aanchal Kochhar, head of product at Capital One Trade Credit, told PYMNTS. “There is a lot of growth potential [when leveraging AI].”

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.