CSuites Worry About AI's Soaring Costs, But Some Think It May Still Pay Off

The Csuite case for AI to justify its staggering upfront costs is based on faith in its power to repeat what electricity, automobiles, oil, telephones, computers and mobile phones have done - with an added fillip of confirmation bias based on the sums already invested. 

But interestingly, there is also a belief among the AI optimists that they will be able to identify how consumers will want to use it and monetize that, not repeating the failed expectations of Siri, Alexa, the metaverse and countless other "next new things." It's a tall order and not out of the question - but also, perhaps surprisingly uncertain, not a slam dunk - with lots of jobs and bonuses are riding on it. Smart companies are wary of relentlessly positive hockey stick projections. JL

Steven Rosenbush reports in the Wall Street Journal:

Nvidia infrastructure would have to generate $200 billion in lifetime revenue to justify companies’ spending on AI over the course of just one year—and the revenue isn’t anywhere near on track to hit that mark soon. To justify a sufficient return on the world’s total AI investment, AI must drive revenue on a comparable scale, transforming companies, entire industries and the economy. (But) CEOs believe they can make AI pay off despite the staggering costs. “As a nation we pumped trillions of dollars into modernizing the electrical and infrastructure to create the economy we have today. If AI is a parallel to that the best products will require as much intuition about how customers want to interact with AI as they’ll require technical expertise,”

A few months ago, it might have seemed that the stakes of the artificial-intelligence investment wave couldn’t have gotten much higher. Well, they have.

Sequoia Capital last September published a report arguing that Nvidia infrastructure would have to collectively generate $200 billion in lifetime revenue to justify companies’ spending on those advanced AI systems over the course of just one year—and the revenue wasn’t anywhere near on track to hit that mark soon.

Given the torrid growth since then at Nvidia, whose dominant role in AI infrastructure makes it a good proxy for how much companies are spending on AI chips and systems, I asked Sequoia Partner David Cahn to update his research.

“AI’s $200 billion question is now AI’s $600 billion question,” Cahn answered.

For those who want to run the numbers, Cahn arrived at his original $200 billion figure by taking investor estimates for Nvidia’s data-center revenue in the final quarter of 2023 and multiplying it by four to arrive at a run rate of $50 billion. He calculated that Nvidia customers would spend an additional $50 billion on energy, buildings, backup generators and the like. Finally, he assumed that Nvidia users would seek a 50% gross margin on these investments, which would require the infrastructure to drive $200 billion in revenue.

But some investors are now forecasting that Nvidia’s run rate for data-center revenue will reach $150 billion by the end of its fourth-quarter, according to Cahn.

That would imply the infrastructure will need to generate $600 billion in lifetime AI-related revenue.

 

AI is also driving enormous spending on data centers to house the clusters of Nvidia AI systems, which the company describes as robot-assembled supercomputers that weigh 70 pounds and contain 35,000 parts.

The clusters of systems needed to train the new AI models until now could fit into existing data centers. But the coming generation of models are aiming for a 10-fold increase in model size, Cahn said in a blog post, and “to house one of these models, you need to build an entire new data center.”

Construction of new data centers costs time as well as money, so the expected return on these investments won’t appear until late 2025 or early 2026. No wonder investors are nervous. (Nvidia shares fell 9.5% Tuesday amid a broader selloff in chip stocks. Investors continued to chew on last week’s earnings report which, while robust, showed Nvidia’s supercharged growth beginning to slow.)

The bull case

CEOs believe they can make AI pay off despite the staggering costs. 

The network equipment company Nokia is collaborating with Nvidia on projects including developing ways for AI to play a bigger role in the evolution of networks.

AI is helping complex 5G radios become continuously self-learning and self-adjusting in the realm of power consumption management, for example, optimizing for current and expected traffic, Nokia Chief Executive Officer Pekka Lundmark told me.

“This is the first application of AI in the radio network and this is already in production, but there is a lot more to come,” Lundmark said. 

 

Intuit embraced AI shortly after the appointment of CEO Sasan K. Goodarzi in 2019. The financial software company, known for products including TurboTax and QuickBooks proclaimed its strategy to be the creation of an AI-driven “expert platform.”

“We’ve transformed our platform from a place where our customers do the work, to an AI-driven expert platform where AI does the work for them,” Goodarzi said.

AI was a driving force behind Intuit’s plan, announced in July, to cut 10% of its workforce, or 1,800 jobs. It said the goal was to accelerate innovation in critical growth areas such as generative AI, and that it also planned to hire about 1,800 new people primarily in engineering, product and customer-facing roles. 

To justify a sufficient return on the world’s total AI investment, AI must drive revenue on a comparable scale, transforming not just companies but entire industries and ultimately the economy in its wake.

But the extraordinary scale of AI investments already under way across many industries isn’t without precedent. 

 

“As a nation—forget about the world—we pumped trillions of dollars into modernizing the electrical infrastructure, the energy infrastructure in order to create the economy we have today,” said Intuit Chief Data Officer Ashok Srivastava, an adjunct professor in the engineering department at Stanford University and a former chief data scientist at Verizon.

“If AI is a parallel to that, I do envision a lot of money going into revolutionizing the way systems operate. And the companies that really build those new applications are the ones that are going to succeed in the future,” Srivastava said.

Cahn said he believes the $600 billion AI spending challenge will be met, but that it will take time. Ultimately, the value in AI will accrue to companies that connect the underlying technology to end-customer value, according to Cahn.

“The best products will require as much intuition about how customers want to interact with AI as they’ll require technical expertise,” he said.