COLUMN-AI and gravity-defying US GDP: Mike Dolan

By Mike Dolan

LONDON, July 23 (Reuters) - For the second time in three years - and straddling two separate presidencies - heightened U.S. fears of recession have proved wide of the mark. The artificial intelligence boom is once again the prime suspect.

The AI explosion and arms race since the launch of OpenAI's ChatGPT in November 2022 had its own mini crisis of confidence earlier this year with the emergence of China's cheaper DeepSeek rival. But neither those doubts nor the market angst about President Donald Trump's tariff war seems to have slowed the staggering capital spend and data center building supporting the AI push.

As the brave new world of generative AI enters its third year, the financial impact has expanded well beyond the share price of chipmakers, as data center construction and capex spending more broadly are flattering U.S. GDP to a remarkable degree.

Carlyle's Head Investment Strategist Jason Thomas has been arguing all year that this capex spending is an effective re-industrialization of corporate America, reshaping its focus away from software and intangible assets and toward investing in plants and machinery and energy like never before.

The GDP impact from that is huge.

RE-INDUSTRIALIZE AND BORROW

Thomas reckons the AI-related capex boom – in data centers, graphics processing units, server infrastructure, power and related hardware and applications – already accounts for more than one-third of this year's second-quarter U.S. GDP growth. And he points out that order books in the relevant industries continue to grow at annual rates of more than 40%.

With Q2 GDP data due out next week expected to show annualized growth of 2.5%, that's a sizeable AI tailwind.

The scale of AI-related spending that has already occurred suggests that these are not just finger-in-the-wind estimates.

Spending on physical data center infrastructure alone is up fourfold from 2020 levels, offsetting weakness in other construction categories.

Among AI chip giant Nvidia's NVDA.O top U.S. customers, Thomas writes, capex has grown 1.5 times as fast as revenue. Property, plant and equipment now account for 70% of typical book value for these firms, up from just 20% before the GenAI scramble began.

The big question remains whether AI's contribution to economic growth is now peaking - even if that's difficult to deduce given the hype and optimism around the sector and evidence of robust chip demand growth worldwide.

FINANCING GAP

But Morgan Stanley this week analyzed this question by considering the borrowing needed to finance the ongoing AI infrastructure and data center buildout.

Analysts Vishwanath Tirupattur and Vishwas Patkar forecast roughly $2.9 trillion of additional global data center spending through 2028, comprising $1.6 trillion on chips and servers and $1.3 trillion on real estate, building costs and maintenance.

For context, that suggests an annual global outlay over the next three years that's close to the $950 billion spent by all S&P 500 firms combined last year. While some of this can be funded with cash, Morgan Stanley estimates that a $1.5 trillion financing gap remains, which will have to be raised via debt, likely a mix of loans, bonds, asset-backed securities and private funding.

On top of this, they reckon that investment spending related to data center construction and power generation will add up to 0.4 percentage points to U.S. GDP through this year and next.

Once again, AI is helping frustrate all other macro bets. And megacap quarterly earnings due over the coming weeks will again be examined forensically for signs of a sting in the tail.

The opinions expressed here are those of the author, a columnist for Reuters.

-- Enjoying this column? Check out Reuters Open Interest (ROI), your essential new source for global financial commentary. Follow ROI on LinkedIn. Plus, sign up for my weekday newsletter Morning Bid U.S.