(Reuters) -Citigroup has raised its forecast for AI-related infrastructure spending by tech giants to surpass $2.8 trillion through 2029, from $2.3 trillion estimated earlier, citing aggressive early investments by hyperscalers and growing enterprise appetite.

The AI boom ignited by ChatGPT's launch in late 2022 has continued to fuel staggering capital outlays and data center expansion despite a brief crisis of confidence sparked by China's cheaper DeepSeek model and lingering market concerns over U.S. President Donald Trump's tariff policies.

The Wall Street brokerage sees AI capex across hyperscalers to reach $490 billion by the end of 2026, up from its earlier estimate of $420 billion.

Data center operators - or hyperscalers - including Microsoft ( MSFT ), Amazon ( AMZN ) and Alphabet have already spent billions of dollars in investments to ease capacity constraints that have hampered their ability to meet surging AI demand.

Citi analysts said hyperscalers were likely to reflect this incremental spend in their third-quarter earnings calls, with the guidance expected to be "building ahead of visible enterprise demand".

Citi estimates global AI compute demand would need 55 gigawatt of new power capacity by 2030, translating to $2.8 trillion in incremental spend, $1.4 trillion in the U.S. alone.

The brokerage said big tech firms are no longer relying only on profits to fund AI infrastructure. The costs are extremely high - about $50 billion for every 1 GW of compute capacity - and the companies are borrowing to keep up.

This shift is already showing up in their financials, with spending starting to eat into free cash flows. Investors are now asking how the tech companies will fund this scale of investment, especially as traditional models fall short.

"Enterprises have provided a clear external validation of value," Citi said, pointing to production deployments at companies such as Eli Lilly, Hitachi and Wolters Kluwer.