Global AI infrastructure investment is accelerating at a pace that few analysts fully anticipated, with hyperscaler capital expenditure projections rising steadily throughout 2026.
The four major AI hyperscalers are collectively on track to spend around $650 billion on data center capital expenditures this year, a figure that is already staggering by historical standards.
Nvidia (NASDAQ: NVDA) has projected that total AI spending could reach $1 trillion in 2027, a forecast that aligns with signals already coming from major players in the space.
Alphabet told investors during its Q1 2026 conference call to expect “significantly” higher capital expenditures in 2027, telegraphing an unusually large spending increase well ahead of schedule.
Three companies are best positioned to capture the bulk of that windfall: Nvidia, Broadcom (NASDAQ: AVGO), and Taiwan Semiconductor Manufacturing (NYSE: TSM).
Nvidia remains the dominant force in AI infrastructure, with its GPUs establishing themselves as the industry standard for data center computing power across a wide range of tasks.
Last quarter, Nvidia posted revenue growth of 85%, with next quarter’s revenue expected to nearly double year over year, yet the stock trades at just 22 times forward earnings.
Broadcom is carving out its own space in the AI chip market by partnering directly with hyperscalers to design custom application-specific integrated circuits purpose-built for narrow workloads.
These custom chips are far more cost-effective than GPUs for the specific tasks they handle, and Broadcom expects to generate $100 billion or more in AI semiconductor revenue during 2027 alone.
For context, Broadcom generated $75 billion in total revenue over the past 12 months, meaning AI chip sales alone could exceed the company’s entire recent annual revenue base.
Taiwan Semiconductor Manufacturing operates differently from Nvidia and Broadcom, functioning as a chip manufacturer rather than a designer competing for data center market share.
TSMC benefits regardless of which chip designer wins the AI race, since it manufactures chips for both Nvidia and Broadcom, meaning rising demand across the board drives its growth.
TSMC CEO C.C. Wei stated during the company’s most recent conference call that he sees chip demand staying strong through at least 2029 to 2030, citing the AI build-out as having created an entirely new industry segment.
That long-term visibility makes TSMC a compelling investment not just for the remainder of this year, but across the broader arc of this decade’s AI-driven infrastructure expansion.
