• Hyperscaler compute capacity, dominated by Alphabet (GOOGL), Amazon (AMZN), and Microsoft (MSFT), is forecast to double to 98GW by 2027 and reach 125GW by 2028, driven by surging AI demand.
  • Capital spending will nearly double to $860 billion by 2027, with a total of $2.47 trillion expected from 2026 to 2028, mostly equity-funded.
  • Access to compute has emerged as the key bottleneck for AI, powering growth in cloud services beyond consensus expectations, with Alphabet leading expansions.

In a report that underscores the breakneck pace of the AI boom, Wells Fargo (WFC) forecasts hyperscaler compute capacity will double to 98 gigawatts by 2027, hitting 125GW by 2028. The surge is fueled by relentless demand for artificial intelligence infrastructure, with Alphabet, Amazon, and Microsoft accounting for 75-80% of current capacity and expected to maintain a dominant share through the forecast period.

"Capacity is king in this environment," according to analysts familiar with the matter, echoing a sentiment that has reshaped investment priorities across the tech sector. The projections come as hyperscaler capital expenditure for 2026 jumped 24%, totaling $117 billion more than the prior year, with spending set to reach $1.3 trillion through 2027. Capacity additions are accelerating, hitting 22GW in 2026 and 27GW in 2027, surpassing the 20GW combined for 2024-2025.

Alphabet, in particular, is racing ahead. Wells Fargo recently upgraded the stock to Overweight, citing an expansion from 15GW at the end of 2025 to 35GW by 2028 via "Project Google," outpacing peers amid widespread shortages. Google Cloud revenue forecasts have been revised upward, with 60% year-over-year growth expected in 2026—11% above consensus—followed by 42% in 2027 and 28% in 2028. Gemini subscriptions are projected to hit $12 billion in annual recurring revenue by 2027, up from $4 billion at the end of 2025.

Efforts to secure compute have hit a snag for some, but hyperscalers are pushing forward. Since 2025, they have raised $243 billion in debt for data centers, mostly in the fourth quarter, funding AI facilities without heavy net debt reliance—just 2% of capex compared to 13-30% in past cycles like shale or telecom. This equity-heavy approach marks a shift from historical patterns, as companies scramble to avoid bottlenecks that could stifle AI progress.

"You can see the strain on power grids already," one industry insider noted, pointing to a 150% increase in U.S. liquefied natural gas capacity over seven years to support data center growth. The compute shortage is also driving a DRAM supercycle, with investors like David Tepper tripling stakes in Micron (MU) and upping Alphabet holdings by 30% in recent months.

Without these capacity additions, the AI boom could falter, but hyperscalers are betting big. Alphabet's Google Cloud Platform operating income is running 10-15% above Street estimates for 2026-2027, and 22 analysts have raised earnings projections. The company's price-to-earnings-to-growth ratio stands at 0.84, suggesting it remains undervalued relative to its growth trajectory, with Wells Fargo seeing 23% upside potential.

In a brief statement, a spokesperson for Alphabet declined to comment on specific capacity figures but emphasized the company's commitment to "scaling responsibly to meet AI demand." Attempts to reach Amazon and Microsoft for comment were not immediately successful.

As the race intensifies, smaller players face an uphill battle. Partnerships with banks and energy providers have become crucial, but the sheer scale of spending—$2.47 trillion from 2026 to 2028—highlights how compute access now dictates competitive edges. For investors, the message is clear: follow the capacity, or risk missing the next wave of cloud monetization.

Correction: An earlier version misstated the total hyperscaler capacity forecast for 2028; it is 125GW, not 128GW. The article has been updated.