Even fiercer than the internet bubble! Tech giants make a $2 trillion bet on AI, with unprecedented capital intensity.

The wave of AI infrastructure investment is pushing tech giants into an unprecedented heavy-asset cycle. Morgan Stanley’s latest research shows that hyperscalers represented by Amazon, Google, Meta, Microsoft, and Oracle are expected to see capital expenditure intensity comprehensively surpass the historical peak levels of the Internet bubble era. The business model of the technology industry is undergoing structural transformation.

According to Wind Trading Desk, and Morgan Stanley’s report published on February 26, 2026, the bank expects the capex-to-sales ratio for these five hyperscalers to reach 34%, 39%, and 37% in 2026–2028 respectively, surpassing the peak of about 32% reached during the Internet bubble period.

If financing leases are included in the calculation, these ratios will further climb to 38%, 44%, and 45%. Meanwhile, over the next three years, their total capital expenditure will exceed $2 trillion, accounting for roughly 40% of the total capex among the Russell 1000 Index constituents.

However, the explosive growth in capital expenditure has not resulted in a proportional upward revision of revenues. Morgan Stanley points out that in the past six months, consensus expectations for 2026–2027 capital expenditure have been collectively revised up by over $630 billion, but revenue expectations have seen far more limited revisions, causing hyperscalers’ free cash flow (FCF) forecasts to keep trending downward. By contrast, for semiconductor AI-empowered companies, consensus expectations for 2026 sales revenue have been revised up about 60% over the past two years—far exceeding the circa 8% increment for hyperscalers. This makes semiconductor firms the most direct financial beneficiaries in the current round of AI investment.

Capital Intensity Breaks Through Historical Highs Set During the Internet Bubble

Morgan Stanley’s report notes that six months ago, the bank described the AI construction boom as "close to but not surpassing" the peak capital intensity of fiber optic construction during the Internet bubble. The latest forecast now shows capital intensity will "far exceed" the approximately 32% peak of the Internet bubble, with capex-to-sales expected to reach 34%, 39%, and 37% in 2026–2028 respectively.

The report also emphasizes that measuring capital expenditure solely by traditional metrics actually underestimates the size of this investment cycle. Financing leases, being a debt-based method of acquiring assets, should also be included in total investment assessment. These were rarely used during the Internet bubble, but today hyperscalers are signing data center lease agreements worth hundreds of billions. Morgan Stanley’s software industry analysts estimate that leases at Microsoft and Oracle alone are enough to bring the overall hyperscalers’ capex-to-sales ratio to 38%, 44%, and 45% for 2026–2028.

As for the Russell 1000 Index, hyperscalers are expected to account for over 150% of the capex increase in 2025—meaning the rest of the constituents are actually net shrinking in capex. Hyperscalers’ capex is expected to grow by about 70% year-on-year, while that of the rest of the index will drop by 6%. Morgan Stanley forecasts that by 2026, hyperscalers’ contribution to overall Russell 1000 capex will reach about 40%, doubling from 2024, and potentially even rising to 49% by 2028.

Record Capital Expenditure Revisions, Revenue Forecasts Lag Severely

A defining feature of this investment cycle is the unprecedented speed and scale of capital expenditure forecast revisions. Since September 2025, consensus expectations for hyperscalers’ 2026 and 2027 capex have been raised by 1.5 times respectively, while Morgan Stanley’s own analyst forecasts have been raised by up to 1.8 times.

Looking at individual companies, Google’s consensus 2026 capex forecast has been revised up 117% year over year, META up 96%, Amazon up 75%, and Oracle up an astonishing 264%. Morgan Stanley analyst Todd Castagno’s team notes these revisions are "step changes" rather than gradual adjustments, highlighting how difficult this investment cycle is to predict—management teams continually update their data center expansion plans and compete to secure critical supply chains, further increasing uncertainty.

In sharp contrast to the rapid upward revisions of capex expectations, revenues are hardly revised and FCF forecasts are falling. The report shows that over the past year, the collective capex revision for 2026 among these five companies has exceeded $310 billion, but revenue revisions totaled just $130 billion. Morgan Stanley points out that as the base of fixed costs expands, operating leverage will increase, and future profitability and FCF will become far more sensitive to changes in revenue expectations.

Financing Leases Significantly Amplify Actual Investment Scale

Hyperscalers have recently expanded their use of financing leases, further raising the level of actual capital intensity. According to their latest financial reports, the total commitment value of future leases among these five companies has now surpassed $660 billion—Oracle at about $248 billion, Microsoft at $155 billion, META at $104 billion, Amazon at $96 billion, and Google at $59 billion. Notably, Google’s lease commitments have grown almost sevenfold since 2024, and META’s have increased by over 200% in the same period.

The impact of financing leases on the capital intensity of individual companies is particularly prominent. For Microsoft, if only measured by traditional capex, its capex-to-sales for FY26 and FY27 would be about 29%. Including financing leases, these ratios leap to around 43% and 42% respectively. Oracle’s situation is even more extreme—the company is acquiring all data center shells via leases, so its traditional capex-to-sales ratio is expected to be 75% and 119% for FY26 and FY27, while including leases those ratios jump to 107% and 201%, meaning total reinvestment in both fiscal years will exceed total revenue for the year.

Semiconductor Firms Are the Biggest Winners; Hyperscalers Still Need to Prove Returns

Although capital expenditure is highly concentrated among hyperscalers, the group that has clearly benefited the most in this investment cycle is semiconductor AI-empowered companies.

The core reason for this divergence lies in differences in revenue certainty: Hyperscalers make large upfront purchases of GPUs and other chip components, providing chip suppliers with clear near-term revenue sources, while hyperscalers themselves must, over the coming years, monetize these compute assets through large language model monetization, continued compute demand, and product differentiation—posing much greater uncertainty.

Capital market performance also validates this logic. Since December 2023, the share prices of North American semiconductor AI-empowered companies have outperformed hyperscalers and the broader AI sector by 272% and 224% respectively. The market is currently more willing to pay a premium for the confirmed near-term earnings of semiconductor companies, while remaining cautious about revenue realization for hyperscalers and the broader AI cohort.

Morgan Stanley analyst Brian Nowak believes that META, Google, and Amazon are using AI investment, data accumulation, and scale advantages to accelerate user engagement and commercial monetization; Keith Weiss characterizes Oracle’s data center expansion as a potential revenue opportunity, but stresses this requires massive financial backing. The ongoing trend in capex revisions will also trigger sustained increases in depreciation, which, in the absence of synchronized revenue upgrades, will cause significant margin pressure.

~~~~~~~~~~~~~~~~~~~~~~~~

The above excellent content comes from Wind Trading Desk.

For more detailed interpretation, including real-time analysis and frontline research, please join [Wind Trading Desk▪Annual Membership]

Risk Warning and DisclaimerThe market is risky. Investment must be cautious. This article does not constitute individual investment advice, nor does it take into account the special investment objectives, financial conditions, or needs of any individual user. Users should consider whether any opinions, views, or conclusions herein suit their specific circumstances. Invest accordingly at your own risk.